@ids-group-ltd/ids-design-system 0.11.0 → 0.11.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (50) hide show
  1. package/CHANGELOG.md +1368 -1021
  2. package/components.md +49 -26
  3. package/design-system.md +18 -18
  4. package/fesm2022/ids-group-ltd-ids-design-system.mjs +1750 -598
  5. package/fesm2022/ids-group-ltd-ids-design-system.mjs.map +1 -1
  6. package/migrations/0.10.0.json +19 -19
  7. package/migrations/0.10.1.json +20 -7
  8. package/migrations/0.10.2.json +8 -8
  9. package/migrations/0.10.3.json +2 -2
  10. package/migrations/0.11.0.json +9 -54
  11. package/migrations/0.11.1.json +279 -0
  12. package/migrations/0.11.2.json +158 -0
  13. package/migrations/0.2.0.json +17 -5
  14. package/migrations/0.2.2.json +1 -1
  15. package/migrations/0.3.0.json +1 -1
  16. package/migrations/0.4.0.json +1 -1
  17. package/migrations/0.5.1/badge-statuses.md +1 -3
  18. package/migrations/0.5.1.json +1 -1
  19. package/migrations/0.6.3.json +1 -1
  20. package/migrations/0.7.1.json +6 -6
  21. package/migrations/0.8.0.json +5 -5
  22. package/migrations/0.9.0.json +12 -11
  23. package/migrations/0.9.1.json +2 -2
  24. package/migrations/README.md +34 -52
  25. package/package.json +1 -1
  26. package/styles/_dropdown-overlay.scss +64 -23
  27. package/styles/_ds-color.scss +3 -3
  28. package/styles/_field-box.scss +2 -6
  29. package/styles/_field-label.scss +5 -10
  30. package/styles/_field-state.scss +3 -2
  31. package/styles/_focus.scss +0 -2
  32. package/styles/_layout-utils.scss +2 -2
  33. package/styles/_overlay-motion.scss +2 -3
  34. package/styles/_press.scss +3 -5
  35. package/styles/_reset.scss +0 -1
  36. package/styles/_scrollbar.scss +1 -3
  37. package/styles/_status.scss +1 -2
  38. package/styles/_surface-motion.scss +3 -4
  39. package/styles/_text-clip.scss +1 -1
  40. package/styles/_theme-activation.scss +1 -1
  41. package/styles/_tokens-charts.scss +1 -3
  42. package/styles/_tokens.scss +5 -4
  43. package/styles/_typography.scss +2 -2
  44. package/themes/README.md +6 -6
  45. package/themes/_base-palette.scss +2 -2
  46. package/themes/_semantic.scss +2 -2
  47. package/themes/dark/_palette.scss +1 -2
  48. package/themes/light/_palette.scss +1 -1
  49. package/themes/light/_theme.scss +1 -1
  50. package/types/ids-group-ltd-ids-design-system.d.ts +608 -72
package/CHANGELOG.md CHANGED
@@ -8,6 +8,471 @@ While the package is on `0.x`, minor releases may contain breaking changes.
8
8
 
9
9
  ## Unreleased
10
10
 
11
+ ## 0.11.2 — 2026-09-01
12
+
13
+ ### Added
14
+
15
+ - **`PageUp` and `PageDown`.** Ten rows a press, which is the native `<select>` step and the APG
16
+ listbox example’s — a fixed number rather than a measured screenful, so the jump is the same
17
+ wherever the panel happens to be clipped. No picker had these.
18
+
19
+ - **`clear()` on `ds-select` and `ds-multi-select`.** `ds-autocomplete` has had it; the other two
20
+ offered only the in-trigger button, so an app with its own “reset filters” had to write the empty
21
+ value by hand and guess what empty means for that control. It does NOT move focus, deliberately:
22
+ the caller may be navigating away.
23
+
24
+
25
+ - **Panel header and footer slots.** The panel was a closed box between the filter row and the
26
+ options: "3 of 40 selected", a "clear all" button, a hint or a link under the list had nowhere to
27
+ go. `dsPanelHeaderTemplate` and `dsPanelFooterTemplate` fill a strip above and below, with
28
+ `count`, `total` and `query` in context. Both sit OUTSIDE the `listbox`, so nothing you put there
29
+ is counted as an option or breaks a reader's "N of M" position, and neither scrolls with the list —
30
+ the panel clips and the list is the scroller.
31
+
32
+ - **`dsGroupLabelTemplate`.** A group heading rendered its text and nothing else, so a count, a team
33
+ glyph or a second line was impossible. The heading ELEMENT stays the library's — it keeps the id the
34
+ group's `aria-labelledby` points at — and only its contents are yours.
35
+
36
+ - **The option row's geometry is tokens.** `--ds-option-gap`, `--ds-option-padding`,
37
+ `--ds-option-radius` and `--ds-option-row-gap`, plus `--ds-option-bg-color-active` for the
38
+ hover/keyboard fill. A denser panel meant re-declaring the row before, which is exactly how a second
39
+ copy of the panel ended up inside a component.
40
+
41
+ - **`searchPlaceholder` on `ds-autocomplete`.** Its `placeholder` was doing two jobs — the
42
+ empty-field caption and the hint for typing over a settled value. `ds-multi-select` already
43
+ separated them; now the name means one thing on both. Empty by default, so nothing moves.
44
+
45
+ - **The result count is announced.** A reader typed into the filter and heard nothing: no live region
46
+ existed in any picker. One is created per picker now, outside the listbox and empty at first — a
47
+ region created together with its text is the case screen readers reliably do not announce.
48
+
49
+ - **Type-ahead on `ds-select` and `ds-multi-select`.** Typing "ba" jumps to the first option starting
50
+ with "ba", the way a native `<select>` does, and repeating one letter cycles through the options
51
+ beginning with it. On a closed `ds-select` the jump commits, matching the native control.
52
+ `ds-autocomplete` is deliberately excluded: its trigger IS a text field, so typing there is the
53
+ filter and a jump would fight it.
54
+
55
+ - **`disabledReason` on an option, rendered by `ds-select`.** A disabled row said it could not be
56
+ picked and never why. On `ds-select` the reason reaches the row through `aria-describedby` and is
57
+ visually hidden — the row already looks disabled; the text is for whoever cannot see that. The
58
+ field is on the shared `DsOption`, so `SelectOption`, `MultiSelectOption` and `AutocompleteOption`
59
+ all accept it and it type-checks on all three pickers; `ds-multi-select` and `ds-autocomplete` do
60
+ not yet render it, so a reason set there is carried and not shown.
61
+
62
+
63
+ - **The keyboard cursor keeps its row in view.** On a list longer than the panel the arrows walked the
64
+ cursor out of sight — the reader pressed down and could not see where they were. All three pickers
65
+ scroll the active row into view now, on by default; `scrollActiveIntoView="false"` gives the scroll
66
+ position back to you. It uses `block: 'nearest'`, so a row that is already visible does not move and
67
+ the mouse is never fought.
68
+
69
+ - **`Home` and `End` on every picker.** They existed on `ds-select` alone, while ten other components
70
+ in the library already had them. The index arithmetic all three had written out separately — next,
71
+ previous, first, last, skipping disabled and wrapping — is one shared helper now, so the three
72
+ behave identically. It is internal: nothing new is exported.
73
+
74
+ - **`compareWith` on `ds-autocomplete`.** The component is generic over its value type but compared
75
+ with `===`, which is reference identity: a re-fetch that returned equal data as new objects silently
76
+ dropped the selection. Pass a comparator keyed on what identifies the value —
77
+ `[compareWith]="(a, b) => a?.id === b?.id"`. The default is unchanged, so string and number values
78
+ behave exactly as before. `ds-select` and `ds-multi-select` key on `string` and do not need it.
79
+
80
+ - **`loading`.** While options were being fetched the panel said "No matches", which is a lie the
81
+ consumer had no way to correct. It shows a spinner instead now, and the listbox reports
82
+ `aria-busy`. Spelled `booleanAttribute` to match `ds-button`, `ds-input` and `ds-loading-overlay`.
83
+
84
+ - **`opened` and `closed`.** There was no way to know a panel had opened, so a picker whose options
85
+ are fetched on demand had nowhere to start the request. `closed` is the name eight other components
86
+ already use; `opened` is its missing pair. They fire on the transition, not on every write, so
87
+ closing by Escape emits once rather than twice.
88
+
89
+ - **`filterWith` and `filterChange` — one input for both server-side search and a custom match.**
90
+ The query matcher was hardcoded to a case-insensitive substring of `label`: matching a second field,
91
+ ignoring diacritics or scoring fuzzily was impossible, and so was letting a server do the filtering.
92
+ `filterWith` takes a predicate for the first and the literal `'none'` for the second, where every
93
+ option handed in is rendered and `filterChange` reports what was typed. One input rather than two,
94
+ because they are one decision: who decides what the query matches.
95
+
96
+ - **`name`, `readOnly` and `clearable` on `ds-multi-select` and `ds-autocomplete`.** The last three
97
+ form-contract members that were `ds-select`'s alone. `name` reaches seven other controls in the
98
+ library, so the two pickers were the outliers; `readOnly` takes focus without opening, and carries
99
+ the same `readonly` binding alias `ds-select` has for signal forms; `clearable` uses the same button
100
+ and the same accessible name as `ds-select`'s.
101
+
102
+
103
+ - **`ds-table` row expansion: one-at-a-time, expand-all, and events.** The base feature shipped
104
+ already; these are the three edges it was missing. `expandMode="single"` collapses the open row
105
+ when another expander is clicked, so a detail panel cannot stack up down the page — enforced when
106
+ the user clicks, not on the binding, because `expandedIds` is the host's model to write and
107
+ rewriting it underneath would be the table arguing with its own consumer. `expandAllToggle` puts
108
+ a chevron in the expander column header, page-scoped when paginating exactly as select-all is.
109
+ `(rowExpanded)` and `(rowCollapsed)` name the row as it opens and closes, expand-all included,
110
+ which is where panel content gets lazy-loaded; a collapse the host writes into `expandedIds`
111
+ itself stays silent, and so does an id whose row is no longer in `data()` — an event carries a
112
+ row, and there is none to carry. Opening two rows or more now also reaches the table's live
113
+ region, which a full page of panels appearing from one keypress previously did not.
114
+
115
+
116
+ - **Column resizing works on a touch device, where the drag handle never could.** The handle
117
+ paints its line on `:hover` and `:focus-visible`; a coarse pointer produces neither, so it was
118
+ invisible — while `touch-action: none` still took the horizontal swipe that scrolls the table,
119
+ leaving a dead 8px strip at the right edge of every header. Under
120
+ `(hover: none) and (pointer: coarse)` the handle now stands down and width moves into the column
121
+ panel: each row carries −, + and Fit next to its visibility and order controls, and the panel
122
+ opens as a `ds-bottom-sheet` rather than a popover. The buttons render on desktop as well, where
123
+ the handle remains the primary route. All three routes share one step, so drag, arrow keys and
124
+ buttons move a column by the same amount, and all three clamp at the scroll viewport.
125
+
126
+
127
+ ### Changed
128
+
129
+ - **Exported class member surface grew — only matters if you SUBCLASS a DS component.**
130
+ `SelectComponent`, `MultiSelectComponent`, `AutocompleteComponent`, `TableComponent`,
131
+ `LoadingOverlayComponent` and `ToastRegionComponent` all gained members for the inputs and outputs
132
+ above, and `DsPanelHeaderTemplateDirective`,
133
+ `DsPanelFooterTemplateDirective` and `DsGroupLabelTemplateDirective` are new classes.
134
+ Subclassing a DS component is unsupported but consumers do it, and a
135
+ member of yours that shares a name with one of ours stops the build. Rename yours, or add `override`
136
+ if you meant to replace ours — the compiler enumerates every collision, so building after the
137
+ upgrade is the whole check.
138
+
139
+ - **BREAKING for code that CONSTRUCTS a `ToastEntry`: `durationMs` is now a required
140
+ `readonly number`.** The field did not exist on the type before, so it arrives required rather
141
+ than widened: an object literal typed as `ToastEntry` — a test double, a hand-rolled region, a
142
+ store that holds toasts — stops compiling until it carries one. The entry has to know its own
143
+ duration now that the countdown belongs to the region rather than the service.
144
+ `ToastOptions.durationMs` is untouched — still optional, still `5000` by default, still `0` for a
145
+ loading toast — so a caller of `ToastService` writes nothing new.
146
+
147
+ - **BREAKING visually where a stack of toasts is part of the page: the newest toast enters at the
148
+ top of a top-anchored region.** Every toast was appended, so in a region resolving to a top
149
+ position — the default — the newest one landed at the BOTTOM of the stack, furthest from the
150
+ corner the stack grows out of, and the thing that had just happened was the last one you read. A
151
+ top position prepends now; a bottom position still appends. Either way the newest toast is the one
152
+ nearest the screen edge the region is anchored to.
153
+
154
+ - **`ds-select`'s option list caps 40px shorter.** The list's `max-height` was declared in
155
+ `ds-multi-select`'s own stylesheet; it sits in the shared panel partial now, so it reaches
156
+ `ds-select` and `ds-autocomplete` too. `ds-select`'s list was previously bounded only by the panel
157
+ cap `--ds-dropdown-panel-maxh` (320px) and now stops at `--picker-maxh` (280px) — roughly one
158
+ option row — which is the height the other two panels already used. `ds-autocomplete` had pinned
159
+ its panel to that value itself and does not move. Set `--picker-maxh` to take the height back, on
160
+ the panel where only that select should change or globally where all three should follow; the
161
+ panel cap is unchanged and is simply no longer what decides where a `ds-select` list stops.
162
+
163
+ - **`skeleton-mode` is gone from the inner `<table>` element of `ds-table` — only matters if you
164
+ reach in with `::ng-deep`.** The class was bound on that element, not the host, and nothing in the
165
+ library ever painted it, so a rule keyed on it was the only thing giving it meaning. If you have
166
+ one, key it off a class of your own on the host instead; if you do not, nothing about the rendered
167
+ table changes.
168
+
169
+ ### Fixed
170
+ - **Dev-mode warnings no longer ship to production bundles.** They were guarded by
171
+ `isDevMode()`, which is a function CALL — a bundler cannot prove the branch dead, so all fourteen
172
+ warning sites and their message closures reached production builds. Each site is now guarded inline
173
+ with `typeof ngDevMode === 'undefined' || ngDevMode`, which `@angular/build` folds away whenever
174
+ script optimization is on, closures included. **Measured: 1 590 B gzip off an app using most of the
175
+ library, 827 B of it off a single `<ds-button>`.** Nothing changes for you — no API, no behaviour,
176
+ and the warnings still fire in development builds exactly as before.
177
+
178
+
179
+ - **A loading scrim stopped the pointer and let the keyboard walk straight under it.**
180
+ `ds-loading-overlay` paints an absolutely positioned scrim over its content, which intercepts
181
+ every pointer event — so a control behind it could not be clicked, and looked unavailable. Nothing
182
+ stopped focus: Tab reached the same control, Enter activated it, and `ds-table` with
183
+ `rowInteractive` would emit `(rowClicked)` for a row whose data the host was in the middle of
184
+ replacing. Neither `onRowClick` nor `onRowKey` consulted the loading state, and a row's
185
+ `tabindex` never did either. The overlay now marks the content it covers `inert` while loading,
186
+ which is the one mechanism that stops pointer and keyboard together and takes the covered controls
187
+ out of the accessibility tree; `aria-busy` stays on the host, so a reader is still told why.
188
+ `inertContent="false"` turns it off for a scrim that covers only PART of its content, and the
189
+ consumer then marks the covered subtree itself — `ds-table` does this, because its scrim starts
190
+ below the header and re-sorting during a refresh is exactly what puts the scrim up.
191
+
192
+ - **A toast queued behind `maxVisible` expired while it was still waiting.** The auto-dismiss
193
+ countdown started when `ToastService` was called, not when the toast reached the screen, so a
194
+ burst past the cap retired toasts that had never been visible — the more you fired at once, the
195
+ fewer were read, and the ones lost were the ones at the back. The countdown starts when the region
196
+ shows the toast now, promotion out of the queue included, so a queued toast gets its full duration
197
+ on screen. One gap remains: `ToastService.promise()` runs its own timers, so a promise toast held
198
+ behind the cap can still be removed before it is shown.
199
+
200
+ - **Dismissing a toast that was still queued promoted a different one instead of removing it.** The
201
+ id belonged to an entry that had not become visible yet, so the entry stayed in the queue and
202
+ another toast was pulled forward: the one you dismissed still appeared later, and one you had not
203
+ touched was hurried on screen ahead of its turn. A queued id is spliced out of the queue now and
204
+ nothing is promoted in its place.
205
+
206
+ - **A column resize handle announced no position and no limits.** The handle is a `role="separator"`,
207
+ which a screen reader presents as a slider, but an auto-sized column carried no `aria-valuenow` at
208
+ all — the user was handed a control with no current value and no way to hear how far it could go in
209
+ either direction. While the handle holds focus it now carries a measured `aria-valuenow` in px,
210
+ `aria-valuemin` of `48` and an `aria-valuemax` of the scroll viewport's width, which is the same
211
+ ceiling the drag, the arrow keys and the column-panel buttons clamp at. All three are present only
212
+ while the handle is focused.
213
+
214
+ - **A table's loading scrim slid off the table as it scrolled sideways.** `ds-loading-overlay` sat
215
+ inside the scroll container, and the scrim is `position: absolute; inset: 0` — so it was sized to a
216
+ single viewport of the table and then travelled with the content: at 400px of horizontal scroll it
217
+ sat 400px to the left of where it was needed, and every column the scroll had revealed was left
218
+ uncovered. The overlay wraps the scroll container now, so the scrim covers what is on screen at any
219
+ scroll offset.
220
+
221
+ - **A compact trigger showed one chip where several fitted.** `triggerDisplay="count"` and
222
+ `"labels"` make the trigger hug its content, so the width the chip fit measured was decided by the
223
+ chips it was about to cut — each pass shrank its own input and the split settled on one. The fit
224
+ reads the space AVAILABLE now, the trigger's parent, which no cut can move. `ds-autocomplete` gets
225
+ the same correction since both pickers measure through one helper, and a `width="fill"` trigger is
226
+ unaffected because there the two numbers are identical.
227
+
228
+ ## 0.11.1 — 2026-08-31
229
+
230
+ ### Added
231
+
232
+ - **Group helpers, for building and narrowing an options array.** Alongside `DsOptionGroup<T>` —
233
+ the shared group type behind each picker's own name — the package exports the union
234
+ `DsOptionItem<T> = DsOption<T> | DsOptionGroup<T>`, the type guard `isDsOptionGroup<T>(item)`
235
+ and `flattenDsOptions<T>(items)`, which returns the options in source order and stamps
236
+ `disabled: true` onto every option of a disabled group. `isDsOptionGroup` is the narrowing to
237
+ read an `options` array back through, now that any element may be a group.
238
+
239
+ - **A written contract for the picker family.**
240
+ Five rules now hold across the three pickers: one meaning per member name, difference only by a
241
+ stated reason, one panel implementation, no library-invented string without a slot, and one way
242
+ to clip a line. A divergence from them is a defect unless the reason is stated.
243
+
244
+ - **`dsEmptyTemplate` and `dsCreateTemplate`.** "No matches" and "Create «…»" were literals with no
245
+ way to replace them. Both are slots now, on all three pickers and on `ds-autocomplete`
246
+ respectively. `ds-select` also gained an empty state at all — with no options it used to open a
247
+ blank panel.
248
+
249
+ - **`selectAll` takes `'group'` and `'both'`.** A row per group, or a group row beside the
250
+ panel-wide one. A group's row is disabled when the group is, and shows the mixed state when part
251
+ of it is picked.
252
+
253
+ - **`triggerDisplay="labels"`.** Same compact footprint as `count`, but it names the selection
254
+ instead of counting it, cut to the trigger's width with the rest behind "+N" — the same
255
+ measurement the chips already use.
256
+
257
+ - **Parity the family was missing.** `label`, `required` and `optional` on `ds-multi-select` and
258
+ `ds-autocomplete` — those three existed on `ds-select` alone — and `size` on `ds-multi-select`,
259
+ which `ds-select` and `ds-autocomplete` both already had. It
260
+ meant a form built from the other two could not be labelled or marked required at all.
261
+
262
+ - **`searchable` on `ds-multi-select` — one input, three meanings.** `true` always shows the
263
+ panel's filter row (the default, unchanged), `false` never, and a **number** is the threshold at
264
+ which it appears — stated where it applies rather than hidden in a config. Set
265
+ `DS_CONFIG.picker.searchable` to give a whole app one threshold and override it per panel with a
266
+ number.
267
+
268
+ Hiding the row is not just an `@if`: the filter input is what takes focus on open and what carries
269
+ `aria-activedescendant`, the arrows, Enter, Escape and the Backspace that removes the last chip.
270
+ Without it the TRIGGER runs that model instead, and the active-descendant id rides on the trigger.
271
+
272
+ - **A compact filter trigger, and a slot to write your own.** `ds-multi-select` takes
273
+ `triggerDisplay="count"`: a `label` caption plus a selected-count `ds-badge` at intrinsic
274
+ width, with an accent border once something is picked. The chip field it replaces is as wide as
275
+ the control and its content changes with every pick, which is why a dense toolbar could not budget
276
+ space for it. (The chips do not grow the trigger: `maxVisibleChips` defaults to `"fit"` and clips
277
+ behind "+N more".)
278
+
279
+ `dsPickerTrigger` projects your own trigger contents into `ds-select`, `ds-multi-select` and
280
+ `ds-autocomplete`. **A projected template WINS over `triggerDisplay`**, which is then not read at
281
+ all, so there is no "compact plus a template" combination to document. The trigger ELEMENT stays
282
+ library-owned in both cases: `aria-haspopup`, `aria-expanded`, `aria-controls`, the roving
283
+ `aria-activedescendant` and Enter/Space/ArrowDown/Escape are not projected.
284
+
285
+ **Neither `count` nor `labels` is offered on `ds-autocomplete`.** Its trigger holds the text
286
+ input the control exists for, and a caption-plus-count summary would remove it. The slot is there instead, where a
287
+ consumer decides what happens to the input.
288
+
289
+ - **`dsOptionTemplate` and `dsChipTemplate` — the option row and the selected chip take your content.**
290
+ An option was `{ value, label, disabled?, icon? }` and nothing else, so an avatar, a colour dot, a
291
+ priority glyph or a trailing count had no way in. Two directives now fill those two
292
+ places: `dsOptionTemplate` on `ds-select`, `ds-multi-select` and `ds-autocomplete`, and
293
+ `dsChipTemplate` on the two that render chips — so a visual that identifies an option does not
294
+ vanish the moment it is picked.
295
+
296
+ Two directives rather than one keyed one: there are exactly two roles and they differ, so a key
297
+ would be an enum of two. The row and the chip stay library-owned — `role="option"`,
298
+ `aria-selected`, the active-row class, the checkbox, the check glyph and the chip close button are
299
+ not projected.
300
+
301
+ - **`width` on `ds-multi-select` and `ds-select`, the input `ds-autocomplete` already had.**
302
+ `fill` (the default, unchanged) stretches the control to its container and pins the panel to the
303
+ trigger's width. `content` hugs the trigger's own content and takes that width as the panel's
304
+ FLOOR instead, so the panel keeps growing to its options — a short trigger no longer forces a
305
+ panel too narrow to read its own labels. Nothing moves until you ask for `content`.
306
+
307
+ The three pickers now share one `DsControlWidth` behind their own exported names
308
+ (`SelectWidth`, `MultiSelectWidth`, `AutocompleteWidth`), the same way 0.10.0 unified
309
+ `DsOption`. Two new panel-level tokens bound the result in `content` mode:
310
+ `--ds-dropdown-panel-minw` (default `0`) and `--ds-dropdown-panel-maxw` (default `none`).
311
+
312
+ `ds-select` also gained token forwarding onto its portaled pane, which it did not have: without
313
+ it a `ds-select { --ds-dropdown-panel-*: … }` never reached the panel, because CDK portals it out
314
+ of the component.
315
+
316
+ ### Changed
317
+
318
+ - **One stylesheet owns the dropdown panel.** The panel is portaled to `document.body`, where
319
+ Angular's encapsulation does not reach, so the shared partial and two component stylesheets were
320
+ all declaring its `background`, `border-radius`, `max-height`, `overflow` and `display` at
321
+ identical specificity — which rule won depended on the order the stylesheets happened to load. The
322
+ partial owns that geometry now, behind `--ds-dropdown-panel-bg-color` beside the `-radius`,
323
+ `-maxh`, `-minw` and `-maxw` knobs, and each picker declares its own name as an ALIAS of the
324
+ shared token on its host and forwards it onto the pane. `--ds-multi-select-panel-bg-color`,
325
+ `--ds-autocomplete-panel-bg-color`, `-radius` and `-maxh` all still work; what changed is that
326
+ they reach the panel through one rule instead of a second competing one.
327
+
328
+ The panel is also the flex column and the LIST is the scroller, so a filter row or a select-all row
329
+ stays put while the options move under it. `ds-multi-select` did that per-component before and the
330
+ other two did not.
331
+
332
+
333
+ - **BREAKING for code that READS `ds-autocomplete`'s `options`: the element type is now
334
+ `AutocompleteItem<T>`, not `AutocompleteOption<T>`.** The input holds flat options OR groups,
335
+ so a reader — or a subclass overriding a member typed against the element — that assumed
336
+ every element has `.value` must narrow with `isDsOptionGroup` first. Passing plain options in
337
+ is unchanged.
338
+
339
+ - **BREAKING for code that READS `ds-multi-select`'s `options`: the element type is now
340
+ `MultiSelectItem`, not `MultiSelectOption`.** The same narrowing: read `.value` only after
341
+ `isDsOptionGroup` has ruled out a group. Writers passing options in are unaffected.
342
+
343
+ - **BREAKING for code that READS `ds-multi-select`'s `selectAll` into a `boolean`: the type is
344
+ now `boolean | 'group' | 'both'`.** The default is still `false`, so setting it is unchanged;
345
+ assigning it back into a `boolean` stops compiling. The two new values are under Added.
346
+
347
+ - **BREAKING for CSS overrides: `ds-multi-select`'s panel is the shared dropdown panel too.** `ds-autocomplete` rejoined it
348
+ earlier in this release; the multi-select still carried its own `.panel` / `.options` / `.option`
349
+ copy while its pane ALSO carried `ds-dropdown-panel`, so both rule sets applied at once. One
350
+ implementation now: the class names are `.dropdown-panel` and `.dropdown-list`, and the row is the
351
+ partial's. What stayed is what no other picker has — the filter row, the select-all rows and this
352
+ component's own tokens. Its option row carries only `[disabled]` where `ds-select` sets
353
+ `.is-disabled` beside it, which is why it still declares an `.option:disabled` of its own.
354
+
355
+ - **BREAKING if you set it: `triggerDisplay="compact"` is now `triggerDisplay="count"`.** The name
356
+ says what the trigger shows, which matters once `labels` sits beside it. The union is exported as
357
+ `DsTriggerDisplay` — `'full' | 'count' | 'labels'` — so a variable typed against it fails to
358
+ compile rather than passing a dead string.
359
+
360
+ - **Every single-line label clips through one mixin now.** Seven rules hand-rolled
361
+ `overflow: hidden; text-overflow: ellipsis; white-space: nowrap` — the `ds-select` trigger text,
362
+ the `ds-tree` label, the `ds-sidenav-item` label, three labels in `ds-sidenav-footer`, and the
363
+ table cell in virtual mode. They all use `single-line-ellipsis` instead, which clips with
364
+ `overflow: clip` and an `overflow-clip-margin` rather than `hidden`: where the line box is tighter
365
+ than the glyph box, `hidden` eats the glyph and shaves every label, not only the ones with
366
+ descenders. Nothing to do on your side — labels stop losing their bottoms at tight line-heights.
367
+
368
+ - **Exported class member surface grew — only matters if you SUBCLASS a DS component.**
369
+ `SelectComponent`, `MultiSelectComponent` and `AutocompleteComponent` gained members, and
370
+ `DsOptionTemplateDirective`, `DsChipTemplateDirective`, `DsPickerTriggerDirective`,
371
+ `DsEmptyTemplateDirective` and `DsCreateTemplateDirective` are new classes. Subclassing is unsupported but consumers do it, and a member of yours that shares a name
372
+ with one of ours stops the build. Rename yours, or add `override` if you meant to replace ours —
373
+ the compiler enumerates every collision, so building after the upgrade is the whole check.
374
+ `AutocompleteComponent` also LOST a member — the protected computed `filteredGroups` — and that
375
+ loss fails silently: a subclass that overrode it now overrides nothing and still compiles. It is
376
+ the one case the build will not flag, so check for it by hand.
377
+
378
+ - **BREAKING for `ds-autocomplete`: one `options` input takes groups on all three pickers.**
379
+ `ds-multi-select` had no groups at
380
+ all; `ds-select` took them mixed into `options`; `ds-autocomplete` took them in a second
381
+ `optionGroups` input whose own documentation said `options` was **ignored** when it was set —
382
+ two inputs where one silently won. There is one shape now: `options` accepts flat options,
383
+ groups, or both mixed, and renders them in the order given.
384
+
385
+ The group type is shared as `DsOptionGroup` behind each picker's own name
386
+ (`SelectOptionGroup`, `MultiSelectOptionGroup`, `AutocompleteOptionGroup`), the way
387
+ `DsOption` already was. It carries `disabled`, which `ds-select` had and the other two did
388
+ not; disabling a group disables every option in it.
389
+
390
+ - **BREAKING for CSS overrides and one token: `ds-autocomplete`'s panel is the shared dropdown
391
+ panel again.** It was passing
392
+ `'ds-overlay-pane'` as its overlay's panel class, and that argument REPLACES the default
393
+ `['ds-dropdown-panel', 'ds-overlay-pane']` rather than adding to it — so nothing from the shared
394
+ partial reached it. The panel had `padding: 0` against the other two pickers' `--space-1`, square
395
+ rows against their `--ds-radius-inner`, and none of `--ds-option-selected-bg-color`,
396
+ `-label-color`, `-bg-color-hover` or `--ds-option-check-color` reached it — set on an
397
+ autocomplete they did nothing at all. One partial paints all three pickers now.
398
+
399
+ The panel's classes moved to the shared vocabulary: `.autocomplete-list` → `.dropdown-list`
400
+ inside a new `.dropdown-panel` wrapper, `.autocomplete-optgroup` → `.optgroup`,
401
+ `.autocomplete-optgroup-label` → `.optgroup-label`, and an option's check icon →
402
+ `.option-check`. `--ds-autocomplete-option-bg-color-hover` is gone; the panel-level
403
+ `--ds-option-*` hooks replace it. They paint `.option.is-selected` and the check glyph, so they
404
+ reach `ds-select` and `ds-autocomplete`; `ds-multi-select` marks its selection with a checkbox and
405
+ keeps its own `--ds-multi-select-check-*`. All three forward the four hooks onto the portaled pane,
406
+ so a per-instance `ds-autocomplete { --ds-option-selected-bg-color: … }` reaches the panel.
407
+ `--ds-autocomplete-panel-bg-color`, `-radius` and `-maxh` still work.
408
+
409
+ - **A settled single value in `ds-autocomplete` ellipsizes instead of wrapping.** A value 134px wide in a
410
+ 144px trigger took the trigger from **40px to 63px** — the text
411
+ broke onto a second line and the chevron onto a third. Single mode has no chip row to wrap for, so
412
+ the wrapping had nothing to protect. The ellipsized value carries a `title`, so the full text is
413
+ still there on hover. `valueOverflow="wrap"` restores wrapping; multi mode is untouched,
414
+ its chips already had a measured cap.
415
+
416
+ ### Fixed
417
+
418
+ - **A compact trigger showed one chip where several fitted.** `triggerDisplay="count"` and
419
+ `"labels"` make the trigger hug its content, so the width the chip fit measured was decided by the
420
+ chips it was about to cut — each pass shrank its own input and the split settled on one. The fit
421
+ reads the space AVAILABLE now (the trigger's parent), which no cut can move. `ds-autocomplete` gets
422
+ the same correction, since both pickers measure through one helper, and a `width="fill"` trigger is
423
+ unaffected because there the two numbers are identical.
424
+
425
+
426
+ - **With the filter row hidden, there was no keyboard way to pick an option.** `Enter` and `Space`
427
+ were consumed by the trigger before the panel's own handler saw them, so the panel toggled shut
428
+ instead of selecting the active row: open, arrow down, Enter, nothing chosen. The trigger delegates
429
+ first now. `searchable` shipped in this same release, so only work in flight is affected.
430
+
431
+ - **An option inside a `disabled` group could not be addressed.** The pickers marked such options
432
+ disabled for rendering, and the flattening helper marked them again — a second copy, so the object
433
+ the template rendered and the object in the flat list were no longer the same one. Every row in the
434
+ group then rendered `id="…-opt--1"`, duplicated across the group, and `aria-activedescendant`
435
+ pointed at an element that did not exist. In `ds-autocomplete`, where a fully disabled group also
436
+ leaves the cursor at `-1`, every row in it painted as the active row at once.
437
+
438
+ - **`selectAll="group"` also rendered the panel-wide row**, making it indistinguishable from
439
+ `"both"`, and rendered that row even on a flat list with no groups at all.
440
+
441
+ - **`dsChipTemplate` did nothing on `ds-autocomplete`.** The slot was declared, exported and
442
+ documented; the template never read it, so a projected chip was discarded without a word.
443
+
444
+ - **`ds-autocomplete`'s `label` was not the accessible name and `required` was not announced.**
445
+ The label rendered an `id` nothing referenced while `aria-label` was set unconditionally from the
446
+ placeholder, and `aria-required` was never emitted despite the input's own documentation. Both
447
+ now match `ds-select` and `ds-multi-select`.
448
+
449
+ - **`ds-multi-select` keyboard and focus behaviour joined the other two.** Its option rows were tab
450
+ stops (a fifty-option panel was fifty of them), `Tab` did not close the panel, the initial active
451
+ row did not skip a disabled first option, and removing a chip dropped focus to `<body>`.
452
+
453
+ - **A bare `searchable` attribute meant "unset" instead of "true"**, so it could not override an
454
+ app-level `DS_CONFIG.picker.searchable` of `false` — the one case that key exists for.
455
+
456
+ - **A compact trigger no longer pins the panel to its own caption width**, which produced a panel too
457
+ narrow to read its options unless `width="content"` was also passed.
458
+
459
+ - **Re-opening a picker inside its close animation left the panel visible while the component
460
+ believed it was closed**, and leaked that panel's listeners. The detach handler now ignores an
461
+ overlay that is no longer the current one.
462
+
463
+ - **`ds-select` accepts a `readonly` options array**, which the other two already did, and tracks
464
+ its top-level items by index rather than by label, which is not unique.
465
+
466
+
467
+ ### What you can now DELETE
468
+
469
+ None of this is required — skipping it leaves working code that is merely redundant. Each row is a
470
+ workaround this release made unnecessary. Generated from the `deleteList` in the release manifest
471
+ this package ships under `migrations/`.
472
+
473
+ | If you have | Delete it, because |
474
+ | ----------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
475
+ | `.ds-overlay-pane .autocomplete-list`, `padding`, `border-radius` | The autocomplete panel shipped with `padding: 0` and square rows while ds-select's dropdown had `--space-1` padding and rounded rows, so consumers patched the difference globally. The panel now uses the same shared partial as the other two pickers and matches without help. Replace with: nothing — the padding is `--space-1` and the row radius is `--ds-radius-inner` out of the box. |
11
476
  ## 0.11.0 — 2026-08-26
12
477
 
13
478
  ### Added
@@ -15,18 +480,16 @@ While the package is on `0.x`, minor releases may contain breaking changes.
15
480
  - **`ds-checkbox` and `ds-radio-group` now show a validation state on the CONTROL, not only in the
16
481
  message under it.** Both already rendered `ds-field-message` and already took `hint`/`error`/`warning`;
17
482
  what was missing was the paint, so on a checkbox the only sign of an invalid state was the red line
18
- underneath — which `color-not-only` says is not enough on its own.
483
+ underneath — which WCAG 1.4.1 says is not enough on its own.
19
484
  **Two treatments, because the shapes differ.** The checkbox moves its **border** and nothing else: at
20
485
  16px a tinted fill reads as "checked", so the shared field mixin — built for a 39px box with a
21
486
  background — could not be reused. The radio group colours its **label**: when a required group is
22
487
  unanswered no single radio is wrong, so reddening every ring would point five fingers at one
23
488
  unanswered question. Both also carry `warning`.
24
- **Deliberately NOT extended to `ds-radio`, `ds-switch` or `ds-slider`**, and the reasoning is on the
25
- record rather than left to taste: a lone radio's choice belongs to its group, a switch applies on flip
26
- rather than on submit, and a slider always holds a value inside bounds the control enforces itself —
27
- "required" is not even expressible there. Counted across 39 usages in one consumer's three apps, the
28
- only `[invalid]` binding in the whole codebase is on a radio group; every switch there carries a hint
29
- and no error, and the slider is unused. Those three keep their message line, which is what they use.
489
+ **Deliberately NOT extended to `ds-radio`, `ds-switch` or `ds-slider`**: a lone radio's choice belongs
490
+ to its group, a switch applies on flip rather than on submit, and a slider always holds a value inside
491
+ bounds the control enforces itself — "required" is not even expressible there. Those three keep their
492
+ message line under the control instead.
30
493
 
31
494
  - **`ds-sidenav` `brandRoute` — the brand row can lead home.** Set it and the mark plus wordmark become
32
495
  one link; leave it unset and the row stays the plain block it was, because a logo that goes nowhere
@@ -37,22 +500,20 @@ While the package is on `0.x`, minor releases may contain breaking changes.
37
500
 
38
501
  - **`ds-button` can say WHY it is refused, without a tooltip.** A greyed button that does not explain
39
502
  itself is a dead end, and the only way to explain one was `[dsTooltip]` — reachable by mouse alone.
40
- One consumer carried that workaround on three buttons; two sibling cases on `ds-menu-item` already
41
- went away in 0.10.2 when it gained `[description]`. Which input to reach for depends on whether the
42
- user can lift the refusal on this screen:
503
+ `ds-menu-item` gained `[description]` in 0.10.2 for the same purpose. Which input to reach for
504
+ depends on whether the user can lift the refusal on this screen:
43
505
  - **Standing refusal** — no permission, no integration, nothing here will change it. Set
44
506
  **`disabledReason`** beside `disabled` and the reason renders as a second line INSIDE the button.
45
507
  Setting it also flips HOW the button is disabled: **`aria-disabled="true"` instead of the native
46
- attribute**, because a natively disabled button never takes focus and its description is therefore
47
- never announced (measured: `document.activeElement` never becomes it, Tab skips it). The click is
48
- swallowed in the handler instead, so **`(pressed)` still never fires** — verified against a live
49
- control that does fire. Chrome's own accessibility tree reads name `"Approve invoice"`, description
50
- `"Needs the Finance role"`, state disabled: the reason is the DESCRIPTION, not part of the name.
51
- **Length is deliberately uncapped** — the text wraps and the button grows rather than truncating the
52
- one thing that explains the refusal.
508
+ attribute**, because a natively disabled button never takes focus, so its description would
509
+ otherwise never be announced. The click is swallowed in the handler instead, so **`(pressed)` still
510
+ never fires**. The reason is exposed as the button's accessible DESCRIPTION rather than folded into
511
+ its NAME, so a screen reader announces the button's label first and the reason as a separate line
512
+ after it. **Length is deliberately uncapped** — the text wraps and the button grows rather than
513
+ truncating the one thing that explains the refusal.
53
514
  - **Temporary refusal** — the form simply is not filled in yet. Do NOT disable the button: leave it
54
515
  live, judge validity in the submit handler, and carry the text UNDER it with **`hint` / `warning` /
55
- `error`** — the same three inputs the fields take, resolved by the same `resolveFieldMessage` and
516
+ `error`** — the same three inputs the fields take, resolved the same way and
56
517
  rendered by the same `ds-field-message`, with **`messageIcon`**, **`reserveMessage`** and the same
57
518
  `DS_CONFIG.field.*` fallbacks. `reserveMessage` holds the line's height while it is empty, so a
58
519
  button that gains an error mid-form does not shove the layout below it.
@@ -78,8 +539,8 @@ While the package is on `0.x`, minor releases may contain breaking changes.
78
539
  - **`--ds-card-overflow`, so a card can hold a full-bleed child inside its own radius.** A child whose
79
540
  background runs to the card's edge — an image, a grey footer strip — paints its square corners over
80
541
  the card's rounded ones, and the host cannot fix it: the rounded box is `ds-card`'s inner element, out
81
- of reach. Five screens in one consumer's portal carried the same
82
- `:host ::ng-deep ds-card .card { overflow: hidden }` to get at it; that rule can go.
542
+ of reach. The usual workaround was `:host ::ng-deep ds-card .card { overflow: hidden }`; that rule
543
+ can go.
83
544
  **Set `--ds-card-overflow: clip`** — `clip` rather than `hidden`, so the card does not also become a
84
545
  scroll container.
85
546
  **Opt-in, not the new default.** Clipping every card would cut the focus ring off any child sitting at
@@ -92,9 +553,9 @@ While the package is on `0.x`, minor releases may contain breaking changes.
92
553
  only way in was clicking the row: `rowInteractive` gives the `<tr>` a tabindex and Enter/Space, but a
93
554
  row has no href, so it never enters the tab order **as a link**, a screen reader announces a table
94
555
  row, and Ctrl/Cmd-click cannot open it in a new tab. `interactive: true` only tinted the cell while
95
- the row was hovered — invisible to a keyboard or touch user. Six lists in one consumer's portal
96
- compensated by repeating the destination as a `View Details` item in the row's overflow menu, which
97
- is a menu entry performing navigation rather than an action.
556
+ the row was hovered — invisible to a keyboard or touch user. A common workaround repeated the
557
+ destination as a `View Details` item in the row's overflow menu, which is a menu entry performing
558
+ navigation rather than an action.
98
559
  `link?: (row) => string | readonly unknown[] | null | undefined` returns an Angular route target and the column's
99
560
  value renders as a real `<a>`. **`null` for a row leaves that cell plain text**, so a list where only
100
561
  some records are reachable needs no second column. External URLs go through a `[dsCellTemplate]` —
@@ -109,49 +570,42 @@ While the package is on `0.x`, minor releases may contain breaking changes.
109
570
  ### Changed
110
571
 
111
572
  - **Every migration item now carries `action`, and the reasoning has moved out of its way.** The
112
- manifests in `dist/ds/migrations/` — the ones your own upgrade tooling reads — gained three prose
573
+ manifests the package ships in `migrations/` — the ones your own upgrade tooling reads — gained three prose
113
574
  fields per item: `action` is the imperative alone and the only field to execute, `appliesWhen` is
114
575
  the boundary (which components, which bindings, and what is explicitly NOT affected), and
115
576
  `pitfall` is the way to apply the item mechanically correctly and still ship a broken app. `note`
116
- keeps what is left, the WHY. All **189 items across 22 published manifests** were rewritten by
117
- hand for this: note prose fell from 12,547 words to 4,263, the **103 ALL-CAPS markers** that had
118
- been standing in for the missing fields are gone, and no note runs past 120 words. Four new lint
119
- rules hold it — `action` required (M039), no marker in a note (M040), the 120-word ceiling (M041)
120
- and a key allowlist so a typo cannot sit unread (M042) — and `tools/migration-lint-selftest.mjs`
121
- mutation-tests all 34 rules to prove none of them is dead. **If you read these files with your own
122
- tooling, read `action` first and `pitfall` before you report an item done.** Nothing was removed:
123
- every field that was there is still there, with the same name and meaning.
577
+ keeps what is left, the WHY. **Every item in every published manifest carries the new shape**, and
578
+ the ALL-CAPS markers that used to stand in for the missing fields — `NO ACTION`, `ACTION only if`,
579
+ `DELETE-LIST` — are gone from `note`: if your tooling matched on those, match on `action` and
580
+ `deleteList` instead. **Read `action` first and `pitfall` before you report an item done.** No
581
+ field was removed: every one that was there is still there, with the same name and meaning.
124
582
  - **If you EXTEND a DS field, its message inputs now live one class further up.** `hint`, `error`,
125
- `warning`, `messageIcon`, `reserveMessage` and `reserveMessageResolved` moved out of `DsFieldBase`
126
- into a new `DsMessageBase`, which `DsFieldBase` extends — so `ds-button` can take the message half
583
+ `warning`, `messageIcon`, `reserveMessage` and `reserveMessageResolved` moved out of the internal `DsFieldBase` into a new internal `DsMessageBase`, which `DsFieldBase` extends — so `ds-button` can take the message half
127
584
  without also inheriting `errors`, `invalid` and `touched`, which describe a form control's validation
128
- state and mean nothing on a button. **`DsFieldBase`'s own surface is unchanged**: every one of those
129
- members still resolves through it, with the same names, types and defaults. Called out only because
130
- `npm run class-surface` cannot see a member that moved to a base class — it reports members per class
131
- and this is the case it warns about. A subclass redeclaring one of these collides exactly as it did
585
+ state and mean nothing on a button. **That internal base's own surface is unchanged**: every one of those
586
+ members still resolves through it, with the same names, types and defaults. A subclass redeclaring one of these collides exactly as it did
132
587
  before; subclassing a DS component stays unsupported (`components.md` § Components).
133
588
 
134
- - **If you EXTEND a DS component, six of them grew members.** `npm run class-surface` against
135
- 0.10.3: `ButtonComponent` gained `disabledReason`, `ariaDescribedBy`, `softDisabled`, `message`,
589
+ - **If you EXTEND a DS component, six of them grew members.** Against 0.10.3: `ButtonComponent` gained `disabledReason`, `ariaDescribedBy`, `softDisabled`, `message`,
136
590
  `fieldMessage`, `hasMessageRow`, `describedBy`, `hintId`, `reasonId` and `uid`;
137
591
  `IconButtonComponent` gained `ariaDescribedBy`; `TableComponent` grew one member, `cellLink`, which
138
592
  backs the new `link` column by resolving a column’s `link(row)` to the value the template binds;
139
593
  `SidenavComponent` gained `brandRoute`, `brandLinkLabel` and the protected `brandAriaLabel`; and
140
594
  `CheckboxComponent` and `RadioGroupComponent` each gained the protected `controlState`.
141
- A subclass that happens to declare the same name now fails with
142
- TS4114 (“must have an ‘override’ modifier”), or TS2415 if the signature differs — and `protected` is
143
- no protection, since a redeclared protected member collides exactly as a public one does.
595
+ A subclass that happens to declare the same name now fails with TS4114 (“must have an ‘override’
596
+ modifier”), or TS2415 if the signature differs — and `protected` is no protection, since a
597
+ redeclared protected member collides exactly as a public one does.
144
598
  Subclassing a DS component stays unsupported (`components.md` § Components); compose instead.
145
599
 
146
600
  ### Fixed
147
601
 
148
602
  - **A `ds-sidenav-item` with no destination was `<a href="#">`, so clicking it navigated.** With neither
149
603
  `routerLink` nor `href` the template fell back to `href="#"` — which rewrites the URL and throws the
150
- scroll position away, on every click. That is what made a sidenav SPECIMEN move the docs page.
604
+ scroll position away, on every click.
151
605
  It renders a `<button type="button">` now: an item that only emits `activate` is a button, not a link.
152
- Verified on the built docs — the element is a `BUTTON`, carries no `href`, and a click leaves both the
153
- URL and the scroll position untouched. **Styling is unaffected** (the rules are class-based, and the
154
- UA's button chrome is reset on `.item`), and both `<a>` variants are unchanged.
606
+ The element carries no `href`, and a click leaves both the URL and the scroll position untouched.
607
+ **Styling is unaffected** (the rules are class-based, and the UA's button chrome is reset on `.item`),
608
+ and both `<a>` variants are unchanged.
155
609
  **If you relied on the anchor**: an item you want to be a link needs `href` or `routerLink`, which it
156
610
  always did — `href=""` no longer silently becomes `#`.
157
611
 
@@ -159,59 +613,49 @@ While the package is on `0.x`, minor releases may contain breaking changes.
159
613
  border.** The footer carried `padding-top` alone — no horizontal, no bottom — while every cell is
160
614
  inset by `--ds-table-cell-pad-x`. That padding is what insets the content of a `padding: 0` container,
161
615
  which is how tables are normally placed, so the summary line landed **16px left of the column text**
162
- with its baseline **on the border**. Measured before: footer padding `12px 0 0 0`, footer content at
163
- 0 against the cells' 16. After: `12px 16px 12px`, footer content at 16 — aligned — and the summary
164
- 21px clear of the bottom edge. **It follows `--ds-table-cell-pad-x` with the cells**, verified: set
165
- the hook to 40px and both move together, so a retuned table keeps its footer in line.
616
+ with its baseline **on the border** — footer padding was `12px 0 0 0`. It is now `12px 16px 12px`,
617
+ matching the cells' inset, with the summary clear of the bottom edge. **It follows `--ds-table-cell-pad-x` with the
618
+ cells**, so retuning that token moves the footer's inset along with the columns'.
166
619
 
167
620
  - **`ds-banner`'s body had no size of its own, so a 16px page rendered it LARGER than its own 14px
168
621
  title** — and in `compact` the title and body ran together with nothing between them. Both came from
169
622
  the same place: the gap lived on the TITLE, where it also fired with no body under it and could not
170
623
  fire at all in `compact` (which zeroes that margin). The gap now belongs to the body — it appears only
171
624
  when a body is actually projected — and the body takes `--font-size-s`, overridable through the new
172
- **`--ds-banner-body-font-size`**. Measured: title 14px and body 14px in a 16px context (was 16px);
173
- 4px between title and body in `compact` (was 0px); a title-only `compact` banner unchanged, because
174
- the wrapper is `:empty` and contributes nothing.
175
- **The "compact has no body" line in the docs was never true** — nothing was ever hidden, and 17 of
176
- one consumer's compact banners project one. It is a supported pairing now rather than a claim.
625
+ **`--ds-banner-body-font-size`**. Title and body both render at 14px in a 16px context, with 4px
626
+ between them in `compact`; a title-only `compact` banner is unchanged, because the wrapper is
627
+ `:empty` and contributes nothing. A body inside a `compact` banner is a supported pairing.
177
628
 
178
629
  - **A `compact` banner whose title was longer than its column overflowed the column and cut the text
179
630
  with nothing to show it had been cut.** `compact` is the one-line strip, so its title carries
180
631
  `white-space: nowrap` — but nothing made it DEGRADE: there was no `text-overflow`, and the host had
181
632
  no `min-width: 0`, so as a flex or grid ROW item it refused to shrink below its own min-content.
182
- Measured: a **260px column held a 319px banner**, with the title clipped mid-word and no ellipsis.
183
- The title now ellipsises and the host can be squeezed — re-measured in the same 260px column, the
184
- banner is 227px and the title ends in `…`. A non-compact banner is unaffected: its title still wraps
185
- (two lines at 260px) rather than clipping, which is the difference between the two variants.
186
- **This is what was behind the "crooked banner" report**, not the title alignment it was filed as —
187
- that part did not reproduce, and is covered by the alignment entry above.
188
-
189
- - **`ds-banner` no longer inherits its alignment from whatever it is dropped into.** Reported as "the
190
- title is centred while the body stays at start" (mm-broker). The DS does not centre the title — measured
191
- across all twelve banners in the docs, in both layouts, title and body both compute `start` and there
192
- is no `text-align` anywhere in the component. What IS true is the underlying defect: the banner owned
193
- no alignment at all, so a centred ancestor dragged **both** of its text blocks along (measured: both
194
- flip to `center` under a centred parent). It now sets `text-align: start` on its own box, so a banner
195
- reads as one alignment wherever it sits. If you were centring a banner deliberately by centring its
196
- container, that stops working — say so and it becomes an input.
633
+ A 260px column held a **319px** banner with the title clipped mid-word; the title now ellipsises and
634
+ the host can be squeezed, fitting at 227px instead of overflowing.
635
+ A non-compact banner is unaffected: its title still wraps rather than clipping, which is the
636
+ difference between the two variants.
637
+
638
+ - **`ds-banner` no longer inherits its alignment from whatever it is dropped into.** The component
639
+ itself never set an alignment, so a centred ancestor dragged both the title and the body to `center`
640
+ along with it. It now sets `text-align: start` on its own box, so a banner reads as one alignment
641
+ wherever it sits. If you were centring a banner deliberately by centring its container, that stops
642
+ working — say so and it becomes an input.
197
643
 
198
644
  - **`ds-sidenav-footer`'s account menu now uses `ds-menu-divider`, so the divider tokens reach it.** It
199
645
  hand-rolled a `div` with local styles instead, which meant `--ds-menu-divider-color` and
200
646
  `--ds-menu-divider-spacing` did nothing there — an override written against the documented tokens
201
- silently skipped the one menu the DS ships itself. **The line also sat 6px from its neighbours where
202
- every other menu puts it 2px away**, because the hand-rolled rule added `--space-1` of its own on top
203
- of the menu's gap. It now behaves like the component it should always have been: **the divider adds
204
- no space of its own** — measured 2px above and 2px below the line, which is exactly the gap between
205
- any two ordinary items, so the line takes one row's place instead of pushing the rows apart. Reported
206
- as UNEVEN spacing; that part did not reproduce — the gaps were symmetric both before and after. The
207
- defect was that there was extra spacing at all.
647
+ silently skipped the one menu the DS ships itself. **The line also sat 6px from its neighbours
648
+ where every other menu puts it 2px away**, because the hand-rolled rule added `--space-1` of its
649
+ own on top of the menu's gap. It now behaves like the component it should always have been: **the
650
+ divider adds no space of its own** — 2px above and 2px below the line, exactly the gap between any
651
+ two ordinary items, so the line takes one row's place instead of pushing the rows apart.
208
652
 
209
653
  - **`ds-menu-divider` sits FLUSH now — a boundary is not a row, so it takes no space of its own.** Its
210
654
  own padding already defaulted to 0 (0.8.0), but `ds-menu` is a flex column and its row gap still
211
655
  landed above AND below the line, so every divider sat one row's width away from the items it
212
- separates. It now cancels that gap: **measured 0px above the line and 0px below**, while two ordinary
213
- items keep their 2px. `--ds-menu-divider-spacing` is untouched and still the way to add air —
214
- verified: `8px` gives exactly 8px each side, and clearing it returns to flush. A divider used outside
656
+ separates. It now cancels that gap: **0px above the line and 0px below**, while two ordinary items
657
+ keep their 2px. `--ds-menu-divider-spacing` is untouched and still the way to add air — set it to
658
+ `8px` and you get exactly 8px each side, and clearing it returns to flush. A divider used outside
215
659
  a `ds-menu` is unaffected (the cancel falls back to `0px`).
216
660
  **If you set `--ds-menu-divider-spacing`** — including via 0.8.0's restore recipe — **your boundary
217
661
  tightens by the menu's row gap**, because that gap no longer stacks on top of your value. Add it back
@@ -221,14 +665,13 @@ While the package is on `0.x`, minor releases may contain breaking changes.
221
665
  `/// daring.lion.race`; a what3words identifier is `///daring.lion.race`, flush, which is what the
222
666
  component's own `aria-label` has said all along — so the screen and the accessible name disagreed.
223
667
  The cause was the template's line breaks: the prefix sat on its own line inside the mark element, and
224
- in inline flow that leading/trailing whitespace collapses to a rendered space. Measured at the default
225
- size, the pill was **7.8px wider** than the format allows. The mark row is a flex container now, which
226
- drops whitespace-only text nodes outright, so the identifier renders flush and the template stays free
227
- to be formatted for reading. No API change, and nothing to do on your side.
668
+ in inline flow that leading/trailing whitespace collapses to a rendered space. The mark row is a flex
669
+ container now, which drops whitespace-only text nodes outright, so the identifier renders flush and
670
+ the template stays free to be formatted for reading. No API change, and nothing to do on your side.
228
671
 
229
672
  - **Nine defects in six already-published manifests, and a repo upgraded before today may be holding
230
673
  the result.** The prose split above rewrote every item; this is the separate, smaller set where the
231
- DATA was wrong — found by auditing each manifest against the code of its own release. It matters
674
+ DATA was wrong. It matters
232
675
  because these files are executed: your upgrade tooling read them, acted, and reported done.
233
676
  **Three changes had no item at all.** `0.2.0` never filed the `--surface-subtle` →
234
677
  `--surface-subtitle` rename, so an override still naming the old token resolves to nothing and says
@@ -238,7 +681,7 @@ While the package is on `0.x`, minor releases may contain breaking changes.
238
681
  TS2416 a subclass hits had no explanation anywhere. `0.10.0` never filed
239
682
  `ModalService.confirm({ tone: 'error' })` gaining the modal's error badge in the header.
240
683
  **Two instructions were wrong, so following them edited the wrong thing.** `0.7.1`’s à-la-carte
241
- `@use` lines were written as bare `styles/…` specifiers, which resolve only inside this repo and
684
+ `@use` lines were written as bare `styles/…` specifiers, which resolve only inside the library's own build and
242
685
  fail from an app; they are package-qualified now. And `0.7.1`’s modal `.title` was filed as a
243
686
  `class-rename` to `.heading` when no rename happened — it is a `behavior` item now, so anyone who
244
687
  renamed that class did so on our word and can put it back.
@@ -254,7 +697,7 @@ While the package is on `0.x`, minor releases may contain breaking changes.
254
697
  those files changed meaning. Also corrected: `0.10.2`’s two `maxVisibleChips` items now declare
255
698
  that they supersede `0.10.0#max-visible-chips-zero-inverted`, which an upgrade stepping through both
256
699
  had been applying for nothing.
257
- **Pure upside in the same pass: sixteen more rows** in the "What you can now DELETE" tables of
700
+ **Sixteen more rows** in the "What you can now DELETE" tables of
258
701
  0.6.1, 0.7.1, 0.8.0, 0.9.0, 0.10.0 and 0.10.3 — workarounds those releases made redundant and never
259
702
  listed.
260
703
 
@@ -271,7 +714,6 @@ workaround written against a hook that only landed in this release.
271
714
  | `::ng-deep ds-button button` kept alive only by a resting `box-shadow` | `--ds-button-shadow` (and `--ds-button-shadow-hover`) on the host |
272
715
  | A `View Details` `ds-menu-item` in `rowActions` that only repeats where the row already leads | `link` on the column that names the record — a real `<a>`, tabbable and Ctrl/Cmd-clickable, announced as a link rather than as a table row |
273
716
 
274
-
275
717
  ## 0.10.3 — 2026-08-24
276
718
 
277
719
  ### Added
@@ -280,11 +722,11 @@ workaround written against a hook that only landed in this release.
280
722
  classes). `focus.owns-the-focus-ring($control)` — include it in a shell that paints its own ring
281
723
  on `:focus-within`, naming the ONE control inside that must not paint the base indicator too.
282
724
  `outline: 0` cannot do that, because the base indicator is a box-shadow, and a box-wide reset
283
- instead silences the tab stops living in the box — a clear button, a stepper arrow. It is the
284
- mixin `tools/check-focus-rings.mjs` checks for. `text-clip.single-line-ellipsis` — a one-line
725
+ instead silences the tab stops living in the box — a clear button, a stepper arrow. `text-clip.single-line-ellipsis` — a one-line
285
726
  label that truncates with an ellipsis, via `overflow: clip` plus `overflow-clip-margin`, because
286
727
  `hidden` under a line-height tighter than the font glyph box shaves the top and bottom off every
287
- glyph. `@use 'styles/focus'` / `@use 'styles/text-clip'` to reach them; a mixin emits nothing until
728
+ glyph. `@use '@ids-group-ltd/ids-design-system/styles/focus'` /
729
+ `@use '@ids-group-ltd/ids-design-system/styles/text-clip'` to reach them; a mixin emits nothing until
288
730
  it is included.
289
731
 
290
732
  ### Fixed
@@ -294,48 +736,39 @@ workaround written against a hook that only landed in this release.
294
736
  lands on the inner element, the box paints its own halo on `:focus-within`, and the `outline: 0`
295
737
  both already had cannot cancel a shadow. So the two plainest fields in the library went on drawing
296
738
  a 4px control ring around the `<input>`/`<textarea>` **and** a 3px field halo around the box.
297
- **They were missed because that pass trusted the shell list written in `styles/_focus.scss` rather
298
- than measuring.** A browser sweep — 9,968 focusable elements over 64 pages, transitions disabled so
299
- every computed shadow is final — found exactly these two and nothing else in the library. The reset
300
- stays scoped to the field element, so `clear-btn` and `password-toggle` inside the same box keep
301
- their own rings; they are real tab stops.
302
- - **The token reference was missing 14 public tokens that were always overridable.**
303
- `tools/extract-tokens.mjs` searched for the literal `var(--ds-`, so any resolver whose declaration
304
- wrapped across lines put a newline between `var(` and the name and the token silently never reached
305
- the reference. Newly listed and described: `--ds-container-pad-x`, `--ds-multi-select-border-color`
739
+ A sweep of every focusable element in the library found these two and nothing else. Both now paint
740
+ a single ring, and the reset stays scoped to the field element, so `clear-btn` and
741
+ `password-toggle` inside the same box keep their own rings; they are real tab stops.
742
+ - **Fourteen public override tokens are now listed in the token reference.** They were always
743
+ overridable and nothing in the components changed — only the reference was missing them. Newly
744
+ listed and described: `--ds-container-pad-x`, `--ds-multi-select-border-color`
306
745
  (+`-focus`), `--ds-segmented-selected-bg-color` / `-label-color` / `--ds-segmented-radius`,
307
746
  `--ds-time-picker-option-bg-color-selected` / `-text-color-selected` / `-bg-color-hover`,
308
747
  `--ds-time-picker-trigger-bg-color` / `-border` / `-color`, `--ds-field-message-icon-offset` and
309
- `--ds-time-picker-focus-shadow`. Nothing changed in the components —
310
- these overrides already worked; only the documentation was blind to them.
748
+ `--ds-time-picker-focus-shadow`.
311
749
  - **Two buttons inside portaled panels had no focus indicator, and one panel had square corners — a
312
750
  private token read from outside the host that declares it.** `ds-date-picker`'s `.period-trigger`
313
751
  (the month/year header, `tabindex="0"`) and `ds-time-picker`'s `.time-action` both did
314
752
  `box-shadow: var(--_…-focus-shadow)` beside `outline: none`. Those private tokens are declared on
315
753
  `:host`, the panels are portaled out of it by the CDK, so the `var()` resolved to nothing — and an
316
- unresolvable `var()` makes `box-shadow` its initial value, `none`. Measured: resting and focused
317
- paint were byte-identical on both, so a keyboard user landing there saw nothing. Another WCAG 2.4.7
318
- failure, and older than the 0.10.x focus work. `ds-time-picker`'s panel `border-radius` read the
319
- same way and fell back to `0`, so the panel rendered square instead of `--ds-radius-panel`.
320
- **Found by opening each overlay and probing inside it** — the earlier sweeps only ever saw
321
- focusables that exist while the overlay is closed. `ds-select`, `ds-multi-select` and
322
- `ds-autocomplete` panels were checked the same way and are clean.
754
+ unresolvable `var()` makes `box-shadow` its initial value, `none`, so a keyboard user landing there
755
+ saw nothing. Another WCAG 2.4.7 failure, and older than the 0.10.x focus work. `ds-time-picker`'s
756
+ panel `border-radius` read the same way and fell back to `0`, so the panel rendered square instead
757
+ of `--ds-radius-panel`. Both now paint correctly; `ds-select`, `ds-multi-select` and
758
+ `ds-autocomplete` panels are unaffected.
323
759
  - **`ds-search`'s clear button had no focus indicator at all — introduced by 0.10.2's own ring fix.**
324
760
  That fix silenced the base indicator with a shell-wide reset, which also hit `.search-clear`, a
325
- `tabindex="0"` tab stop. Measured: with the field's halo already lit, tabbing from the input to the
326
- × changed **no** paint property anywhere on screen — box-shadow, outline, border, background and
327
- colour were byte-identical in both states, so a keyboard user could not tell the × had focus. A
328
- WCAG 2.4.7 failure, and the same defect as the double ring seen from the other side. Every shell now
761
+ `tabindex="0"` tab stop: with the field's halo already lit, tabbing from the input to the × changed
762
+ no paint property anywhere on screen, so a keyboard user could not tell the × had focus. A WCAG
763
+ 2.4.7 failure, and the same defect as the double ring seen from the other side. Every shell now
329
764
  names the one control it speaks for through the new `focus.owns-the-focus-ring($control)` mixin
330
765
  instead of blanketing its box, and `ds-field-group` drops its blanket entirely — each segment
331
- already silences its own. `tools/check-focus-rings.mjs` fails the build when a component paints a
332
- ring on `:focus-within` without silencing the base rule; it replaces the hand-kept shell list in
333
- `styles/_focus.scss` that had been wrong for two releases.
766
+ already silences its own.
334
767
 
335
768
  ### What you can now DELETE
336
769
 
337
770
  None of this is required — skipping it leaves working code that is merely redundant. Each row is a
338
- workaround this release made unnecessary. Generated from the release manifest’s `deleteList`.
771
+ workaround this release made unnecessary. Generated from the `deleteList` in `migrations/0.10.3.json`.
339
772
 
340
773
  | If you have | Delete it, because |
341
774
  | --- | --- |
@@ -353,16 +786,14 @@ workaround this release made unnecessary. Generated from the release manifest’
353
786
  **both, never hue alone** — through `--ds-table-cell-link-color` (default `--accent`) and
354
787
  `--ds-table-cell-link-decoration` (default `underline`). Off unless a column asks for it, and inert
355
788
  on a passive table for the same reason the row tint is: nothing should promise a click that does not
356
- happen. Four list screens in one consumer had written this as
357
- `:host ::ng-deep ds-table tr:hover .cell-name { color: var(--accent); text-decoration: underline }`,
358
- which was their last table reach that no token could replace.
789
+ happen. This is commonly written by hand as `:host ::ng-deep ds-table tr:hover .cell-name { color:
790
+ var(--accent); text-decoration: underline }`.
359
791
 
360
792
  - **`TableRowAction` can vary its items per row: `hidden` and `disabledReason`.** The contract could
361
793
  only GREY an item that was always there, so a menu whose items swap by status — a draft offering Edit
362
- + Launch where a live record offers Revert + Delete — grew a permanently dead half on every row. That
363
- is why the row-action column shipped in 0.6.3 and one consumer adopted it in **zero** of eleven row
364
- menus. `hidden?: (row) => boolean` leaves the item out instead, and **a row with every item hidden
365
- renders no kebab at all** rather than opening an empty menu. `disabledReason?: string | ((row) =>
794
+ + Launch where a live record offers Revert + Delete — grew a permanently dead half on every row. Now
795
+ `hidden?: (row) => boolean` leaves the item out instead, and **a row with every item hidden renders
796
+ no kebab at all** rather than opening an empty menu. `disabledReason?: string | ((row) =>
366
797
  string)` puts the why on the item's second line, read only when `disabled` returns true for that row —
367
798
  a reason on a live item would read as a caveat on something that works.
368
799
 
@@ -370,14 +801,13 @@ workaround this release made unnecessary. Generated from the release manifest’
370
801
  the action carries ("only the first 1,000 rows", "read-only in this workspace"), visible without hover
371
802
  or focus, which is what separates it from a `[dsTooltip]`. **It reaches assistive tech as
372
803
  `aria-describedby`, not as part of the item's name:** the line is `aria-hidden` and referenced by id,
373
- so the projected label alone names the action. Seven items across four files in one consumer
374
- hand-rolled this with the same five-line rule; the width cap is now
375
- `--ds-menu-item-description-max-width` instead of each app's own `14rem`.
804
+ so the projected label alone names the action. The width cap is now
805
+ `--ds-menu-item-description-max-width` instead of a hand-written `14rem` cap.
376
806
 
377
807
  - **`data-focus="inset"` — a focus style that survives a clipping ancestor.** The ring is a box-shadow
378
808
  outside the border box, so any `overflow: hidden` ancestor ate it and left the element focusable with
379
- **no indicator at all** — measured in four places in one app, one of them a `ds-button` inside a
380
- wrapper that clips for its own animation. Clipping a card to hold a full-bleed child inside a rounded
809
+ **no indicator at all** — for example a `ds-button` inside a wrapper that clips for its own
810
+ animation. Clipping a card to hold a full-bleed child inside a rounded
381
811
  corner is normal, and `ds-card` + `padding: none` invites exactly that. **Unlike the other four focus
382
812
  styles this one is NOT pinned to `:root`:** a clip is a local problem, so the attribute works wherever
383
813
  you set it and the declarations inherit into the subtree. Put it on the clipping element.
@@ -391,7 +821,7 @@ workaround this release made unnecessary. Generated from the release manifest’
391
821
  - **`ds-search` binds to a form — both systems — and a disabled search now looks disabled.** It held a
392
822
  value that nothing could bind: no `NG_VALUE_ACCESSOR`, so `[(ngModel)]`, a `formControlName` and a
393
823
  signal form all slid off it while its own `[(value)]` kept working. It extends
394
- `DsControlValueAccessor` now, like every other control that carries a value, and a new `disabled`
824
+ the internal `DsControlValueAccessor` now, like every other control that carries a value, and a new `disabled`
395
825
  input folds together with the form's own disabled state. The shell paints the disabled surface from
396
826
  the inner input's state, which it never did — a disabled search looked live and still took a click.
397
827
  **Deliberately no label and no message row.** A toolbar filter is not a validated field, and giving
@@ -399,15 +829,15 @@ workaround this release made unnecessary. Generated from the release manifest’
399
829
  needs a label.
400
830
 
401
831
  - **`--ds-menu-radius` — the bare `ds-menu` corner, which two token descriptions already promised.**
402
- The component hardcoded `var(--radius-md)` while `token-descriptions.json` referred to
403
- `--ds-menu-radius` as though it were public; the dead-token audit caught it. Same default, so
404
- nothing moves — this only makes the documented hook real. The panelled case stays separate: inside a
405
- `ds-popover` the menu clips to `--ds-menu-clip-radius`, which the popover publishes.
832
+ The component hardcoded `var(--radius-md)` while its documentation referred to `--ds-menu-radius`
833
+ as though it were public. Same default, so nothing moves — this only makes the documented hook
834
+ real. The panelled case stays separate: inside a `ds-popover` the menu clips to
835
+ `--ds-menu-clip-radius`, which the popover publishes.
406
836
 
407
837
  ### Changed
408
838
 
409
839
  - **`ds-search` no longer takes the field message inputs, because it never rendered them.**
410
- Since it gained form binding it extended `DsControlValueAccessor`, which sits on `DsFieldBase` — so
840
+ Since it gained form binding it extended the internal `DsControlValueAccessor`, which sits on the equally internal `DsFieldBase` — so
411
841
  the API reference advertised `hint`, `error`, `warning`, `errors`, `invalid`, `touched`,
412
842
  `reserveMessage` and `messageIcon` on a component whose template has no `ds-field-message` and whose
413
843
  class reads none of them. **A consumer binding `error="Too short"` on a `ds-search` got silence, and
@@ -418,17 +848,17 @@ workaround this release made unnecessary. Generated from the release manifest’
418
848
  toolbar filter is not a validated field, which is the same reasoning that kept the message row off it
419
849
  in the first place.
420
850
  **How the split works, since it is a shape other controls may want:** the `ControlValueAccessor`
421
- plumbing is now `valueControlBridge()`, a function rather than a base class, so a control can bind to
422
- a form without inheriting the message API. `DsControlValueAccessor` is that function plus
423
- `DsFieldBase` and delegates to it, so the logic exists once rather than twice. `hidden` stayed on
424
- `ds-search`: it belongs to the forms contract — signal forms reaches every control through it — not to
425
- the message API.
851
+ plumbing is a function rather than a base class, so a control can bind to a form without
852
+ inheriting the message API; the internal CVA base is that function plus the shared field base and
853
+ delegates to it, so the logic exists once rather than twice. `hidden` stayed on `ds-search`: it
854
+ belongs to the forms contract — signal forms reaches every control through it — not to the
855
+ message API.
426
856
  - **A field's message line now carries a glyph for `error`, `warning` and `success` — and not for
427
- `hint`.** Colour was the whole signal before, and in the light theme warning (#8A7400) against success
428
- (#156822) is a hue apart and nothing else: WCAG 1.4.1 is about exactly that, and four of the six
429
- systems audited for the fill question carry a per-state icon in the field. **`hint` deliberately stays
430
- bare** — it is the resting text rather than a state, and it is the majority of rows on any real form
431
- (25 of 43 on our own input page), so a glyph there is noise rather than signal.
857
+ `hint`.** Colour was the whole signal before, which fails WCAG 1.4.1: in the light theme warning
858
+ (#8A7400) against success (#156822) is a hue apart and nothing else. Four of six major design
859
+ systems carry a per-state icon in
860
+ the field. **`hint` deliberately stays bare** — it is the resting text rather than a state, and it
861
+ is the majority of rows on any real form, so a glyph there is noise rather than signal.
432
862
  **Per state, independently, app-wide:**
433
863
  `provideDsConfig({ field: { stateIcons: { warning: false, hint: 'info' } } })` — a name replaces the
434
864
  glyph, `false` removes it, each key is independent of the others, and **`stateIcons: false` is the one
@@ -447,15 +877,15 @@ workaround this release made unnecessary. Generated from the release manifest’
447
877
  `number`, `currency`, `percent` and `date` are all typed `string | null`, so
448
878
  `[value]="count | number"` failed `strictTemplates` with TS2322 and consumers wrote
449
879
  `count.toLocaleString('en-US')` in the template instead — which pins a locale rather than reading
450
- `LOCALE_ID`, and re-runs on every change-detection pass where a pure pipe is cached. One consumer's
451
- KPI row carries both idioms side by side for exactly this reason. `null` renders as nothing, like the
452
- empty default, and the four pipes now have a demo on the docs page — which is also the test, since
453
- that section compiles under `strictTemplates`.
454
-
455
- - **If you EXTEND a DS component, eleven classes grew and six members changed type.** `npm run
456
- class-surface` against 0.10.1: `DsFieldBase` (`messageIcon` — the state glyph, declared once for
457
- every field rather than per component), `DsControlValueAccessor` (`bridge`),
458
- `FieldMessageComponent` (`config`, `iconResolved`), `AutocompleteComponent` and
880
+ `LOCALE_ID`, and re-runs on every change-detection pass where a pure pipe is cached. `null` renders
881
+ as nothing, like the empty default, and the four pipes have a demo on the docs page.
882
+
883
+ - **If you EXTEND a DS component, eleven classes grew and six members changed type.** Against
884
+ 0.10.1, and the first three are internal so they reach only a fork: `DsFieldBase`
885
+ (`messageIcon` — the state glyph, declared once for every field rather than per component),
886
+ internal `DsControlValueAccessor` (`bridge`), internal `FieldMessageComponent` (`config`,
887
+ `iconResolved`),
888
+ `AutocompleteComponent` and
459
889
  `MultiSelectComponent` (`chipsMode`, `showsEveryChip`), `MenuItemComponent` (`description`,
460
890
  `descriptionId`), `SearchComponent` (the CVA surface it now declares itself: `bridge`,
461
891
  `disabled`, `formDisabled`, `hidden`, `isDisabled`, `markTouched`, `registerOnChange`,
@@ -466,14 +896,13 @@ workaround this release made unnecessary. Generated from the release manifest’
466
896
  collides just as a public one does, and a name on `DsFieldBase` reaches every field through it.
467
897
  **Six members kept their name and changed type**, which is the case that compiles for a caller
468
898
  and breaks an overrider: `maxVisibleChips` on both pickers (widened to
469
- `number | 'fit' | 'all' | null`, default now `null`), `FieldMessageComponent.icon` (widened to
899
+ `number | 'fit' | 'all' | null`, default now `null`), internal `FieldMessageComponent.icon` (widened to
470
900
  `IconName | false | null`, so `false` can mean "no glyph" against a state that supplies one),
471
901
  `DsConfig.field` (its shape grew the glyph defaults), `SearchComponent.clear` (`protected` →
472
902
  public), `StatComponent.value` (widened to `string | number | null`) and
473
- `TimeColumnsComponent.onKey` (its parameter list changed). **What the report does NOT show:**
474
- `ds-search` no longer extends the field base, so it loses eight inherited inputs — the tool
475
- compares declarations per file and a dropped `extends` moves nothing, so that break is documented
476
- in its own entry above and not here.
903
+ `TimeColumnsComponent.onKey` (internal, and its parameter list changed). **One more break outside this list:**
904
+ `ds-search` no longer extends the field base, so it loses eight inherited inputs — documented in
905
+ its own entry above.
477
906
 
478
907
  - **Every readback that resolves through `DS_CONFIG` is now public.** `sizeResolved` on fourteen
479
908
  controls, plus `elevationResolved` / `paddingResolved` (card), `withDotResolved` (badge),
@@ -507,19 +936,19 @@ workaround this release made unnecessary. Generated from the release manifest’
507
936
  - **The active `ds-sidenav-item` is a tinted wash with a coloured label, not a solid brand pill** — and
508
937
  the edge rail is an input rather than a token you blank out. The item filled with `--accent` and put
509
938
  `--text-on-primary` on its label, making the loudest element on the screen out of a state that only
510
- says "you are here"; two apps independently replaced it, which is the signal that decided this.
511
- Material fills its navigation drawer's active item with `secondary-container` and labels it
939
+ says "you are here". Material fills its navigation drawer's active item with `secondary-container` and labels it
512
940
  `on-secondary-container`; Carbon uses `$background-selected` with a border and a weight bump.
513
941
  Neither fills with the brand accent. Ours is `--accent-subtitle` behind an `--accent-strong` label,
514
942
  and hover deepens toward the accent (`--accent-muted`) instead of lightening toward white — the old
515
943
  88%-toward-white mix was invisible on a wash. **The state still does not rest on colour alone:** the
516
944
  edge rail and the weight bump both stay — and **the rail is now `--accent-strong`, the active label's
517
- own colour, instead of `--secondary`.** An orange bar across a monochrome-blue row measured **2.33:1**
518
- on the light wash, under the 3:1 floor a non-text indicator needs; the accent measures 6.16:1 light and
945
+ own colour, instead of `--secondary`.** An orange bar across a monochrome-blue row measured
946
+ **2.33:1** on the light wash, under the 3:1 floor a non-text indicator needs; the accent measures
947
+ 6.16:1 light and
519
948
  5.28:1 dark. Carbon's indicator is the interactive colour and Material's IS the wash — neither
520
949
  introduces a second hue. Restore the old bar with `--ds-sidenav-item-rail: var(--secondary)`.
521
- Every `--ds-sidenav-item-*-active` token keeps its name, so
522
- an app that already retuned them keeps its own look and needs no edit.
950
+ Every `--ds-sidenav-item-*-active` token keeps its name, so an app that already retuned them keeps
951
+ its own look and needs no edit.
523
952
  **`[rail]="false"` on `ds-sidenav`** drops the edge marker for every item, and an item can override
524
953
  it for itself (`[rail]` on `ds-sidenav-item`, unset = follow the parent). The old route was
525
954
  `--ds-sidenav-item-rail: transparent` or a `::ng-deep` box-shadow reset — both still work, but the
@@ -527,35 +956,31 @@ workaround this release made unnecessary. Generated from the release manifest’
527
956
  the library is expected to implement it.
528
957
 
529
958
  - **The shipped migration manifests say WHERE to look, not only what changed — and a lint now holds
530
- them to it.** `migrations/<version>.json` (the files `dist/ds/migrations/` carries into your
531
- `node_modules`, which the auto-update plugin runs) gained `surface`, `detect`, `verify`, `restore`,
532
- `supersedes`, and a structured `deleteList`. The reason is a consumer upgrade that
533
- went 0.6.0 → 0.10.1 and passed over two entries in silence: `0.7.1`'s
534
- `table-sortkey-non-null-when-unsorted` pointed at `ds-table[sortKey]`, which appears in nobody's
535
- template, while the code that breaks was their own store —
536
- `sorting: store.sortKey() ? { path: store.sortKey()!, … } : undefined` — so the search found
537
- nothing, the item was recorded not-applicable, and the re-check's clean grep agreed. Three
538
- server-paged screens kept sorting by a column whose header showed no sort. `surface` and `detect`
539
- make that item searchable in the artifact that actually changes; `verify` gives a behavioural entry
540
- the positive assertion a re-check can test, since there is no removed symbol whose absence proves
541
- anything. **Every published manifest was back-filled the same day:** 167 items across 19 releases
542
- got a `surface`, every behavioural one a `verify`, every changed default a `restore`, and the 32
543
- prose `DELETE-LIST:` lines became `deleteList` arrays with a greppable `find` each — four of those
544
- "lines" turned out to be the words "the DELETE-LIST" inside ordinary prose. Two default flips that
545
- reverse each other are linked by `supersedes` on the LATER item, one way only because a published
546
- manifest cannot know its own future: `0.7.0` / `0.7.1` on `reserveMessage`, and `0.8.0` / `0.9.0`
547
- on the table's sort mode. `npm run migration-lint` enforces the contract with no warning level, and
548
- both the release preflight and the PR gate run it, so a manifest nobody can act on stops the
549
- pipeline instead of reaching a consumer. Three CHANGELOG blocks also gained the "What you can now
550
- DELETE" table their manifests had been carrying alone — 0.10.0's nineteen rows, 0.10.1's and
551
- 0.6.2's.
959
+ them to it.** `migrations/<version>.json` (the files the package carries into your
960
+ `node_modules`) gained `surface`, `detect`, `verify`, `restore`,
961
+ `supersedes`, and a structured `deleteList`. The reason: a migration's `find` pattern can point at a
962
+ prop or class name that never appears in your own code even though the underlying behaviour broke
963
+ there instead — a mechanical grep finds nothing, so an item gets recorded not-applicable while the
964
+ break survives. That is not hypothetical: `0.7.1#table-sortkey-non-null-when-unsorted` named
965
+ `ds-table[sortKey]`, while the code that actually breaks is a host's own
966
+ `sorting: store.sortKey() ? { … } : undefined` — so an upgrade that crossed 0.7.1 before this
967
+ release may have recorded it not-applicable. Re-check it if you server-page a table. `surface` and
968
+ `detect` make that item searchable in the artifact that actually changed; `verify` gives a behavioural entry the positive assertion a re-check can test, since there
969
+ is no removed symbol whose absence proves anything. **Every published manifest now carries the same
970
+ fields, back to the earliest release**, so an upgrade tool works consistently across the
971
+ whole upgrade chain. Two default flips that reverse each other are linked by `supersedes` on the
972
+ LATER item, one way only because a published manifest cannot know its own future: `0.7.0` / `0.7.1`
973
+ on `reserveMessage`, and `0.8.0` / `0.9.0` on the table's sort mode. Three CHANGELOG blocks also
974
+ gained the "What you can now DELETE" table their manifests had been carrying alone — 0.10.0's,
975
+ 0.10.1's and 0.6.2's. A lint enforces the contract at release and on every merge, so a manifest
976
+ nobody can act on cannot reach you.
552
977
 
553
978
  ### Fixed
554
979
 
555
980
  - **One focus ring per field, not two.** Since 0.10.0 the universal `:where(:focus-visible)` rule has
556
981
  given every focusable element a `box-shadow`, and these shells paint their own on `:focus-within` —
557
982
  `outline: 0`, which each inner control already had, cannot cancel a shadow. So a focused field drew a
558
- tight ring around its inner control AND a halo around the box. Measured on five shapes: `ds-search`,
983
+ tight ring around its inner control AND a halo around the box. Affected five shapes: `ds-search`,
559
984
  `ds-autocomplete`, `ds-date-picker`, `ds-number-stepper`, and any field inside a `ds-field-group`
560
985
  (fixed once in the shared segment mixin). It is the same trap 0.10.0 warned consumers about — an app
561
986
  rule that styles focus with an `outline` only — landing in our own components.
@@ -564,35 +989,33 @@ workaround this release made unnecessary. Generated from the release manifest’
564
989
  the halo is already on, so it cannot say which of them focus moved to. The date picker's calendar
565
990
  toggle is `tabindex="-1"` and correctly loses its ring.
566
991
 
567
- - **A flat line box stopped shaving descenders off clipped labels** (mm-broker MBP-245). `.option` sets
992
+ - **A flat line box stopped shaving descenders off clipped labels.** `.option` sets
568
993
  `line-height: 1` while `.option-label` adds the `overflow: hidden` its ellipsis needs and no padding,
569
- so the label box was exactly the font size tall and the glyph box was taller: measured **2.13px above
570
- the cap line and 1.47px below the baseline** clipped, on every option rather than only the ones with a
571
- descender. Fixed with `overflow: clip` plus a `var(--space-1)` clip margin, which **moves no geometry**
572
- — the option row stays 30px, the multi-select row 36px, the table row 38px.
994
+ so the label box was exactly the font size tall and the glyph box was taller, clipping the top and
995
+ bottom of every option's text rather than only the ones with a descender. Fixed with
996
+ `overflow: clip` plus a `var(--space-1)` clip margin, which **moves no geometry** — the option row
997
+ stays 30px, the multi-select row 36px, the table row 38px.
573
998
  **Six sites:** the shared dropdown overlay (`ds-select`, `ds-autocomplete`), `ds-multi-select`'s own
574
- copy, `.cell-clip` in `ds-table` and `.label` in `ds-breadcrumbs` (both measured), and the `.crumb`
575
- rules in `ds-topbar` / `ds-page-header`, which share the shape and are unreachable in our own docs.
576
- **Two behaviours checked rather than assumed:** the ellipsis still applies (`text-overflow` reads any
999
+ copy, `.cell-clip` in `ds-table`, `.label` in `ds-breadcrumbs`, and the `.crumb` rules in
1000
+ `ds-topbar` / `ds-page-header`, which share the same shape.
1001
+ **Two behaviours are unaffected:** the ellipsis still applies (`text-overflow` reads any
577
1002
  non-`visible` overflow), and `ds-table`'s clipped-cell popover still fires, since `scrollWidth >
578
1003
  clientWidth` still reports true under `clip`.
579
1004
  - **A field bound with signal forms threw once per field, and then showed an error it had already
580
- cleared.** `resolveFieldMessage` subscribes to `NgControl`'s `events` stream to follow reactive and
1005
+ cleared.** The shared message resolver subscribes to `NgControl`'s `events` stream to follow reactive and
581
1006
  template-driven forms. The `NgControl` that signal forms bridges carries no such stream, so the
582
- subscription threw `TypeError: Cannot read properties of undefined (reading 'pipe')` — four times on
583
- our own Practice → Signal Forms page, swallowed into the console. **The second half is what a user
584
- saw:** the throw left that snapshot frozen at first render, and the message resolver falls back to it
585
- whenever the live errors go empty — so a field that had just become VALID went on showing its FIRST
586
- failure, with `aria-invalid="true"`. Measured through CDP with real input: empty → "We need an email",
587
- `nope` → "That does not look like an email address", a valid address → **the required message again**.
588
- Now the `NgControl` path bows out entirely when it has no live stream, leaving the component's own
589
- inputs — which is where a signal-forms field's errors and touched arrive — as the single source. The
590
- same sequence now clears the message and drops `aria-invalid`. **Reactive and template-driven forms are
591
- untouched** and were re-checked on `/components/input`: too short, empty and valid each render
592
- correctly, as before.
1007
+ subscription threw `TypeError: Cannot read properties of undefined (reading 'pipe')`, swallowed into
1008
+ the console on every signal-forms field. **The second half is what a user saw:** the throw left
1009
+ that snapshot frozen at first render, and the message resolver falls back to it whenever the live
1010
+ errors go empty — so a field that had just become VALID went on showing its FIRST failure, with
1011
+ `aria-invalid="true"`. For example: empty → "We need an email", `nope` → "That does not look like an
1012
+ email address", a valid address → **the required message again**. Now the `NgControl` path bows out
1013
+ entirely when it has no live stream, leaving the component's own inputs — which is where a
1014
+ signal-forms field's errors and touched arrive — as the single source. The same sequence now clears
1015
+ the message and drops `aria-invalid`. **Reactive and template-driven forms are untouched.**
593
1016
 
594
1017
  - **`components.md` was missing two layouts and naming three sidenav defaults that are not the
595
- defaults.** The digest `tools/build-schematics.mjs` copies into `dist/ds/` is what a consumer's own
1018
+ defaults.** The digest shipped in the package is what a consumer's own
596
1019
  agent reads to answer "what does this library have", and `ds-app-layout` and `ds-auth-layout` were
597
1020
  absent from it entirely — an agent asked to build a shell or a sign-in screen could not find either
598
1021
  and had no way to know they existed. It also named `--primary`, `--text-on-primary` and
@@ -604,8 +1027,8 @@ workaround this release made unnecessary. Generated from the release manifest’
604
1027
 
605
1028
  - **Five more overlays dismiss on a click in the host's own empty width.** `ds-select`,
606
1029
  `ds-multi-select`, `ds-autocomplete`, `ds-time-picker` and `ds-color-picker` all tested the whole
607
- HOST in their outside-pointer guard, which is the defect `ds-date-picker` was fixed for in 0.10.1
608
- (MBP-225): the trigger box hugs its own content while the host takes whatever width a consumer's
1030
+ HOST in their outside-pointer guard, which is the defect `ds-date-picker` was fixed for in 0.10.1 —
1031
+ the trigger box hugs its own content while the host takes whatever width a consumer's
609
1032
  layout gives it, so the label, the message row and any bare strip beside the trigger swallowed the
610
1033
  dismiss instead of performing it. Each guard now tests its TRIGGER, as `ds-popover` always did;
611
1034
  `ds-time-picker` tests whichever of its three triggers are present, since a range renders two.
@@ -614,14 +1037,14 @@ workaround this release made unnecessary. Generated from the release manifest’
614
1037
  caret, and re-focus).
615
1038
 
616
1039
  - **`variant="tonal"` now reads as disabled in the dark theme, where it rested on desaturation alone.**
617
- Measured on `/components/button`: the disabled label was **1.11:1** from the enabled one (#778ff8 →
618
- #979eb5) and the fill **1.20:1**, so the only signal was a hue going grey — and the label sat at 5.51:1
619
- on its own fill, which reads as a live control. Tonal now takes the same 40% label mix ghost, text and
620
- outline already use, toward `--surface-default`: the state gap becomes **2.90:1** in dark and **3.39:1**
621
- in light, and the label keeps its hue instead of turning grey. **Toward the surface and not toward
622
- tonal's own fill, even though the label sits on that fill** — the disabled fill has already moved
623
- toward the label, so mixing into it yields only 2.56:1. The fill itself still barely moves (1.09:1
624
- light, 1.20:1 dark) and does not need to: the label carries the state, which is the same division of
1040
+ The disabled label was **1.11:1** from the enabled one and the fill **1.20:1**, so the only signal was
1041
+ a hue going grey — and the label sat at **5.51:1** on its own fill, which reads as a live control.
1042
+ Tonal now takes the same 40% label mix ghost, text and outline already use, toward
1043
+ `--surface-default`: the state gap
1044
+ becomes **2.90:1** in dark and **3.39:1** in light, and the label keeps its hue instead of turning
1045
+ grey. **The mix moves toward the surface, not toward tonal's own fill**, even though the label sits
1046
+ on that fill: mixing into the fill instead would keep the gap too small to read. The fill itself
1047
+ barely moves and does not need to: the label carries the state, which is the same division of
625
1048
  labour the transparent variants shipped with in 0.10.1. Override with
626
1049
  `--ds-button-label-color-disabled` as before.
627
1050
 
@@ -634,7 +1057,7 @@ workaround this release made unnecessary. Generated from the release manifest’
634
1057
  outside-pointer guard treated the whole host element as the trigger, but the visible trigger box
635
1058
  hugs its mask while the host takes whatever a consumer's layout gives it — `flex-grow: 1` beside a
636
1059
  Clear button leaves a wide bare strip between the field and the button, and a click there kept the
637
- calendar open with no visible reason (mm-broker MBP-225). The guard now tests the trigger box
1060
+ calendar open with no visible reason. The guard now tests the trigger box
638
1061
  itself: a press on the field or its calendar glyph still keeps the panel up, while the label, the
639
1062
  message row and the host's empty width dismiss it like any other outside click.
640
1063
 
@@ -662,54 +1085,12 @@ workaround written against a hook that only landed in this release.
662
1085
 
663
1086
  ## 0.10.1 — 2026-08-20
664
1087
 
665
- ### Removed
666
-
667
- - **The seven inputs every field shares are declared once, in `DsFieldBase`.** `hint`, `error`,
668
- `warning`, `errors`, `invalid`, `touched` and `reserveMessage` — plus the `reserveMessageResolved`
669
- helper — were hand-copied into all fourteen fields, 98 declarations. **Nothing about the public API
670
- changed:** every input keeps its name, type, default and documentation, and the docs API tables are
671
- byte-identical apart from the eight new `success` rows (verified by diffing the generated
672
- reference: 0 lost, 8 gained).
673
- **Why it is worth knowing anyway:** the duplication had already drifted twice — `success` was
674
- missing from eight fields, and `ds-date-picker` had quietly moved to a different `invalid` source.
675
- A new state added to the base now reaches every field at once.
676
- `DsControlValueAccessor` and `DsBooleanControlValueAccessor` extend the base (so single inheritance
677
- still holds for the thirteen fields that already extended one of them) and `ds-radio` extends it
678
- directly. Only the UNIVERSAL inputs moved: `success`, `label`, `required` and `optional` stay on the
679
- components, because they exist on 9, 9, 7 and 5 fields respectively and lifting them would hand a
680
- checkbox a `success` or a slider a `label` it never had.
681
-
682
- - **`ds-icon-button` with `variant="standard"` mixes its disabled glyph toward the surface too.**
683
- New default on that variant only: `color-mix(in srgb, <icon colour> 40%, var(--surface-default))`.
684
- Same defect as ghost, and it matters more here — there is no word to read, just a 16px glyph.
685
- Measured: light **4.17 → 5.19:1**, dark **1.46 → 3.58:1**. Light was already healthy, because the
686
- enabled glyph is `--text-secondary` at OKLCH L 35% and the old recipe moved it to L 70%; **dark was
687
- the break**, where the enabled glyph is L 80% and a disabled one pinned at L 70% left ten points
688
- and a 1.46:1 gap.
689
- **Scoped to `standard` deliberately, and this is the trap worth knowing:** the resolver is declared
690
- once on the base and no variant overrode it, so touching the base would have hit `filled` too —
691
- whose glyph is `--c-on`, i.e. white, and white mixed into a white surface does not move at all,
692
- which would have put an invisible white glyph on a pale disabled fill. `filled`, `tonal` and
693
- `outlined` keep the base recipe unchanged. Consumers who set
694
- `--ds-icon-button-icon-color-disabled` are unaffected.
695
-
696
- - **`--opacity-muted` is gone from the token scale.** It was the last token in the file that no
697
- component read: the disabled states moved onto `--text-disabled` + `--opacity-disabled` in
698
- 0.10.0, and it had been sitting there since, labelled a legacy alias. Keeping a dead number in a
699
- purpose-named scale is worse than removing it — it invites exactly the override that 0.10.0's
700
- DELETE-LIST had to warn about twice, where a consumer retunes it on `ds-button` /
701
- `ds-icon-button` and gets a silent no-op. **Nothing in the library changes.** If you set it, drop
702
- the declaration; if you READ it in your own CSS, inline the 0.6 or move to the token that names
703
- the job — `--opacity-disabled` (0.5) for a disabled state where opacity is the only signal,
704
- `--opacity-text-disabled` (0.38) for disabled text. The comment above the block claims every
705
- token in it has a real consumer; that is true again.
706
-
707
1088
  ### Added
708
1089
 
709
1090
  - **`ds-segmented` joins the form systems, and gains a label and a message row.** It carried a value
710
1091
  and no form could bind it — no `NG_VALUE_ACCESSOR`, so `[(ngModel)]`, a `formControlName` and
711
- signal forms all slid off it. It now extends `DsControlValueAccessor` and resolves its message
712
- through the same `resolveFieldMessage` as every other field, mirroring `ds-radio-group`: `label`,
1092
+ signal forms all slid off it. It now extends the internal `DsControlValueAccessor` and resolves its message
1093
+ the same way as every other field, mirroring `ds-radio-group`: `label`,
713
1094
  `hint`, `error`, `warning` and `errors` all behave as they do elsewhere, and `disabled` folds
714
1095
  together with the form's own disabled state. **The DOM grew a level to fit them:** the pill moved
715
1096
  off `:host` onto an inner `.track`, so the host can stack label, pill and message. The pill's own
@@ -717,23 +1098,6 @@ workaround written against a hook that only landed in this release.
717
1098
  `::ng-deep` rule written against the host AS the pill (a background, a border-radius, a grid) now
718
1099
  targets the stack instead, and wants `.track`.
719
1100
 
720
- - **`invalid` paints the error state only — it no longer promotes `hint` to the error message.**
721
- Every state carries its own text now, and `hint` is never recoloured into somebody else's state.
722
- **BREAKING:** `[invalid]="true" hint="Must be before end date"` used to render that hint in red,
723
- through `role="alert"`. It now renders a red border with a **grey** hint. Use
724
- `error="Must be before end date"` instead — the state and the message come together.
725
- **Why, beyond tidiness.** The idiom lost the actual hint: when both were set, the help text the
726
- author wrote was replaced by the error rather than shown alongside it. It also announced help text
727
- as an alert, and it made a template unreadable — whether `hint="…"` was a hint or an error depended
728
- on a different attribute. It was the last input that borrowed another one's text; `success` stopped
729
- doing it in the same release.
730
- **`invalid` is not going away and still has its own job:** the error state with NO message, for when
731
- the text lives elsewhere — a form-level error summary, or one message covering a whole
732
- `ds-field-group` — plus `aria-invalid` for that case. Carbon and Material split the boolean from
733
- the text the same way.
734
- Our own docs taught the old idiom on 13 components (14 of 19 `[invalid]` bindings also bound
735
- `hint`); all of them now use `error`.
736
-
737
1101
  - **A form model can now raise a WARNING, not only an error** — `DS_FIELD_WARNING_KIND`, newly
738
1102
  exported. A validation error whose `kind` is `'warning'` paints the field's warning state and shows
739
1103
  its message, but leaves the control **valid**: `aria-invalid` stays off and a submit guard reading
@@ -791,7 +1155,7 @@ workaround written against a hook that only landed in this release.
791
1155
  - **Per-side padding on `ds-modal` / `ds-modal-card`.** Each region resolved ONE `pad-y` applied to
792
1156
  both of its edges, so a layout whose header pads only at the top and whose actions row pads only
793
1157
  at the bottom — the body's own padding carrying the whole gap between them — could not be said
794
- through the API at all. mm-broker has held three `!important` rules on `.modal > .header` /
1158
+ through the API at all. A common workaround has been three `!important` rules on `.modal > .header` /
795
1159
  `> .body` / `> .actions` for exactly this, reaching past the component into its internal class
796
1160
  names, and `DS_CONFIG.modal.dividers` could not retire them. Six new tokens —
797
1161
  `--ds-modal-header-pad-top` / `-pad-bottom` and the same for `body` and `actions` — each fall back
@@ -806,6 +1170,69 @@ workaround written against a hook that only landed in this release.
806
1170
  `--ds-table-icon-btn-focus-color`, defaulting to exactly what was hardcoded, so nothing moves unless
807
1171
  you set them.
808
1172
 
1173
+ ### Changed
1174
+
1175
+ - **`invalid` paints the error state only — it no longer promotes `hint` to the error message.**
1176
+ Every state carries its own text now, and `hint` is never recoloured into somebody else's state.
1177
+ **BREAKING:** `[invalid]="true" hint="Must be before end date"` used to render that hint in red,
1178
+ through `role="alert"`. It now renders a red border with a **grey** hint. Use
1179
+ `error="Must be before end date"` instead — the state and the message come together.
1180
+ **Why, beyond tidiness.** The idiom lost the actual hint: when both were set, the help text the
1181
+ author wrote was replaced by the error rather than shown alongside it. It also announced help text
1182
+ as an alert, and it made a template unreadable — whether `hint="…"` was a hint or an error depended
1183
+ on a different attribute. It was the last input that borrowed another one's text; `success` stopped
1184
+ doing it in the same release.
1185
+ **`invalid` is not going away and still has its own job:** the error state with NO message, for when
1186
+ the text lives elsewhere — a form-level error summary, or one message covering a whole
1187
+ `ds-field-group` — plus `aria-invalid` for that case. Carbon and Material split the boolean from
1188
+ the text the same way.
1189
+
1190
+ - **A disabled `ds-icon-button variant="standard"` mixes its glyph toward the surface too.** It is
1191
+ the only icon-button variant with neither a fill nor a border, so the glyph's colour is the whole
1192
+ disabled signal — and the shared fixed-lightness recipe could not carry it in dark, where the
1193
+ enabled icon is already light (OKLCH L 80%) and a disabled one pinned at L 70% sat **1.46:1** away. The new
1194
+ default is `color-mix(in srgb, <icon> 40%, var(--surface-default))`, the same mix the
1195
+ borderless button variants take, which picks its direction per theme: light **4.17 → 5.19:1**, dark
1196
+ **1.46 → 3.58:1**. Scoped to `standard` deliberately, and that is the trap worth knowing: the resolver is declared
1197
+ once on the base and no variant overrode it, so moving the base would have hit `filled` too —
1198
+ whose glyph is `--c-on`, i.e. white, and white mixed into a white surface does not move at all,
1199
+ which would have left an invisible glyph on a pale disabled fill. `filled`, `tonal` and
1200
+ `outlined` keep the base recipe unchanged. `--ds-icon-button-icon-color-disabled` is the same
1201
+ hook as before and still wins; only the default under it changed.
1202
+
1203
+ - **If you EXTEND a DS field component, read this before comparing member lists.** A diff of
1204
+ declared members against 0.10.0 says fourteen field components LOST eight members each — `hint`, `error`,
1205
+ `warning`, `errors`, `invalid`, `touched`, `reserveMessage` and `reserveMessageResolved`.
1206
+ **Nothing was removed.** The duplication had drifted — `ds-date-picker` had been reading a different
1207
+ `invalid` source, and now resolves it like every other field. Those inputs were declared once per
1208
+ component and are now declared once in total, on a new internal `DsFieldBase`, which every field
1209
+ reaches either directly or through the internal
1210
+ `DsControlValueAccessor` / `DsBooleanControlValueAccessor`. All fourteen still carry all eight by
1211
+ inheritance, so a template binding, a TS read and an existing `override` all behave exactly as
1212
+ before. A member that moves to a base class is easy to misread as removed; it is not.
1213
+ What genuinely grew is smaller: **eight bordered fields gained `success`** (`ds-autocomplete`,
1214
+ `ds-color-picker`, `ds-date-picker`, `ds-multi-select`, `ds-otp-input`, `ds-select`,
1215
+ `ds-textarea`, `ds-time-picker`), `ds-input` having had it all along. **Only the universal inputs moved:** `success`, `label`, `required` and `optional` stay on the
1216
+ components, because they exist on 9, 9, 7 and 5 fields respectively — lifting them would hand a
1217
+ checkbox a `success` or a slider a `label` it never had. `ds-segmented`
1218
+ is the tenth grown class and the one with real new surface — joining the form systems gave it
1219
+ `label`, `disabled`, `isDisabled`, `message`, `invalidResolved`, `fieldMessage`, `hintId` and
1220
+ `uid`. A subclass declaring `success` on one of those eight now fails with TS4114; declaring any
1221
+ of the eight moved names still collides exactly as it did, because an inherited member collides
1222
+ the same way an own one does.
1223
+ `DsFieldBase` is internal and deliberately not exported.
1224
+
1225
+ ### Removed
1226
+
1227
+ - **`--opacity-muted` is gone from the token scale.** It was the last token in the file that no
1228
+ component read: the disabled states moved onto `--text-disabled` + `--opacity-disabled` in
1229
+ 0.10.0. Keeping a dead number in a purpose-named scale is worse than removing it — it invites the
1230
+ override that silently does nothing, where a consumer retunes it on `ds-button` /
1231
+ `ds-icon-button` and gets no visible effect. **Nothing in the library changes.** If you set it,
1232
+ drop the declaration; if you READ it in your own CSS, inline the 0.6 or move to the token that
1233
+ names the job — `--opacity-disabled` (0.5) for a disabled state where opacity is the only signal,
1234
+ `--opacity-text-disabled` (0.38) for disabled text.
1235
+
809
1236
  ### Fixed
810
1237
 
811
1238
  - **`ds-field-message`'s docs claimed `state` picks an icon. It never did.** The input's docblock
@@ -832,9 +1259,9 @@ workaround written against a hook that only landed in this release.
832
1259
  showed that square fill through it. The popover now publishes `--ds-menu-clip-radius` — its own
833
1260
  radius less the border it contains — and the menu clips to that, concentric at 11px inside a 12px
834
1261
  panel. A brand overriding `--ds-popover-radius` is followed automatically.
835
- **Not fixed with `overflow: hidden` on the panel**, which was the obvious move: a submenu is a CDK
1262
+ **Not fixed with `overflow: hidden` on the panel:** a submenu is a CDK
836
1263
  overlay portaled to `body`, so clipping the panel would not have reached it, and the panel's clip
837
- would have cut an edge row's focus ring — the same defect fixed in `ds-filter-bar` this month.
1264
+ would have cut an edge row's focus ring — the same defect `ds-filter-bar` also had.
838
1265
  Nothing else changes: the item radius token `--ds-menu-item-radius` still defaults to 0, because the
839
1266
  MENU owns the edge corners, not the row.
840
1267
 
@@ -845,20 +1272,17 @@ workaround written against a hook that only landed in this release.
845
1272
  has no error of its own to bind, and an empty string falls straight back to the picker's.
846
1273
  The `value → text` effect now clears both flags. It is the right place because every typing path
847
1274
  holds an `editing` guard while it runs, so reaching the effect means the value did NOT come from the
848
- keyboard — and a leftover error no longer describes what the field is about to show. Verified that
849
- neither typing path writes `value` on the invalid branch, so blur validation still stands.
850
- **This retires two consumer workarounds.** One app was tearing down a whole modal body on `closed`
851
- purely to get a fresh picker — and that had a hole, since reopening inside the 300 ms leave animation
852
- cancels the close, `closed` never fires and the stale error survived. A second dialog had no
853
- workaround at all, so a mistyped date of birth read "Enter a valid date" over an empty field the next
854
- time it opened.
1275
+ keyboard — and a leftover error no longer describes what the field is about to show. Neither typing
1276
+ path writes `value` on the invalid branch, so blur validation still stands. If you tore a modal body
1277
+ down on `(closed)` purely to get a fresh picker, that can go — and it never worked reliably anyway,
1278
+ because reopening inside the 300 ms leave animation cancels the close and `closed` never fires.
855
1279
 
856
1280
  - **A disabled `ds-button` with `variant="ghost"`, `variant="text"` or `variant="outline"` mixes its
857
1281
  label toward the surface instead of taking the shared fixed-lightness recipe.** New default:
858
1282
  `color-mix(in srgb, <label> 40%, var(--surface-default))`. These three share one accent source and
859
1283
  have no fill that could carry the state, so the label is the whole signal — and the shared recipe
860
- was failing it, badly in dark. Measured against the enabled label: light **2.54 → 3.41:1**, dark
861
- **1.11 → 2.91:1**. `outline` does have a border, but its disabled border resolves to only
1284
+ was failing it, badly in dark. Against the enabled label, the disabled label goes from **2.54 → 3.41:1**
1285
+ in light and **1.11 → 2.91:1** in dark. `outline` does have a border, but its disabled border resolves to only
862
1286
  **1.30:1** against the page — less than its own *enabled* border at 1.49:1 — so it never covered
863
1287
  for the label.
864
1288
  **The dark number is the real defect.** `--primary-strong` there is already light (OKLCH L 67.9%),
@@ -867,48 +1291,16 @@ workaround written against a hook that only landed in this release.
867
1291
  lighter than the accent; in dark the accent is already light, so disabled has to go darker. The
868
1292
  surface knows that direction and a constant does not — which is the whole reason the mix target is
869
1293
  `--surface-default` rather than a number.
870
- **The trade:** the disabled label is fainter against its own ground — **2.00:1** light, 1.99:1
871
- dark, where the old recipe gave 2.69 / 6.45. Disabled text is exempt from WCAG 1.4.3 and 1.4.11
872
- and Material ships its equivalent at 1.93:1, so this is inside practice. Chroma rises with the
873
- mix (0.036 → 0.099 light), so the disabled label stays tinted rather than going grey; that was
874
- chosen deliberately over a desaturating variant after looking at both.
1294
+ **The trade:** the disabled label is fainter against its own ground — **2.00:1** light, **1.99:1**
1295
+ dark, where the old recipe gave 2.69 / 6.45. Disabled text is exempt from WCAG 1.4.3 and 1.4.11 and Material ships its equivalent at
1296
+ 1.93:1, so this is inside practice. Chroma rises with the mix (0.036 → 0.099 light), so the disabled
1297
+ label stays tinted rather than going grey.
875
1298
  **`solid` and `tonal` are unchanged**, deliberately: a white label mixed into a white surface does
876
1299
  not move, and tonal's disabled fill already carries the state on its own.
877
- **Known limitation:** the mix target is `--surface-default`, a stand-in for "whatever is behind" — CSS cannot read the actual
878
- backdrop, so on a tinted ground (a card at `--surface-secondary`, an alert) the disabled label is
879
- computed against the page colour rather than that ground. Consumers who set
880
- `--ds-button-label-color-disabled` are unaffected.
881
- ### Changed
882
-
883
- - **A disabled `ds-icon-button variant="standard"` mixes its glyph toward the surface too.** It is
884
- the only icon-button variant with neither a fill nor a border, so the glyph's colour is the whole
885
- disabled signal — and the shared fixed-lightness recipe could not carry it in dark, where the
886
- enabled icon is already light (OKLCH L 80%) and a disabled one pinned at L 70% sat **1.46:1** away.
887
- The new default is `color-mix(in srgb, <icon> 40%, var(--surface-default))`, the same mix the
888
- borderless button variants take, which picks its direction per theme: light **4.17 → 5.19:1**, dark
889
- **1.46 → 3.58:1**. Scoped to `standard` deliberately — `filled`'s icon is `--c-on`, and white
890
- mixed into a white surface does not move at all. `--ds-icon-button-icon-color-disabled` is the same
891
- hook as before and still wins; only the default under it changed.
892
-
893
- - **If you EXTEND a DS field component, read this before trusting `class-surface`.** The report
894
- against 0.10.0 says fourteen field components LOST eight members each — `hint`, `error`,
895
- `warning`, `errors`, `invalid`, `touched`, `reserveMessage` and `reserveMessageResolved`.
896
- **Nothing was removed.** Those inputs were declared once per component and are now declared once in
897
- total, on a new internal `DsFieldBase`, which every field reaches either directly or through
898
- `DsControlValueAccessor` / `DsBooleanControlValueAccessor`. All fourteen still carry all eight by
899
- inheritance, so a template binding, a TS read and an existing `override` all behave exactly as
900
- before. The tool counts declarations per FILE, and `migrations/AUTHORING.md` already names a member
901
- that moved to a base class as one of the two things it cannot see — this release is that case.
902
- What genuinely grew is smaller: **eight bordered fields gained `success`** (`ds-autocomplete`,
903
- `ds-color-picker`, `ds-date-picker`, `ds-multi-select`, `ds-otp-input`, `ds-select`,
904
- `ds-textarea`, `ds-time-picker`), `ds-input` having had it all along. `ds-segmented`
905
- is the tenth grown class and the one with real new surface — joining the form systems gave it
906
- `label`, `disabled`, `isDisabled`, `message`, `invalidResolved`, `fieldMessage`, `hintId` and
907
- `uid`. A subclass declaring
908
- `success` on one of those eight now fails with TS4114; declaring any of the eight moved names still
909
- collides exactly as it did, because an inherited member collides the same way an own one does.
910
- `DsFieldBase` is internal and deliberately not exported — it holds only the inputs every field has,
911
- which is why `success`, `label`, `required` and `optional` stayed on the components.
1300
+ **Known limitation:** the mix target is `--surface-default`, a stand-in for "whatever is behind" —
1301
+ CSS cannot read the actual backdrop, so on a tinted ground (a card at `--surface-secondary`, an
1302
+ alert) the disabled label is computed against the page colour rather than that ground. Consumers
1303
+ who set `--ds-button-label-color-disabled` are unaffected.
912
1304
 
913
1305
  ### What you can now DELETE
914
1306
 
@@ -917,13 +1309,14 @@ None of this is required — skipping it leaves working code that is merely redu
917
1309
  | If you have | Delete it, because |
918
1310
  | --- | --- |
919
1311
  | A `?? ''` guard wrapped around a message binding - `[error]="form.errors?.message ?? ''"` - added because a null pushed into `hint` / `error` / `warning` / `success` threw a TypeError during render. | nothing - bind the expression directly, the inputs absorb null now |
1312
+ | A modal body torn down on `(closed)`, or guarded behind `@if (open())`, purely so a re-opened `ds-date-picker` starts clean | The picker clears its own stale error and text on the value-to-text effect now - and the teardown never worked reliably, because reopening inside the 300 ms leave animation cancels the close and `closed` never fires |
920
1313
 
921
1314
  ## 0.10.0 — 2026-08-18
922
1315
 
923
1316
  ### Added
924
1317
 
925
- - **A capped chip trigger no longer grows a second line on a narrow screen.** Found by Stas testing the
926
- docs on a phone: `[maxVisibleChips]="2"` honoured the count, ran out of width, and wrapped the text
1318
+ - **A capped chip trigger no longer grows a second line on a narrow screen.** On a narrow trigger,
1319
+ `[maxVisibleChips]="2"` honoured the count, ran out of width, and wrapped the text
927
1320
  input onto its own line — the exact growth the cap exists to prevent. **A number is now a CEILING**
928
1321
  ("at most N"), measured like `"fit"`, so a 412px trigger shows one chip and `+3 more` on one line
929
1322
  instead of two chips over two lines. `null` is unchanged and still the only mode that skips measuring.
@@ -938,26 +1331,25 @@ None of this is required — skipping it leaves working code that is merely redu
938
1331
  wrapping, so the slack goes to the columns that should absorb it. An auto-layout table spreads
939
1332
  leftover width evenly, which is why a status or a figure ends up as wide as its own header while the
940
1333
  name column that should take the space does not — and a pixel `width` cannot express that, since
941
- pinning one column leaves the rest spreading. Measured in the docs demo: two `'content'` columns come
942
- out at 89px and 94px while the two ordinary ones take 438px and 530px of the same 1152px table.
1334
+ pinning one column leaves the rest spreading — a `'content'` column takes only what its widest cell
1335
+ needs, and the ordinary columns absorb what is left over.
943
1336
  **You can now delete** the `::ng-deep ds-table { .table th, .table td { width: 1%; white-space: nowrap } }`
944
- block — the last reach into the table's markup a consumer had left, still marked in their tree as
945
- `TODO(ds-shrink-to-content)`. Dragging a column's edge still wins, because that is a deliberate
946
- pixel width — but note it wins for the WHOLE table at once: on a `resizableColumns` table the first
947
- drag OR resize keystroke anywhere locks a measured pixel width onto every column that has not got
948
- one, `'content'` columns included. `Home` on a column's own handle hands that one back; the column
949
- panel's "Reset to default" button and clearing the public two-way `columnWidths` hand back all of
950
- them. A `'content'`
951
- column also has no clip width, so `truncate` on it is a no-op rather than an ellipsis against 240px.
1337
+ block — the last reach into a table's markup this idiom required. Dragging a column's edge still
1338
+ wins, because that is a deliberate pixel width — but note it wins for the WHOLE table at once: on
1339
+ a `resizableColumns` table the first drag OR resize keystroke anywhere locks a measured pixel
1340
+ width onto every column that has not got one, `'content'` columns included. `Home` on a column's
1341
+ own handle hands that one back; the column panel's "Reset to default" button and clearing the
1342
+ public two-way `columnWidths` hand back all of them. A `'content'` column also has no clip width,
1343
+ so `truncate` on it is a no-op rather than an ellipsis against 240px.
952
1344
  The `title` attribute and the click-to-popover handler are still emitted on such a cell, but the
953
1345
  popover cannot open: it opens only when the cell measures as clipped, and this one never does.
954
1346
 
955
1347
  - **`ds-table` takes `[loading]="true"` and picks the treatment itself** — skeleton rows while it has
956
1348
  nothing on screen, overlay once there are rows. `'skeleton'` and `'overlay'` still pin either one
957
1349
  deliberately, and `false` is unchanged. Which of the two is right is a function of the row count,
958
- which the table already holds and every consumer was restating: **you can now delete** the
959
- `items().length > 0 ? 'overlay' : 'skeleton'` computed beside each list — one consumer feature had
960
- three copies. The rule is not a taste call, which is why it belongs here: replacing visible rows with
1350
+ which the table already holds, so restating it beside every list is redundant: **you can now
1351
+ delete** the `items().length > 0 ? 'overlay' : 'skeleton'` computed beside each list. The rule is
1352
+ not a taste call, which is why it belongs here: replacing visible rows with
961
1353
  placeholders throws away what the reader is looking at, and dimming an empty card announces nothing
962
1354
  about what is coming, so a hardcoded `'overlay'` makes a first load read as an empty table that
963
1355
  flickers.
@@ -965,16 +1357,15 @@ None of this is required — skipping it leaves working code that is merely redu
965
1357
  - **`ds-table` has a toolbar row you can put things in — one `[toolbar]` slot.** The component had two
966
1358
  projection slots in total (`[empty]`, `[bulk-actions]`) and its toolbar existed only while
967
1359
  `exportButton` was on, so a control that acts on THIS table — a search box, a status filter, a refresh
968
- — had nowhere to sit next to Export. A consumer forked the table for exactly this; retiring the fork
969
- pushed two screens' filter rows out of the table and onto the page, where a search + status pair reads
970
- as page furniture one card away from the rows it narrows. **Project a handful of controls and the row
971
- lays them out** (flex, wrap, 8px gap), so the ordinary case needs no CSS of yours; **project one
972
- element of your own** and it stretches to fill the slot, so your layout is the only one in play — the
973
- slot is the row minus the Export button, and the whole row when `exportButton` is off. The row
974
- **collapses to nothing** when nothing filled it and `exportButton` is off, so a table that uses
975
- neither is pixel-identical — the `div` is still in the DOM, hidden in CSS, so a
976
- `ds-table > :first-child` selector or a DOM snapshot still sees it. Filters that belong to the PAGE
977
- still belong in a `ds-filter-bar` above it.
1360
+ — had nowhere to sit next to Export, so a search + status pair ended up living on the page above the
1361
+ table, reading as furniture one card away from the rows it narrows instead of a part of the table
1362
+ itself. **Project a handful of controls and the row lays them out** (flex, wrap, 8px gap), so the
1363
+ ordinary case needs no CSS of yours; **project one element of your own** and it stretches to fill
1364
+ the slot, so your layout is the only one in play — the slot is the row minus the Export button,
1365
+ and the whole row when `exportButton` is off. The row **collapses to nothing** when nothing filled
1366
+ it and `exportButton` is off, so a table that uses neither is pixel-identical — the `div` is still
1367
+ in the DOM, hidden in CSS, so a `ds-table > :first-child` selector or a DOM snapshot still sees
1368
+ it. Filters that belong to the PAGE still belong in a `ds-filter-bar` above it.
978
1369
 
979
1370
  - **`runExport()`, `exporting()` and `exportStatus()` are public**, so the Export button can live in a
980
1371
  page header, a filter bar or the new toolbar slot and still run the library's path — the
@@ -997,10 +1388,9 @@ None of this is required — skipping it leaves working code that is merely redu
997
1388
  (`table.bordered`, `table.stickyScrollbar`, `badge.withDot`, and `dividers` on both modal components)
998
1389
  used to read a binding evaluating to `null`/`undefined` as `false`, and now read it as *unset* — so it
999
1390
  falls through to the config and then the built-in, which for `dividers` is `true`. A literal
1000
- `[bordered]="false"` is unaffected. **You can now
1001
- delete** the attribute an app repeats on every element: one consumer binds `[stickyScrollbar]="true"`
1002
- on all 35 of its tables and `size="sm"` on 146 buttons, which is a house default written 146 times,
1003
- and each new element is one forgotten binding away from looking wrong. `control.size` is a different
1391
+ `[bordered]="false"` is unaffected. **You can now delete** the attribute an app repeats on every
1392
+ element — a house default rewritten on every table or button — since each new element is one
1393
+ forgotten binding away from looking wrong. `control.size` is a different
1004
1394
  axis from `[data-density]`: the tier rewrites hit targets and padding in CSS for a route, this picks
1005
1395
  which of the three declared sizes a control renders at, and an app can use both. Documented as one
1006
1396
  page under Foundations → Theming rather than in scraps across eight component pages.
@@ -1016,8 +1406,7 @@ None of this is required — skipping it leaves working code that is merely redu
1016
1406
  row carries `.is-disabled` and `aria-disabled`, refuses the click, and **the arrow keys step OVER it** —
1017
1407
  including when the panel opens or the filter re-runs, so the active row is never one that Enter would
1018
1408
  ignore. Behaviourally this matches `ds-select` and `ds-multi-select`, which have honoured it all along.
1019
- The row greys out too: measured in the open panel it comes out `rgb(155,163,180)` against
1020
- `rgb(22,26,40)` for an enabled one, so the state is not subtle.
1409
+ The row greys out too, so the state is not subtle.
1021
1410
 
1022
1411
  - **`maxVisibleChips="fit"` on `ds-autocomplete` and `ds-multi-select`** caps the chips by measured
1023
1412
  width instead of a count: one line holds as many as the trigger is wide, the rest collapse into
@@ -1033,29 +1422,16 @@ None of this is required — skipping it leaves working code that is merely redu
1033
1422
  trigger, empty or not — widths are cached per selection, and a geometry guard skips the re-measure
1034
1423
  when neither width nor count moved.
1035
1424
 
1036
- - **`--ds-dropdown-panel-radius` — the lever 0.9.0's notes promised and never shipped.** That release
1037
- said an 8px select panel could be restored with `--ds-select-panel-radius`; no such token existed,
1038
- so the advice did nothing. This is the honest version, and it is
1039
- panel-level rather than per-component on purpose: retuning one picker's corner is not a thing to
1425
+ - **`--ds-dropdown-panel-radius` — a hook for the dropdown corner.** It is panel-level rather
1426
+ than per-component on purpose: retuning one picker's corner is not a thing to
1040
1427
  encourage, while `--ds-radius-panel` moves modal, popover and date picker too. Defaults to that
1041
1428
  role, so an untouched theme is unchanged, and it is forwarded onto the portaled pane — scope it with
1042
- a selector (`ds-select { --ds-dropdown-panel-radius: 8px }`) and only that panel moves. Verified
1043
- in a browser: one select at 3px, the other 34 on the 12px role. **Two panels read it, not three:**
1429
+ a selector (`ds-select { --ds-dropdown-panel-radius: 8px }`) and only that panel moves.
1430
+ **Two panels read it, not three:**
1044
1431
  `ds-select` and `ds-autocomplete`. `ds-multi-select`'s panel hardcodes `--radius-md` and is reached by
1045
1432
  neither this token nor `--ds-radius-panel` — so only `--radius-md` itself moves it, and that moves
1046
1433
  buttons, fields and tooltips with it.
1047
1434
 
1048
- - **Two migration manifests were corrected and three more were written**, so re-read them if you
1049
- subclass a DS component or upgrade across a version you skipped. `0.9.0`'s named no member collisions
1050
- at all — the note that now fills the gap lists **fifteen** classes whose member surface had grown, plus
1051
- four newly exported symbols that can shadow a local name — and `0.8.0`'s missed the private
1052
- `saveLayout` that actually stopped a build (`restoreLayout` went in beside it). **`0.6.1`, `0.6.2` and
1053
- `0.9.1` had no manifest at all**, and now every released version carries one: the auto-update planner
1054
- prefers a manifest and otherwise re-derives the plan from changelog prose on every run, which is what
1055
- an upgrade crossing those three used to do. `0.6.2`'s turned out to hold something actionable that no
1056
- machine-readable file had ever stated — a `ds-bottom-sheet` with nothing focusable inside it stopped
1057
- closing on Escape — which is exactly the kind a "just fixes" patch is assumed not to have.
1058
-
1059
1435
  - **`ds-avatar` takes a `name`, and the rule that reads it is replaceable.** One binding now does both
1060
1436
  jobs a caller used to wire by hand: it derives the initials and seeds the tint. **The tint hashes the
1061
1437
  NAME, not the initials** — two letters collide wholesale, so Alice Brown and Adam Bell used to share a
@@ -1072,15 +1448,14 @@ None of this is required — skipping it leaves working code that is merely redu
1072
1448
  - **`DS_CONFIG.brand` — set the brand once instead of on every screen.**
1073
1449
  `provideDsConfig({ brand: { name, logoLetter, tagline } })` answers `ds-sidenav` and `ds-auth-layout`
1074
1450
  together — `name` and `logoLetter` on both, `tagline` on the auth layout, which is the only one of the
1075
- two with a strapline row. They ask the same questions and an app answers them identically, so six auth
1076
- screens were binding the same values six times. Each field resolves on its own, so a screen can
1451
+ two with a strapline row. They ask the same questions and an app answers them identically, so
1452
+ restating them per screen is redundant. Each field resolves on its own, so a screen can
1077
1453
  override the name without restating the letter — and `brand=""` drops the wordmark while a
1078
- `logoLetter` in the config
1079
- still paints its tile, so clear both to give a screen no brand row. **The two-tone wordmark stays
1080
- projected:** there is deliberately no accent in the config, because a string would have to be split
1081
- back apart to find one and that is the guessing `[dsBrandName]` exists to avoid. The projection wins
1082
- over both. **You can now delete** an adapter component that existed to inject a brand object and fan
1083
- it out into four bindings.
1454
+ `logoLetter` in the config still paints its tile, so clear both to give a screen no brand row.
1455
+ **The two-tone wordmark stays projected:** there is deliberately no accent in the config, because
1456
+ a string would have to be split back apart to find one and that is the guessing `[dsBrandName]`
1457
+ exists to avoid. The projection wins over both. **You can now delete** an adapter component that
1458
+ existed to inject a brand object and fan it out into four bindings.
1084
1459
 
1085
1460
  - **`rememberLastTab` on `ds-tabs` — reopen the tab a screen was last left on.** Off by default, and
1086
1461
  inert without `urlParam` (and outside a routed context), because the parameter supplies the names: a
@@ -1096,7 +1471,7 @@ None of this is required — skipping it leaves working code that is merely redu
1096
1471
  `provideDsConfig({ modal: { dividers: false } })` instead of a binding on every `<ds-modal>`, with an
1097
1472
  explicit `[dividers]` on an element still winning. It also reaches where an attribute cannot: a
1098
1473
  `ModalService` confirm renders a `ds-modal-card` with no `ds-modal` ancestor to bind, and the injected
1099
- default gets there (verified — the confirm's card comes out `no-dividers`). One limit on that: the
1474
+ default gets there too, and the confirm's card comes out `no-dividers`. One limit on that: the
1100
1475
  service is `providedIn: 'root'` and builds the confirm's injector from the root, so only a ROOT-level
1101
1476
  `provideDsConfig` reaches it — a route- or component-scoped one does not. **You can now delete** a
1102
1477
  `dividers` binding repeated per modal, and any `!important` reach into `.modal > .header|.body|.actions`
@@ -1124,36 +1499,184 @@ None of this is required — skipping it leaves working code that is merely redu
1124
1499
  - **`--ds-stat-value-white-space` and `--ds-stat-value-font-size`, for a tile whose value is a phrase.**
1125
1500
  The value line is `nowrap`, which is right for the number `ds-stat` was built around and wrong for a
1126
1501
  date range or a status — so a text value spilled straight out past the tile's edge, and the way out
1127
- was the projected
1128
- `[dsStatValue]` slot plus two declarations undoing the component. Both default to today's values, so
1129
- nothing moves until you set them. **The size hook works at every `size`**, not just the default: the
1130
- `sm`/`lg` variants now move the same token instead of restating `font-size`, which used to out-rank
1131
- any override. **You can now delete** a `[dsStatValue]` span that exists only to set `white-space`
1132
- or a smaller size.
1502
+ was the projected `[dsStatValue]` slot plus two declarations undoing the component. Both default
1503
+ to today's values, so nothing moves until you set them. **The size hook works at every `size`**,
1504
+ not just the default: the `sm`/`lg` variants now move the same token instead of restating
1505
+ `font-size`, which used to out-rank any override. **You can now delete** a `[dsStatValue]` span
1506
+ that exists only to set `white-space` or a smaller size.
1133
1507
 
1134
1508
  - **The global `.link` class now works on a `<button>`, which it has been documented as doing all
1135
1509
  along.** `ds-button` warns `variant="text"` off in favour of `.link`, and an action that is not
1136
1510
  navigation has to be a button — but the class was written for anchors and reset none of the browser's
1137
- button chrome, so everyone who took that advice wrote the same rule by hand. Measured against an
1138
- `<a class="link">` in the same paragraph, a button came out at 13.3px inside 14px text, with
1139
- `line-height: normal`, `1px 6px` of padding and an outset border; all of that is gone. **You can now
1140
- delete** a local `.link-button`, or the padding/border/background/font reset in your overrides.
1511
+ button chrome, so everyone who took that advice wrote the same rule by hand. A `<button
1512
+ class="link">` used to keep its native chrome — `line-height: normal`, `1px 6px` of padding and an
1513
+ outset border — all of that is gone, so it now renders identically to an `<a class="link">` in the
1514
+ same paragraph. **You can now delete** a local `.link-button`, or the
1515
+ padding/border/background/font reset in your overrides.
1141
1516
  `font-weight` is untouched so `.quiet` and `.standalone` still decide it, and `display` stays
1142
1517
  `inline-block` so a standalone button lays out predictably. The reset is scoped with `:where()`, so
1143
1518
  it sits at `.link`'s own specificity and your rule still wins on source order.
1144
1519
 
1520
+ ### Changed
1521
+
1522
+ - **`avatarTint` switches from OKLCH to `hsl(H 50% 62%)` with white initials, always.** The seeded
1523
+ tints read as dull, and every set of initials came out black. The auto-contrast was not the fault:
1524
+ measured across the hue circle, the old `oklch(62% 0.14 h)` put every hue's luminance between 0.218
1525
+ and 0.271 against a 0.179 black/white crossover, so black was genuinely the better pick every time
1526
+ (5.4–6.5:1) — the switch was right, the lightness never let it switch. The new recipe was adopted deliberately: HSL lightness is not perceptually even, so yellows land far lighter than
1527
+ blues, and that unevenness is exactly what makes the row read as bright.
1528
+ **It is a knowing trade against contrast.** White on the lighter hues measures as little as
1529
+ **1.65:1**, under the 4.5:1 WCAG AA asks of text, and no hue across the full 20–300° range clears
1530
+ it. Suitable where the avatar sits beside a name that is written out anyway;
1531
+ where the initials are the only identifier, pass your own `lightness` / `foreground`.
1532
+ **API:** the `chroma` option is gone, replaced by `saturation` (HSL %, default 50). `lightness` keeps
1533
+ its name and its default of 62 but now means **HSL** lightness, so an existing numeric override will
1534
+ land somewhere different — re-pick it rather than assuming it carried over. `foreground` now defaults
1535
+ to `#fff` instead of auto-choosing; pass it explicitly to get anything else. `avoidHues` is unchanged.
1536
+
1537
+ - **`ds-input`'s `floating` label is removed, with no replacement.** It only ever did half the job: the
1538
+ label was drawn inside a taller control, but it never moved out of the field on focus — the animation
1539
+ from inside-the-box to above-it *is* the pattern, and without it an empty field reads as a
1540
+ permanently mislabelled one rather than a floating-label one. Completing it is a piece of motion work this library is not doing, so it goes. **Migration is deleting the attribute:**
1541
+ `[floating]="true"` → nothing, and the label renders above the field like every other field in the
1542
+ library. One thing to re-check when you do: a floating input stood at `--hit-touch` (48px at default
1543
+ density) and now takes the ordinary size scale (`--field-h-md` → `--hit-cozy`, 40px at default
1544
+ density), so a row whose height was built around one moves. Gone with it are the five internal
1545
+ tokens only this variant read (`--field-h-floating`, `--field-floating-label-fs`,
1546
+ `--field-floating-label-lh`, `--field-floating-label-top`, `--field-floating-pad-t`) — overriding
1547
+ them is now inert.
1548
+ `--ds-elevation-floating` is a different token for toasts and is untouched.
1549
+
1550
+ - **`ds-table`'s `sortKey` and `sortDir` are an `input` + `output` pair instead of a `model`** — the
1551
+ mechanism behind the sort-indicator fix under **Fixed** below, because a `model` cannot tell a host's
1552
+ write apart from the
1553
+ table's own. **Templates are unaffected:** `[(sortKey)]` still works (`sortKeyChange` and
1554
+ `sortDirChange` are emitted), one-way `[sortKey]` with `(sortChange)` still works, and a static
1555
+ `sortKey="project"` still works. What breaks is reaching into the component in TypeScript:
1556
+ `table.sortKey.set(…)` no longer exists — nor `sortDir.set(…)` — so drive it from your own state and
1557
+ bind it, which is what every server-paged list already does. Reading `table.sortKey()` still works and
1558
+ now returns what the host bound; the marked state the header renders is internal. **One emission rule
1559
+ moved with it:** both `…Change` outputs now fire on EVERY sort click, where a `model` suppressed the
1560
+ emit when the value was unchanged. `[(sortKey)]` is unaffected — it rebinds the same value — but a
1561
+ `(sortKeyChange)="reload()"` handler now also runs when only the DIRECTION changed. Listen to
1562
+ `(sortChange)` instead, which carries both halves once.
1563
+
1564
+ - **`ds-modal` / `ds-modal-card`: `error: boolean` → `status: DsStatus`.** A boolean could not express
1565
+ a warning or an info dialog, and the rest of the library settled on the `DsStatus` vocabulary in
1566
+ 0.9.0. **Migration is one attribute:** `[error]="true"` → `status="error"`, and a computed
1567
+ `[error]="isDangerous"` → `[status]="isDangerous && 'error'"`, since the input takes `false` and
1568
+ `undefined` as "no status" — the longer `? 'error' : null` ternary works too. Unset stays neutral,
1569
+ so a dialog that never used `error` looks identical. **What you gain beyond the rename:** a status now
1570
+ drives the icon-badge tint, the title colour, a DEFAULT icon and the ARIA role — the same four
1571
+ things `ds-alert`'s variant drives, so a status dialog and a status alert on one screen cannot
1572
+ disagree. The default icon matters: a tint alone signals by colour only, which fails the
1573
+ don't-rely-on-colour rule. `icon` still overrides the pick, and `error`/`warning` default the
1574
+ dialog to `alertdialog`. `dialogRole` lives on `ds-modal-card` only — it is now
1575
+ `'dialog' | 'alertdialog' | null`, defaulting to `null` rather than `'dialog'` so "unset" is
1576
+ distinguishable from a deliberate `dialog`, and an explicit value still wins. `ds-modal` has no such
1577
+ input, so `<ds-modal status="error">` becomes an `alertdialog` with nothing to override it. What
1578
+ renders is unchanged unless you set a status. Three tokens come with the tints:
1579
+ `--ds-modal-title-text-color-warning`, `-success` and `-info`, joining the `-error` that already
1580
+ existed. **One thing rides along that no attribute of yours controls:** a
1581
+ `ModalService.confirm({ tone: 'error' })` passes `status="error"` down internally, and a status now
1582
+ supplies a default icon — so a destructive confirm that never set `options.icon` opens with a red
1583
+ error badge where it previously had none. Pass `icon` to choose a different glyph; there is no input
1584
+ that turns the badge back off.
1585
+
1586
+ - **Chips cap by width by default.** `maxVisibleChips` on `ds-autocomplete` and `ds-multi-select`
1587
+ now defaults to `"fit"` instead of "show everything": the trigger keeps as many chips as it is
1588
+ wide and collapses the rest into "+N more". A trigger that grew a line per selection was pushing
1589
+ the rest of the form down the page. **To keep the old behaviour, bind `[maxVisibleChips]="null"`.**
1590
+ Two more things moved with the default, and both are easy to miss. A trigger in any measuring mode is
1591
+ now single-line (`flex-wrap: nowrap`), and its chips are held to one line each and clipped rather than
1592
+ wrapped — so a chip carrying more than a short label gets squeezed, where `null` still wraps the row.
1593
+ And **`[maxVisibleChips]="0"` has inverted**: it used to mean "show them all" and now means "show
1594
+ none", putting the whole selection behind `+N more`. If you have a `0` anywhere, it was almost
1595
+ certainly written for the old meaning; `null` is what that meant.
1596
+
1597
+ - **`size` is now `DsSize | null` on the fourteen controls that resolve `DS_CONFIG.control.size`**, so
1598
+ an unbound `size` can fall through to it. `ds-drawer` and `ds-spinner` also take a `size` and are NOT
1599
+ among them — theirs stays a non-null `DsSize` defaulting to `'md'`, so a grep for `DsSize` finds
1600
+ sixteen controls and only fourteen changed. **Nothing changes in a template** — `size="sm"` binds as
1601
+ before and an unbound control still renders `md`. **What changes is reading it back:** `size()` now
1602
+ returns `null` when nothing was bound, where it used to return `'md'`. The same is true of
1603
+ `ds-card`'s `elevation` and `padding`, `ds-icon`'s `size`, `ds-badge`'s `withDot`, `ds-table`'s
1604
+ `bordered` and `stickyScrollbar`, `dividers` on both modal components, `dialogRole` on
1605
+ `ds-modal-card`, `brand` and `logoLetter` on `ds-sidenav` and `ds-auth-layout`, and `tagline` on the
1606
+ auth layout — all of them now answer `null` for "nothing bound" rather than the old default.
1607
+ **There is no public accessor for the resolved value**, and this is the
1608
+ awkward part of the release: each component holds a `sizeResolved()` (or `elevationResolved()`,
1609
+ `withDotResolved()`, …) but declares it `protected`, so reading it from a host, a test or a
1610
+ `viewChild` does not compile. `ds-field-group` is the single exception, where `sizeResolved` is
1611
+ public. Until that is opened up, a host that needs to know what a control is rendering at should keep
1612
+ the value it bound, or resolve `DS_CONFIG.control.size` itself — the fallbacks are unchanged, but
1613
+ they are not all `false`: `'md'` for `size`, `'flat'` for `elevation`, `16` for the icon size,
1614
+ `false` for `bordered` / `stickyScrollbar` / `withDot`, **`true`** for `dividers`, `''` for the brand
1615
+ strings, and the card's own padding when `padding` is unset. **If you implement
1616
+ `DsFieldGroupContext`**, its member is now `sizeResolved` rather than `size`, for the same reason:
1617
+ a segment must not have to resolve the group's size a second time.
1618
+
1619
+ - **`[data-density="ultra-compact"]` now compresses surface padding too** — `--ds-container-pad`,
1620
+ the inner padding `ds-card` and `ds-stat` share, drops from 24px to 16px in that tier only.
1621
+ Everything else is untouched: the default and the other three tiers still resolve to 24px. It was
1622
+ the one padding the tier missed — it already compressed field padding and table cells, so a card
1623
+ kept its full 24px frame while its own cells halved. **If you want the old 24px inside an
1624
+ ultra-compact container**, re-declare `--ds-container-pad: var(--space-6)` on it. For a value
1625
+ between the rungs — 20px, say — set that one token at `:root`; both components default to it, which
1626
+ is why neither `ds-card` nor `ds-stat` should be pinned on its own.
1627
+
1628
+ - **`ds-tabs` `urlParam` now SLUGS a tab's label instead of using it raw.** A tab with no `value` used
1629
+ to put its display string straight into the address — `?tab=License%20Verification%20Status`. It is now
1630
+ kebab-cased and lowercased with diacritics dropped: `?tab=license-verification-status`. **If you
1631
+ adopted `urlParam` in 0.9.0 with tabs that declare no `value`, your existing links change** — an old
1632
+ one no longer matches and lands on the first tab, which is what a stale bookmark has always done.
1633
+ Declare `value` on each tab to pin the names for good: a slug is still derived from copy, so renaming a
1634
+ tab still moves the link. What `tabClosed` emits is unchanged — still the raw label.
1635
+
1636
+ - **`ds-auth-layout`'s projected `[brand]` slot no longer forces `width: auto`.** It shipped in 0.9.0
1637
+ alongside the slot's `max-height`, but at `(0,2,0)` it out-ranks a width the consumer set on their
1638
+ own logo, and with the default `max-height: none` it never did anything: an `<img>` with no CSS
1639
+ width already keeps its ratio under `max-height`. If you opted into the max-height AND set an
1640
+ explicit width on your logo, add `width: auto` yourself to keep the old scaling.
1641
+
1642
+ - **The chips of both pickers now render inside a `<span class="chip-row">`** rather than directly
1643
+ in the trigger — the element the width measurement needs. **On `[maxVisibleChips]="null"` it is a
1644
+ wrapper and nothing more**: `display: contents`, so it adds no box and the chips stay flex items of
1645
+ the trigger, identical in spacing, wrapping and height. In the measuring modes — which includes the
1646
+ new default — the row becomes a `nowrap` flex box that clips, which is how the cap is enforced.
1647
+ Either way, what breaks is a CSS override that walks the structure:
1648
+ `.autocomplete-trigger > ds-tag`, `:nth-child` written against the trigger's children, or
1649
+ `ds-tag + .autocomplete-input` adjacency — the chips are no longer siblings of the filter input,
1650
+ the addon slots or the chevron. Drop the combinator or insert `.chip-row`. Multi-select's trigger is
1651
+ `.field-trigger`, not `.trigger`, if you are rewriting a selector against it. Chips carry `data-chip`
1652
+ and the marker `data-chip-overflow`; the measurement finds them that way, so leave them in place.
1653
+
1654
+ - **If you EXTEND a DS component, this is the release that will break you.** Against 0.9.1,
1655
+ **30 exported classes grew their member surface** — among them `ButtonComponent`,
1656
+ `TableComponent`, `ModalComponent`, `SearchComponent`, `AvatarComponent`, `IconComponent`, both
1657
+ pickers, every bordered field that gained `controlState`, and every control that gained
1658
+ `sizeResolved`. A subclass declaring any of those names now fails with TS4114. **Four classes also
1659
+ LOST members**, which is the case a subclass notices differently — an `override` that now overrides
1660
+ nothing: `ModalComponent.error` and `.onBackdrop`, `ModalCardComponent.error`,
1661
+ `BottomSheetComponent.stop` and `InputComponent.floating`. Subclassing a DS component remains
1662
+ unsupported, so the full list is stated rather than a sample. **Two kinds of break a member-list diff cannot see**, so check them by hand:
1663
+ a member whose SIGNATURE changed keeps its name — `TableComponent.exporting` and `.exportStatus`
1664
+ are still there, but as public readonly signals rather than the protected writables a subclass
1665
+ could `.set()` — and a plain `private pending: … = null` is not a field the parser recognises, which
1666
+ is how `SearchComponent`'s `pending` and `pushed` sit outside its reported seven.
1667
+
1145
1668
  ### Fixed
1146
1669
 
1147
1670
  - **A card no longer eats the Enter and Space of everything inside it.** `ds-card` bound
1148
1671
  `(keydown.enter)` and `(keydown.space)` on its `<article>` with an unconditional `preventDefault()`,
1149
1672
  and keydown bubbles — so a card cancelled the DEFAULT ACTION of both keys for its whole subtree, on
1150
1673
  every card, not only the `elevation="interactive"` ones that are supposed to answer a key at all.
1151
- Measured on a sign-in screen made of `ds-card` + `ds-input` + `ds-checkbox`, driven with real key
1152
- events: Enter in the password field sent **no request at all** — the form could only be submitted by
1153
- clicking the button — Space did not toggle the "Remember me" box, Enter on a link inside the card did
1154
- not follow it, and a space could not be typed into a text field, so `John` + Space + `Smith` came out
1155
- as `JohnSmith`. Both handlers now run only on an interactive card, and only when the key landed on
1156
- the card itself rather than on something projected into it.
1674
+ On a card built from `ds-card` + `ds-input` + `ds-checkbox`, Enter in a password field sent **no
1675
+ request at all** — the form could only be submitted by clicking the button — Space did not toggle a
1676
+ "Remember me" checkbox, Enter on a link inside the card did not follow it, and a space could not be
1677
+ typed into a text field, so `John` + Space + `Smith` came out as `JohnSmith`. Both handlers now run
1678
+ only on an interactive card, and only when the key landed on the card itself rather than on
1679
+ something projected into it.
1157
1680
 
1158
1681
  - **A table row's own controls answer the keyboard again.** `onRowKey` consumed Enter and Space for
1159
1682
  every target inside an interactive `<tr>`, so the row's select checkbox could not be toggled with
@@ -1172,22 +1695,22 @@ None of this is required — skipping it leaves working code that is merely redu
1172
1695
  `Other: ___` row) also keeps its own keys now.
1173
1696
 
1174
1697
  - **A field in error no longer tints itself inside a `ds-field-group`, and an invalid group draws a red
1175
- border rather than a near-black one.** Two faults the field-state work left behind, both found by an
1176
- adversarial re-read of this changelog. The state rules are included at the end of each stylesheet and
1177
- score the same `(0,3,0)` as the `.in-group` block that strips a segment's chrome, so source order
1178
- alone decided: measured, the border stayed 0 (`border: 0` zeroes the width, so a later `border-color`
1179
- paints nothing) but the FILL came back, tinting one segment pink mid-row — the exact thing a group
1180
- exists to prevent. All three states now carry `:not(.in-group)`. Separately `ds-field-group` itself
1181
- read `var(--border-error)`, **a token declared nowhere**, so `[invalid]` fell through to
1182
- `currentColor`; it now resolves through a new `--ds-field-group-border-color-error`, defaulting to
1183
- `--error`, matching every other colour the component publishes.
1698
+ border rather than a near-black one.** Two faults sat behind it: the state rules are included at the
1699
+ end of each stylesheet and score the same `(0,3,0)` as the `.in-group` block that strips a segment's
1700
+ chrome, so source order alone decided — the border stayed 0 (`border: 0` zeroes the width, so a later
1701
+ `border-color` paints nothing) but the FILL came back, tinting one segment pink mid-row — the exact
1702
+ thing a group exists to prevent. All three states now carry `:not(.in-group)`. Separately
1703
+ `ds-field-group` itself read `var(--border-error)`, **a token declared nowhere**, so `[invalid]`
1704
+ fell through to `currentColor`; it now resolves through a new
1705
+ `--ds-field-group-border-color-error`, defaulting to `--error`, matching every other colour the
1706
+ component publishes.
1184
1707
 
1185
1708
  - **`ds-timeline-item`'s marker sits level with its heading.** The node is 28px and the heading's line
1186
1709
  box is 19.6px, and both started at the row's top edge — so the circle's centre sat **4.2px** below the
1187
1710
  heading's and read as floating between the heading and the time beneath it. The node column is now
1188
1711
  pulled up by half the difference, computed from the same `--font-size-s` / `--line-height-base` pair
1189
- the heading uses, so retuning the heading moves the marker with it. Measured across all ten items on
1190
- the docs page: every one now aligns to 0.
1712
+ the heading uses, so retuning the heading moves the marker with it. Every marker now aligns exactly
1713
+ with the heading's centre.
1191
1714
 
1192
1715
  - **The collapsed `ds-sidenav` stopped squashing its items into a lozenge.** Every item in the rail
1193
1716
  measured **20×40** where the rail gives it 40×40, so the active item's blue tile rendered as a
@@ -1225,8 +1748,7 @@ None of this is required — skipping it leaves working code that is merely redu
1225
1748
  joins the set (86 now), available to anything else that wants one. The row also transitions its
1226
1749
  colour over `--duration-base` instead of snapping, because the reader is looking at the field they
1227
1750
  are typing in, so an instant change says a rule flipped but not which one. No reduced-motion guard
1228
- is needed: the duration tokens are already zeroed globally. Row spacing is unchanged at
1229
- `--space-1` — deliberately, after comparing both.
1751
+ is needed: the duration tokens are already zeroed globally. Row spacing is unchanged at `--space-1`.
1230
1752
 
1231
1753
  - **`ds-password-rules` printed its screen-reader text on screen.** Every row rendered a visible
1232
1754
  "— not met" / "— not yet met" beside the label, because the component used `.visually-hidden` and
@@ -1248,10 +1770,10 @@ None of this is required — skipping it leaves working code that is merely redu
1248
1770
  did. Pasting a full code is unaffected — it cancels the native input, so nothing ever reaches the box.
1249
1771
 
1250
1772
  - **A `warning` now shows on the control, and one rule drives all nine bordered fields.** `warning="…"`
1251
- coloured the message and left the box untouched — which is what a phone screenshot caught. On 0.9.1
1252
- six stylesheets styled `.warning` and not one belonged to a bordered field: `ds-field-message`, the
1253
- message this describes, plus `ds-alert`, `ds-banner`, `ds-progress`, `ds-sidenav-footer` and
1254
- `ds-toast`. It now takes a `--warning-text` border. Not `--warning`:
1773
+ coloured the message and left the box untouched. On 0.9.1 six stylesheets styled `.warning` and
1774
+ not one belonged to a bordered field: `ds-field-message`, the message this describes, plus
1775
+ `ds-alert`, `ds-banner`, `ds-progress`, `ds-sidenav-footer` and `ds-toast`. It now takes a
1776
+ `--warning-text` border. Not `--warning`:
1255
1777
  that is `--yellow-400`, **1.47:1** against a white field where WCAG 1.4.11 asks 3:1 for a component
1256
1778
  boundary, so the obvious mirror of the error recipe would have drawn a border nobody can see;
1257
1779
  `--warning-text` is the first rung of the ramp that clears it, at 4.59:1. Border only, deliberately —
@@ -1294,11 +1816,11 @@ None of this is required — skipping it leaves working code that is merely redu
1294
1816
  the five.
1295
1817
 
1296
1818
  - **Binding `error` now tints the control, not just the message.** Every field that takes an `error`
1297
- string — all 14, via the shared `resolveFieldMessage` — put a red message under a box that stayed
1298
- neutral grey, because the invalid state was computed from `[invalid]` and from validator `errors` but
1299
- not from an explicit `error`. That contradicted the input's own contract ("binding it says the
1300
- consumer has already decided this field is in error") and it is what a phone screenshot caught: a red
1301
- line under an untouched field. `error="…"` now drives the red border, the tinted fill, the error focus
1819
+ string — all 14 — put a red message under a box that stayed neutral grey, because the invalid
1820
+ state was computed from `[invalid]` and from validator `errors` but not from an explicit `error`.
1821
+ That contradicted the input's own contract ("binding it says the consumer has already decided this
1822
+ field is in error") — the message showed in red while the box beneath it stayed untouched.
1823
+ `error="…"` now drives the red border, the tinted fill, the error focus
1302
1824
  ring and `aria-invalid`, exactly as `[invalid]="true"` always did — the border and fill on the nine
1303
1825
  bordered fields, and `aria-invalid` plus the message state on the other five, which have no box to
1304
1826
  paint. **If you were binding both**
@@ -1306,33 +1828,28 @@ None of this is required — skipping it leaves working code that is merely redu
1306
1828
  the box staying neutral**, that is now the error box — drop to `warning` for an advisory message, or
1307
1829
  bind `hint` and leave `error` empty.
1308
1830
 
1309
- - **A field inside `ds-field-group` no longer pins its value to the top of the row.** Found by Stas
1310
- testing the docs on a phone, and it affected **every segment except `ds-button`**: the value sat
1311
- against the top edge instead of the vertical centre — **5px out at `sm`, 9px at `md`, 13px at `lg`**,
1312
- measured in one sitting against the 21.6px content height the box fell back to on the 32/40/48 row
1313
- scale. The group stretches its segments to one row height, but a segment's host is a block, so the
1314
- `height: 100%` each one set on its bordered box resolved against `auto` — the box kept that content
1315
- height, and the value with it. `ds-button`
1316
- escaped the defect only because its host already declared `display: inline-flex`; its block was
1317
- rewritten with the rest all the same, to `flex: 1; height: auto`. Every segment now centres at all
1318
- three sizes. The six `.in-group` blocks — identical declaration for declaration apart from the box's
1319
- class name, which is
1831
+ - **A field inside `ds-field-group` no longer pins its value to the top of the row.** Every segment
1832
+ except `ds-button` was affected: the value sat against the top edge instead of the vertical centre —
1833
+ **5px out at `sm`, 9px at `md`, 13px at `lg`** — against the 21.6px content height the box fell back
1834
+ to on the 32/40/48 row scale. The group stretches its segments to one row height, but a segment's
1835
+ host is a block, so the `height: 100%` each one set on its bordered box resolved against `auto` —
1836
+ the box kept that content height, and the value with it. `ds-button` escaped the defect only
1837
+ because its host already declared `display: inline-flex`; its block was rewritten with the rest
1838
+ all the same, to `flex: 1;
1839
+ height: auto`. Every segment now centres at all three sizes. The six `.in-group` blocks —
1840
+ identical declaration for declaration apart from the box's class name, which is
1320
1841
  why the defect sat in five of them at once — are now one `field-group-segment` mixin, so the next
1321
1842
  change to segment chrome cannot land in some of them and miss the rest.
1322
1843
 
1323
- - **`components.md` had drifted, and it ships inside the package.** The file `tools/build-schematics.mjs`
1324
- copies into `dist/ds/` for a consumer's own agents to read was still documenting the API and token names
1325
- of two renames ago: **46 of the 92 `--ds-*` tokens it named no longer exist** (`--ds-button-fg`,
1326
- `--ds-sidenav-item-bg-hover`, the whole `--ds-pagehdr-*` family, …), `ds-button` was described with its
1327
- pre-2026-06 single `variant` axis (`'primary' | 'secondary' | 'tertiary' | …`) instead of
1328
- `color` × `variant`, and the topbar section documented a `--ds-topbar-margin-b` that **has never
1329
- existed** — telling readers the topbar owns the gap below it and that pages must not add their own.
1330
- Every name in the body is now the shipped one, verified against `src/lib` rather than assumed, and the
1331
- two paragraphs are rewritten from the components' actual API. The preamble of its `## Components`
1332
- section still carries the older warning that the `ds-button` section below is out of date — which now
1333
- contradicts the corrected section, and is itself wrong twice over: it calls `--ds-button-bg-color` a
1334
- dead pre-rename name when that is the shipped token this very release tells you to set. Being cleared
1335
- separately.
1844
+ - **`components.md` had drifted, and it ships inside the package.** The digest a consumer's own
1845
+ agents read was still documenting the API and token names of two renames ago: **46 of the 92
1846
+ `--ds-*` tokens it named no longer exist** (`--ds-button-fg`, `--ds-sidenav-item-bg-hover`, the
1847
+ whole `--ds-pagehdr-*` family, …), `ds-button` was described with its pre-2026-06 single `variant`
1848
+ axis (`'primary' | 'secondary' | 'tertiary' | …`) instead of `color` × `variant`, and the topbar
1849
+ section documented a `--ds-topbar-margin-b` that **has never existed** — telling readers the
1850
+ topbar owns the gap below it and that pages must not add their own.
1851
+ Every name in the body is now the shipped one, and the two paragraphs are rewritten from the
1852
+ components' actual API.
1336
1853
 
1337
1854
  - **An outlined button no longer paints its own white background.** `ds-button variant="outline"`,
1338
1855
  `ds-icon-button variant="outlined"` and the outlined half of `ds-split-button` defaulted their fill to
@@ -1350,9 +1867,9 @@ None of this is required — skipping it leaves working code that is merely redu
1350
1867
 
1351
1868
  - **Everything focusable now gets the DS focus ring, not the browser's black outline.** The library
1352
1869
  styled focus on the elements that remembered to — 42 of 81 component stylesheets, and only 35 of those
1353
- with `:focus-visible` — and everything else
1354
- fell through to `outline: auto`, which Chrome paints black: the close button inside `ds-alert`,
1355
- `ds-banner` and `ds-drawer`, a link in projected content, and any plain `<button>` an app writes itself.
1870
+ with `:focus-visible` — and everything else fell through to `outline: auto`, which Chrome paints
1871
+ black: the close button inside `ds-alert`, `ds-banner` and `ds-drawer`, a link in projected
1872
+ content, and any plain `<button>` an app writes itself.
1356
1873
  One base rule now covers them: `:where(:focus-visible)` sets `outline: none` and resolves
1357
1874
  `--ds-control-focus-shadow`, written at **zero specificity** so every component rule and every app rule
1358
1875
  still out-ranks it. The `data-focus` switch is unaffected — all five modes (default, `ring`, `field`,
@@ -1360,9 +1877,9 @@ None of this is required — skipping it leaves working code that is merely redu
1360
1877
  specificity. **If you deliberately suppressed a focus ring** by leaving an element unstyled, it will now
1361
1878
  show one; give that element `box-shadow: none` explicitly. Note the rule also drops the outline such an
1362
1879
  element used to get from the browser, so if an `outline` was what you wanted there, declare it yourself.
1363
- **The case that actually bites, added 2026-08-21 after a consumer hit it fourteen times:** an app rule
1364
- that styles `:focus-visible` with an **`outline` only**. That rule keeps winning on `outline`, but it
1365
- says nothing about `box-shadow`, so the base rule still supplies one and the element renders **both** —
1880
+ **The case that actually bites:** an app rule that styles `:focus-visible` with an **`outline`
1881
+ only**. That rule keeps winning on `outline`, but it says nothing about `box-shadow`, so the base
1882
+ rule still supplies one and the element renders **both** —
1366
1883
  a second ring in the app's colour around the DS ring. Nothing fails a build and nothing greps for it:
1367
1884
  the app rule looks untouched because it is. Add `box-shadow: none` to that rule, or — better, and what
1368
1885
  this release is for — delete the rule and retint the DS ring through `--ds-control-focus-shadow`.
@@ -1370,13 +1887,11 @@ None of this is required — skipping it leaves working code that is merely redu
1370
1887
  - **A disabled `ds-button` / `ds-icon-button` now reads as disabled on a COLOURED accent.** One label recipe
1371
1888
  for every variant — `oklch(from accent 70% 0.035 h)` — replacing the 58% default, and on `ds-button` the
1372
1889
  52% overrides that tonal and outline carried (`ds-icon-button` only ever had the base). Those sat right
1373
- where a saturated accent already is, so the disabled label
1374
- differed from the live one by **1.16–1.60:1** on four of the five variants; only `color="neutral"` was
1890
+ where a saturated accent already is, so the disabled label differed from the live one by
1891
+ **1.16–1.60:1** on four of the five variants; only `color="neutral"` was
1375
1892
  safe, because its enabled label is nearly black. Now **2.40–2.71:1** on the coloured variants that
1376
1893
  were failing, and higher still on `neutral`, which reaches 6.46:1.
1377
- Measured, not guessed: the matrix is in `presentations/disabled-recipe.html` — 5 variants x 3 colours
1378
- x rest/hover/active/disabled in light, and a narrower primary-only sweep in dark. **The trade,
1379
- stated:** the disabled label is fainter
1894
+ **The trade, stated:** the disabled label is fainter
1380
1895
  against its own background than it was — 2.1–2.7:1 where the old recipe gave 3.4–5.2 — which is the point,
1381
1896
  and disabled text is exempt from the WCAG contrast floor. **Override per component if you disagree:**
1382
1897
  `--ds-button-label-color-disabled` / `--ds-icon-button-icon-color-disabled` still win. **And delete this
@@ -1386,14 +1901,13 @@ None of this is required — skipping it leaves working code that is merely redu
1386
1901
  - **A disabled control was dimmed twice, and had nearly disappeared.** Five places each set
1387
1902
  `--text-disabled` AND `opacity: 0.6` on top of it: the closed triggers of `ds-select`,
1388
1903
  `ds-date-picker` and `ds-time-picker`, plus two genuine rows — `ds-menu-item` and the shared dropdown
1389
- option. Measured in the light theme: that lands at **1.68:1** against the
1390
- surface, where the colour alone gives **2.53:1** and an enabled control sits at 17.32:1 — so the
1391
- state was
1392
- never in danger of being missed, the text was. The opacity is gone from those five; the colour carries
1393
- the state, which is what it was chosen for. Where opacity IS the only signal — a disabled calendar day,
1394
- `ds-toggle-button` — it now reads `--opacity-disabled` (0.5) instead of `--opacity-muted` (0.6), the
1395
- token named for the job. **If you retuned `--opacity-muted` to compensate**, drop it: it is a
1396
- de-emphasis level for hints and secondary text, and it no longer carries any disabled state.
1904
+ option. Measured in the light theme: that lands at **1.68:1** against the surface, where the colour
1905
+ alone gives **2.53:1** and an enabled control sits at 17.32:1 — so the state was never in danger of
1906
+ being missed, the text was. The opacity is gone from those five; the colour carries the state, which is what it was chosen for.
1907
+ Where opacity IS the only signal — a disabled calendar day, `ds-toggle-button` — it now reads
1908
+ `--opacity-disabled` (0.5) instead of `--opacity-muted` (0.6), the token named for the job. **If
1909
+ you retuned `--opacity-muted` to compensate**, drop it: it is a de-emphasis level for hints and
1910
+ secondary text, and it no longer carries any disabled state.
1397
1911
 
1398
1912
  - **`ds-table` lost the sort indicator when a host answered a cleared sort with its own default.** Third
1399
1913
  click clears a column, the host replies with its default sort, the rows come back ordered by it — and
@@ -1403,33 +1917,30 @@ None of this is required — skipping it leaves working code that is merely redu
1403
1917
  to `owner`/`desc` had its `desc` dropped and stayed at `null`, while a host defaulting to `asc` was
1404
1918
  restored. That is exactly why the same three clicks worked on one screen and not another. The marked
1405
1919
  state is now a `linkedSignal` over BOTH inputs, so a change in either re-seeds the pair and the
1406
- direction arrives with the key. **Verified against both reported screens** by replaying them through
1407
- Angular's real signal graph with the binding diff modelled: the old shape fails the first and passes
1408
- the second, the new one passes both. A table that binds neither input is unaffected.
1920
+ direction arrives with the key. A table that binds neither input is unaffected.
1409
1921
 
1410
1922
  - **`ds-table` painted its empty state while the first read was still in flight.** Only
1411
1923
  `loading="skeleton"` suppressed the empty row, and `overlay` is the mode a list uses on every read
1412
1924
  after the first — so a zero-row reload showed "No records match your filters" or a whole
1413
1925
  `ds-empty-state` CTA underneath the spinner. The empty row is a claim about the data ("there are
1414
1926
  none"), so it is now withheld whenever `loading` is anything but `false`. **You can delete the guard
1415
- you wrote for this** — and check the ones you did not write: of one consumer's 14 overlay-loading
1416
- tables, 9 were shipping the sentence during a load, because branching failure-vs-empty looks complete
1417
- while saying nothing about loading. Note this closes the door on the `[empty]` slot during a load
1927
+ you wrote for this** — and check the ones you did not write, since branching failure-vs-empty looks
1928
+ complete while saying nothing about loading. Note this closes the door on the `[empty]` slot during a load
1418
1929
  rather than opening another one: there is no `loading` value that renders it, so copy you want a
1419
1930
  reader to see WHILE a load runs has to come from your own page, above or around the table.
1420
1931
 
1421
1932
  - **Releasing the mouse on the backdrop closed `ds-modal` and `ds-bottom-sheet`, even when the press
1422
1933
  started inside the dialog.** Select a sentence in the body and let go past the edge, or overshoot with
1423
- a slider, and the dialog vanished — reported by a consumer as "not a click, the mouse-UP". The cause is
1424
- in the DOM shape: those backdrops WRAP their dialog, and a browser fires `click` at the nearest common
1425
- ancestor of the press and the release, which for that gesture IS the backdrop — so the
1426
- `target === currentTarget` guard let it through. Dismissal now tracks the press and fires only when
1427
- both the press and the release land on the backdrop itself, which is how the CDK has always guarded
1428
- its own outside-click. `ds-drawer` and the `ModalService` confirm were never affected (their scrim is
1429
- a SIBLING of the panel, so the common ancestor is a third element) and are unchanged — all four were
1430
- re-checked. **If a test of yours dismisses a dialog with `backdropEl.click()`**, it now needs
1431
- `pointerdown` + `pointerup` on the backdrop, both with the primary button; a real mouse or touch has
1432
- always sent both. One side effect on `ds-bottom-sheet`: the sheet no longer stops click propagation
1934
+ a slider, and the dialog vanished. The cause is in the DOM shape: those backdrops WRAP their
1935
+ dialog, and a browser fires `click` at the nearest common ancestor of the press and the release,
1936
+ which for that gesture IS the backdrop — so the `target === currentTarget` guard let it through.
1937
+ Dismissal now tracks the press and fires only when both the press and the release land on the
1938
+ backdrop itself, which is how the CDK has always guarded its own outside-click. `ds-drawer` and
1939
+ the `ModalService` confirm were never affected (their scrim is a SIBLING of the panel, so the
1940
+ common ancestor is a third element) and are unchanged — all four were re-checked. **If a test of
1941
+ yours dismisses a dialog with `backdropEl.click()`**, it now needs `pointerdown` + `pointerup` on
1942
+ the backdrop, both with the primary button; a real mouse or touch has always sent both. One side
1943
+ effect on `ds-bottom-sheet`: the sheet no longer stops click propagation
1433
1944
  (the guard it needed for the old `click` listener is gone with it), so a click inside the sheet now
1434
1945
  reaches an ancestor handler of yours, the sheet being rendered inline rather than portaled.
1435
1946
 
@@ -1469,157 +1980,6 @@ None of this is required — skipping it leaves working code that is merely redu
1469
1980
  it. It now reads `env(safe-area-inset-bottom, 0px)`, which is what `ds-auth-layout` — the only other
1470
1981
  component using these insets — already did on all four of its.
1471
1982
 
1472
- ### Changed
1473
-
1474
- - **`avatarTint` switches from OKLCH to `hsl(H 50% 62%)` with white initials, always.** The seeded
1475
- tints read as dull, and every set of initials came out black. The auto-contrast was not the fault:
1476
- measured across the hue circle, the old `oklch(62% 0.14 h)` put every hue's luminance between 0.218
1477
- and 0.271 against a 0.179 black/white crossover, so black was genuinely the better pick every time
1478
- (5.4–6.5:1) — the switch was right, the lightness never let it switch. The new recipe is pinkmoon's
1479
- own, adopted deliberately: HSL lightness is not perceptually even, so yellows land far lighter than
1480
- blues, and that unevenness is exactly what makes the row read as bright.
1481
- **It is a knowing trade against contrast.** White on the lighter hues measures as little as
1482
- **1.65:1**, under the 4.5:1 WCAG AA asks of text — verified across the full 20–300° range, where all
1483
- 29 sampled hues fall short. Suitable where the avatar sits beside a name that is written out anyway;
1484
- where the initials are the only identifier, pass your own `lightness` / `foreground`.
1485
- **API:** the `chroma` option is gone, replaced by `saturation` (HSL %, default 50). `lightness` keeps
1486
- its name and its default of 62 but now means **HSL** lightness, so an existing numeric override will
1487
- land somewhere different — re-pick it rather than assuming it carried over. `foreground` now defaults
1488
- to `#fff` instead of auto-choosing; pass it explicitly to get anything else. `avoidHues` is unchanged.
1489
-
1490
- - **`ds-input`'s `floating` label is removed, with no replacement.** It only ever did half the job: the
1491
- label was drawn inside a taller control, but it never moved out of the field on focus — the animation
1492
- from inside-the-box to above-it *is* the pattern, and without it an empty field reads as a
1493
- permanently mislabelled one rather than a floating-label one. Nobody had asked for it, and finishing
1494
- it is a real piece of motion work, so it goes. **Migration is deleting the attribute:**
1495
- `[floating]="true"` → nothing, and the label renders above the field like every other field in the
1496
- library. One thing to re-check when you do: a floating input stood at `--hit-touch` (48px at default
1497
- density) and now takes the ordinary size scale (`--field-h-md` → `--hit-cozy`, 40px at default
1498
- density), so a row whose height was built around one moves. Gone with it are the five internal
1499
- tokens only this variant read
1500
- (`--field-h-floating`, `--field-floating-label-fs`, `--field-floating-label-lh`,
1501
- `--field-floating-label-top`, `--field-floating-pad-t`) — overriding them is now inert.
1502
- `--ds-elevation-floating` is a different token for toasts and is untouched.
1503
-
1504
- - **`ds-table`'s `sortKey` and `sortDir` are an `input` + `output` pair instead of a `model`** — the
1505
- mechanism behind the indicator fix above, because a `model` cannot tell a host's write apart from the
1506
- table's own. **Templates are unaffected:** `[(sortKey)]` still works (`sortKeyChange` and
1507
- `sortDirChange` are emitted), one-way `[sortKey]` with `(sortChange)` still works, and a static
1508
- `sortKey="project"` still works. What breaks is reaching into the component in TypeScript:
1509
- `table.sortKey.set(…)` no longer exists — nor `sortDir.set(…)` — so drive it from your own state and
1510
- bind it, which is what every server-paged list already does. Reading `table.sortKey()` still works and
1511
- now returns what the host bound; the marked state the header renders is internal. **One emission rule
1512
- moved with it:** both `…Change` outputs now fire on EVERY sort click, where a `model` suppressed the
1513
- emit when the value was unchanged. `[(sortKey)]` is unaffected — it rebinds the same value — but a
1514
- `(sortKeyChange)="reload()"` handler now also runs when only the DIRECTION changed. Listen to
1515
- `(sortChange)` instead, which carries both halves once.
1516
-
1517
- - **`ds-modal` / `ds-modal-card`: `error: boolean` → `status: DsStatus`.** A boolean could not express
1518
- a warning or an info dialog, and the rest of the library settled on the `DsStatus` vocabulary in
1519
- 0.9.0. **Migration is one attribute:** `[error]="true"` → `status="error"`, and a computed
1520
- `[error]="isDangerous"` → `[status]="isDangerous && 'error'"`, since the input takes `false` and
1521
- `undefined` as "no status" — the longer `? 'error' : null` ternary works too. Unset stays neutral,
1522
- so a dialog that never used `error` looks identical. **What you gain beyond the rename:** a status now
1523
- drives the icon-badge tint, the title colour, a DEFAULT icon and the ARIA role — the same four
1524
- things `ds-alert`'s variant drives, so a status dialog and a status alert on one screen cannot
1525
- disagree. The default icon matters: a tint alone signals by colour only, which fails the
1526
- don't-rely-on-colour rule. `icon` still overrides the pick, and `error`/`warning` default the
1527
- dialog to `alertdialog`. `dialogRole` lives on `ds-modal-card` only — it is now
1528
- `'dialog' | 'alertdialog' | null`, defaulting to `null` rather than `'dialog'` so "unset" is
1529
- distinguishable from a deliberate `dialog`, and an explicit value still wins. `ds-modal` has no such
1530
- input, so `<ds-modal status="error">` becomes an `alertdialog` with nothing to override it. What
1531
- renders is unchanged unless you set a status. Three tokens come with the tints:
1532
- `--ds-modal-title-text-color-warning`, `-success` and `-info`, joining the `-error` that already
1533
- existed. **One thing rides along that no attribute of yours controls:** a
1534
- `ModalService.confirm({ tone: 'error' })` passes `status="error"` down internally, and a status now
1535
- supplies a default icon — so a destructive confirm that never set `options.icon` opens with a red
1536
- error badge where it previously had none. Pass `icon` to choose a different glyph; there is no input
1537
- that turns the badge back off.
1538
-
1539
- - **Chips cap by width by default.** `maxVisibleChips` on `ds-autocomplete` and `ds-multi-select`
1540
- now defaults to `"fit"` instead of "show everything": the trigger keeps as many chips as it is
1541
- wide and collapses the rest into "+N more". A trigger that grew a line per selection was pushing
1542
- the rest of the form down the page. **To keep the old behaviour, bind `[maxVisibleChips]="null"`.**
1543
- Two more things moved with the default, and both are easy to miss. A trigger in any measuring mode is
1544
- now single-line (`flex-wrap: nowrap`), and its chips are held to one line each and clipped rather than
1545
- wrapped — so a chip carrying more than a short label gets squeezed, where `null` still wraps the row.
1546
- And **`[maxVisibleChips]="0"` has inverted**: it used to mean "show them all" and now means "show
1547
- none", putting the whole selection behind `+N more`. If you have a `0` anywhere, it was almost
1548
- certainly written for the old meaning; `null` is what that meant.
1549
-
1550
- - **`size` is now `DsSize | null` on the fourteen controls that resolve `DS_CONFIG.control.size`**, so
1551
- an unbound `size` can fall through to it. `ds-drawer` and `ds-spinner` also take a `size` and are NOT
1552
- among them — theirs stays a non-null `DsSize` defaulting to `'md'`, so a grep for `DsSize` finds
1553
- sixteen controls and only fourteen changed. **Nothing changes in a template** — `size="sm"` binds as
1554
- before and an unbound control still renders `md`. **What changes is reading it back:** `size()` now
1555
- returns `null` when nothing was bound, where it used to return `'md'`. The same is true of
1556
- `ds-card`'s `elevation` and `padding`, `ds-icon`'s `size`, `ds-badge`'s `withDot`, `ds-table`'s
1557
- `bordered` and `stickyScrollbar`, `dividers` on both modal components, `dialogRole` on
1558
- `ds-modal-card`, `brand` and `logoLetter` on `ds-sidenav` and `ds-auth-layout`, and `tagline` on the
1559
- auth layout — all of them now answer `null` for "nothing bound"
1560
- rather than the old default. **There is no public accessor for the resolved value**, and this is the
1561
- awkward part of the release: each component holds a `sizeResolved()` (or `elevationResolved()`,
1562
- `withDotResolved()`, …) but declares it `protected`, so reading it from a host, a test or a
1563
- `viewChild` does not compile. `ds-field-group` is the single exception, where `sizeResolved` is
1564
- public. Until that is opened up, a host that needs to know what a control is rendering at should keep
1565
- the value it bound, or resolve `DS_CONFIG.control.size` itself — the fallbacks are unchanged, but
1566
- they are not all `false`: `'md'` for `size`, `'flat'` for `elevation`, `16` for the icon size,
1567
- `false` for `bordered` / `stickyScrollbar` / `withDot`, **`true`** for `dividers`, `''` for the brand
1568
- strings, and the card's own padding when `padding` is unset. **If you implement
1569
- `DsFieldGroupContext`**, its member is now
1570
- `sizeResolved` rather than `size`, for the same reason: a segment must not have to resolve the
1571
- group's size a second time.
1572
-
1573
- - **`[data-density="ultra-compact"]` now compresses surface padding too** — `--ds-container-pad`,
1574
- the inner padding `ds-card` and `ds-stat` share, drops from 24px to 16px in that tier only.
1575
- Everything else is untouched: the default and the other three tiers still resolve to 24px. It was
1576
- the one padding the tier missed — it already compressed field padding and table cells, so a card
1577
- kept its full 24px frame while its own cells halved. **If you want the old 24px inside an
1578
- ultra-compact container**, re-declare `--ds-container-pad: var(--space-6)` on it. For a value
1579
- between the rungs — 20px, say — set that one token at `:root`; both components default to it, which
1580
- is why neither `ds-card` nor `ds-stat` should be pinned on its own.
1581
-
1582
- - **`ds-tabs` `urlParam` now SLUGS a tab's label instead of using it raw.** A tab with no `value` used
1583
- to put its display string straight into the address — `?tab=License%20Verification%20Status`. It is now
1584
- kebab-cased and lowercased with diacritics dropped: `?tab=license-verification-status`. **If you
1585
- adopted `urlParam` in 0.9.0 with tabs that declare no `value`, your existing links change** — an old
1586
- one no longer matches and lands on the first tab, which is what a stale bookmark has always done.
1587
- Declare `value` on each tab to pin the names for good: a slug is still derived from copy, so renaming a
1588
- tab still moves the link. What `tabClosed` emits is unchanged — still the raw label.
1589
-
1590
- - **`ds-auth-layout`'s projected `[brand]` slot no longer forces `width: auto`.** It shipped in 0.9.0
1591
- alongside the slot's `max-height`, but at `(0,2,0)` it out-ranks a width the consumer set on their
1592
- own logo, and with the default `max-height: none` it never did anything: an `<img>` with no CSS
1593
- width already keeps its ratio under `max-height`. If you opted into the max-height AND set an
1594
- explicit width on your logo, add `width: auto` yourself to keep the old scaling.
1595
-
1596
- - **The chips of both pickers now render inside a `<span class="chip-row">`** rather than directly
1597
- in the trigger — the element the width measurement needs. **On `[maxVisibleChips]="null"` it is a
1598
- wrapper and nothing more**: `display: contents`, so it adds no box and the chips stay flex items of
1599
- the trigger, identical in spacing, wrapping and height. In the measuring modes — which includes the
1600
- new default — the row becomes a `nowrap` flex box that clips, which is how the cap is enforced.
1601
- Either way, what breaks is a CSS override that walks the structure:
1602
- `.autocomplete-trigger > ds-tag`, `:nth-child` written against the trigger's children, or
1603
- `ds-tag + .autocomplete-input` adjacency — the chips are no longer siblings of the filter input,
1604
- the addon slots or the chevron. Drop the combinator or insert `.chip-row`. Multi-select's trigger is
1605
- `.field-trigger`, not `.trigger`, if you are rewriting a selector against it. Chips carry `data-chip`
1606
- and the marker `data-chip-overflow`; the measurement finds them that way, so leave them in place.
1607
-
1608
- - **If you EXTEND a DS component, this is the release that will break you.** `npm run class-surface`
1609
- against 0.9.1 reports **30 exported classes whose member surface grew** — among them `ButtonComponent`,
1610
- `TableComponent`, `ModalComponent`, `SearchComponent`, `AvatarComponent`, `IconComponent`, both
1611
- pickers, every bordered field that gained `controlState`, and every control that gained
1612
- `sizeResolved`. A subclass declaring any of those names now fails with TS4114. **Four classes also
1613
- LOST members**, which is the case a subclass notices differently — an `override` that now overrides
1614
- nothing: `ModalComponent.error` and `.onBackdrop`, `ModalCardComponent.error`,
1615
- `BottomSheetComponent.stop` and `InputComponent.floating`. Subclassing a DS component remains
1616
- unsupported, and this is the third release running where it broke someone, so it is stated in full
1617
- rather than by example. **Two kinds of break the report cannot see**, so check them by hand:
1618
- a member whose SIGNATURE changed keeps its name — `TableComponent.exporting` and `.exportStatus`
1619
- are still there, but as public readonly signals rather than the protected writables a subclass
1620
- could `.set()` — and a plain `private pending: … = null` is not a field the parser recognises, which
1621
- is how `SearchComponent`'s `pending` and `pushed` sit outside its reported seven.
1622
-
1623
1983
  ### What you can now DELETE
1624
1984
 
1625
1985
  None of this is required — skipping it leaves working code that is merely redundant. Each row is a workaround this release made unnecessary.
@@ -1635,13 +1995,13 @@ None of this is required — skipping it leaves working code that is merely redu
1635
1995
  | A `[dsStatValue]` projection used only to set `white-space: normal` or a smaller font size on a TEXT value. | the two hooks on the ordinary `value` input - keep the slot for a value that is genuinely another component, like a badge |
1636
1996
  | A wrapper around `ds-search` adding a typing debounce (typically 200ms) with an immediate path when the field empties. | `[debounce]`, which does both plus the Enter flush |
1637
1997
  | `--ds-button-height` / `--ds-button-font-size` / `--ds-button-pad-x` pinned under `ds-filter-bar` to force its trigger smaller. | `triggerSize="sm"` |
1638
- | `[dividers]="false"` repeated on every modal (one consumer had 41), or a reach into `.modal > .header` / `.body` / `.actions` with `!important` because hiding the rules left the padding stacked. | `provideDsConfig({ modal: { dividers: false } })`, which drops the rules AND the padding that existed to clear them |
1998
+ | `[dividers]="false"` repeated on every modal, or a reach into `.modal > .header` / `.body` / `.actions` with `!important` because hiding the rules left the padding stacked. | `provideDsConfig({ modal: { dividers: false } })`, which drops the rules AND the padding that existed to clear them |
1639
1999
  | A helper keeping the last visited tab per path in `sessionStorage`. | `rememberLastTab` on `ds-tabs` (needs `urlParam`, which supplies the names) |
1640
2000
  | An adapter component that injects a brand object and fans it out into `brand` / `logoLetter` / `tagline` bindings on each auth screen. | `provideDsConfig({ brand: … })`; the accent stays a `[dsBrandName]` projection instead of a `brandAccent` string plus an `indexOf` split |
1641
2001
  | The `ds-avatar` wiring around your own initials helper (the helper itself stays yours). | `[name]` plus one `[initialsRule]` |
1642
- | An attribute repeated identically on every element - one consumer binds `[stickyScrollbar]="true"` on all 35 of its tables and `size="sm"` on 146 buttons. | `provideDsConfig` with the matching section (`table`, `control.size`, …) |
2002
+ | An attribute repeated identically on every element — `[stickyScrollbar]="true"` on every table, `size="sm"` on every button. | `provideDsConfig` with the matching section (`table`, `control.size`, …) |
1643
2003
  | A button wired to `exportCsv()` on a SERVER-PAGED table - it serialises the rows the table holds, so the file is one page. | `runExport()` with an `exportFetcher` |
1644
- | The computed beside every list table - `items().length > 0 ? 'overlay' : 'skeleton'` - one consumer feature carried three copies of it. | `[loading]="true"`, which resolves the treatment itself; pin a value only where you mean to override the rule |
2004
+ | The computed beside every list table - `items().length > 0 ? 'overlay' : 'skeleton'`. | `[loading]="true"`, which resolves the treatment itself; pin a value only where you mean to override the rule |
1645
2005
  | An override retuning `--opacity-muted` to compensate for a washed-out disabled state. That token is a de-emphasis level for hints and secondary text and carries no disabled state at all - and on `ds-button` / `ds-icon-button` it was always inert, because neither ever read it. | `--ds-button-label-color-disabled` / `--ds-icon-button-icon-color-disabled` |
1646
2006
  | A `::ng-deep ds-table { .table th, .table td { width: 1%; white-space: nowrap } }` override - the shrink-to-content idiom. | `TableColumn.width: 'content'` |
1647
2007
  | A `stopPropagation()` on a control inside a clickable row, added so the row would not steal the control's KEY. Read each hit before deleting: a CLICK still bubbles (`onRowClick` only skips a target inside an `<input>`), so a click-path guard stays. | nothing on the key path - `onRowKey` now requires `target === currentTarget` |
@@ -1660,84 +2020,41 @@ None of this is required — skipping it leaves working code that is merely redu
1660
2020
  shipping with a name, a type and no description: 10 `--ds-*` tokens (the whole
1661
2021
  `--ds-password-rules-*` set among them, plus `--ds-multi-select-panel-bg-color` and
1662
2022
  `--ds-menu-item-radius`) and 31 inputs and outputs across `ds-calendar-day`, `ds-time-columns`,
1663
- `ds-field-message`, `ds-divider`, `ds-sidenav-footer` and the three layout primitives. **Three of
1664
- them were guessable wrong**, so check yours if you use them: `ds-time-columns` `value` is always
1665
- `HH:mm` in 24h and `format` changes only the DISPLAY, so a 12h picker still emits 24h;
1666
- `minTime`/`maxTime` disable the out-of-range rows rather than removing them, so the columns keep
1667
- their height as a bound moves; and `ds-calendar-day` `tabIndex` is a roving tabindex (one cell at
1668
- 0, the rest at -1), not a tab position you set per day. No behaviour changed — only what the
1669
- reference tells you.
1670
-
1671
- ### Fixed
1672
-
1673
- - **The changelog page on the docs site was rendering with browser defaults.** Its markdown styles
1674
- were scoped by emulated encapsulation, which stamps its attribute on the container but not on the
1675
- nodes the markdown renderer produces — so every rule in that block matched nothing: measured
1676
- border-top 0px and margin-top 0px on the version headings, and `h3` white instead of
1677
- `--text-secondary`. Docs-site only, nothing in the package.
1678
-
1679
- ### Changed
1680
-
1681
- - **`migrations/<version>.json` now carries the delete-list, so an upgrade tool can act on it.**
1682
- Each version's manifest gains `DELETE-LIST: if you have <X> -> <why it can go>` entries — 21 of
1683
- them for 0.9.0 — naming the workarounds that release made unnecessary, from the
1684
- `.button.button.button` triple-class to hand-binding `[value]`/`(valueChange)` on every field of a
1685
- signal form. They existed for 0.9.0 but only in a prose file the tooling does not read, so nothing
1686
- applied them. **Nothing here is required**: skipping it leaves working code that is merely
1687
- redundant. The same table is in this file under 0.9.0 if you would rather read than run.
1688
- `migrations/unreleased/README.md` no longer ships — it duplicated the manifest and had already gone
1689
- stale inside one release.
1690
-
1691
- ## 0.9.0 — 2026-08-10
1692
-
1693
- ### What you can now DELETE
1694
-
1695
- The half a release note usually leaves out. Two dogfooding audits (mm-ecm 2026-08-06, mm-broker
1696
- 2026-08-07) read consumer stylesheets as a record of what this library made them write, and both
1697
- found the same thing: **most of the workarounds were written against hooks that already existed** —
1698
- the consumer had no signal that anything had been tokenised since. That is a failure of these notes,
1699
- not of the consumer. None of this is required; skipping it leaves working code that is merely
1700
- redundant.
1701
-
1702
- | If you have | Delete it, because |
1703
- | --- | --- |
1704
- | `.button.button.button { … }` or rules against `.v-outline` / `.c-secondary` | `ds-button` reflects `[data-variant]` and `[data-color]` on the HOST. Use `ds-button[data-variant="outline"]` |
1705
- | `--radius-sm: 0` to square a checkbox | `--ds-checkbox-radius` |
1706
- | `.title.title { font-size }` inside a modal | `--ds-modal-title-font-size` |
1707
- | Pagination label type overrides — a rule on `ds-pagination` setting `--ds-pagination-label-*` | `--ds-pagination-label-*` |
1708
- | `.size { display: none }` on a pager | `[showPageSize]="false"` — and it is now forwarded through `ds-table` too |
1709
- | `.actions:empty { display: none }` on a modal | The library already ships it |
1710
- | `--ds-modal-divider: transparent`, plus whatever you rewrote the modal padding to afterwards | `[dividers]="false"` — it drops the rules *and* the padding that existed to clear them |
1711
- | A defensive `:hover` rule guarding the selected dropdown row — a selector naming `is-selected` | The `:not(.is-selected)` branch already handles it |
1712
- | `ds-menu-divider { padding: 0 }` | `--ds-menu-divider-spacing`, which already defaults to 0 |
1713
- | A dozen scattered `--ds-*-radius: 0` | For a fully square brand, four lines: `--radius-sm/md/lg/xl: 0` (leave `--radius-pill`). Measured — the three ROLES alone leave options, clear buttons and dropdown panels rounded; they are the tool for squaring one family and keeping another. Worked example on the Theming page |
1714
- | A global rule squaring `.ds-overlay-pane` | Radius and elevation roles now travel onto portaled panes when the host overrides them |
1715
- | Absolutely positioning a search glyph over an autocomplete, or hiding `.autocomplete-chevron` | `dsAutocompletePrefix` and `[showChevron]="false"` |
1716
- | Querying the autocomplete's inner `<input>` and dispatching a synthetic `input` event to clear it | `clear()` |
1717
- | A hand-built KPI tile — a `kpi` block pairing your own markup with `ds-badge`, where `ds-stat` now fits | `ds-stat` takes a projected `[dsStatValue]` |
1718
- | A pinned `--ds-search-height` (or any sibling) to get a shorter control row | `density="compact"` on the container — every field control is 32px there, measured |
1719
- | A hand-rolled `.scrim` + `.panel` dialog, written because the modal header could not hold a badge or a second line | `ds-modal` projects `[title-badge]` and takes a `subtitle`. You get back the focus trap, focus restore, the scroll lock, Escape and the footer |
1720
- | `[value]="f.x().value()" (valueChange)="f.x().value.set($event)"` on every field of a signal form | `[formField]="f.x"` — the control implements the contract already |
1721
- | `white-space: normal` forced onto a badge in a table cell | `--ds-badge-white-space` |
1722
- | `::ng-deep .card { height: 100% }` for equal-height cards | The card's box fills a stretched host on its own |
1723
- | `!important` on a `--ds-table-cell-pad` override under `density="compact"` | Compact no longer out-specifies the padding hooks |
1724
- | A wrapper around `ds-table` for sort/range announcements, a `role="separator"` resize handle, or a header filter the checklist cannot express | All three ship: the live region, the `role="separator"` handle, and `dsHeaderFilter` |
1725
- | `new Event('input'`, `dispatchEvent`, `querySelector` | Querying the autocomplete's inner `<input>` and dispatching a synthetic `input` event to clear it Replace with: `clear()`. |
1726
- | `--ds-autocomplete-input-minw` | A --ds-autocomplete-input-minw override, written to hide the empty filter box that sat beside a chosen value in single mode. The input collapses on its own once a value is picked and the panel closes. Replace with: nothing - unless you genuinely want a wider typing area. |
1727
- | `.trigger`, `:focus-within`, `ds-date-picker` | A :focus-within rule on the date-picker trigger, written because the trigger went dark while its panel was open. The component keeps an "open" class alongside the focus rules now. Replace with: nothing - the component holds the indicator itself. |
1728
- | A wrapper around the export button catching a rejected `exportFetcher` — a `try {` / `catch` around it | ds-table emits `(exportFailed)` now. Replace with: bind (exportFailed) on ds-table. |
1729
- | Your own live region around `ds-table` — an `aria-live` or `role="status"` element announcing sort state or row range | The table announces both through its own role="status" region, so keeping yours announces every change twice. Replace with: nothing - retune the wording through sortAnnouncement and rangeAnnouncement. |
2023
+ `ds-field-message`, `ds-divider`, `ds-sidenav-footer` and the three layout primitives. **Three of
2024
+ them were guessable wrong**, so check yours if you use them: `ds-time-columns` `value` is always
2025
+ `HH:mm` in 24h and `format` changes only the DISPLAY, so a 12h picker still emits 24h;
2026
+ `minTime`/`maxTime` disable the out-of-range rows rather than removing them, so the columns keep
2027
+ their height as a bound moves; and `ds-calendar-day` `tabIndex` is a roving tabindex (one cell at
2028
+ 0, the rest at -1), not a tab position you set per day. No behaviour changed — only what the
2029
+ reference tells you.
1730
2030
 
1731
- **To sweep for the rest:** grep your styles for `::ng-deep`, `!important` and any selector naming a
1732
- DS internal class (`.button`, `.trigger`, `.option`, `.modal`). Each hit is either a row above or a
1733
- genuine gap worth reporting.
2031
+ ### Changed
2032
+
2033
+ - **`migrations/<version>.json` now carries the delete-list, so an upgrade tool can act on it.**
2034
+ Each version's manifest gains a `deleteList` array — `{ id, surface, find, why, replaceWith }`
2035
+ per row — naming the workarounds that release made unnecessary, from the
2036
+ `.button.button.button` triple-class to hand-binding `[value]`/`(valueChange)` on every field of a
2037
+ signal form. They existed for 0.9.0 in prose only, which no tool reads — so an upgrade that
2038
+ crossed 0.9.0 before this release applied none of them. **Nothing here is required**: skipping it leaves working code that is merely
2039
+ redundant. The same table is in this file under 0.9.0 if you would rather read than run.
2040
+ `migrations/unreleased/README.md` no longer ships from the package — its content now lives in the
2041
+ manifest.
2042
+
2043
+ ### Fixed
2044
+
2045
+ - **The changelog page on the docs site was rendering with browser defaults.** Its markdown styles
2046
+ were scoped by emulated encapsulation, which stamps its attribute on the container but not on the
2047
+ nodes the markdown renderer produces, so every rule in that block matched nothing: border-top 0px
2048
+ and margin-top 0px on the version headings, and `h3` white instead of `--text-secondary`.
2049
+ Nothing in the package.
2050
+
2051
+ ## 0.9.0 — 2026-08-10
1734
2052
 
1735
2053
  ### Changed (breaking, behaviour)
1736
2054
 
1737
2055
  - **If you EXTEND a DS component, this release collides with your subclass.** Sixteen exported
1738
2056
  classes gained members, so a subclass declaring the same name stops the build: `TS4114` wants
1739
- `override`, `TS2415` fires on a redeclared private, `TS2416` on a differing signature. Not
1740
- hypothetical — a consumer migrating to 0.8.0 hit exactly this on three `TableComponent` members.
2057
+ `override`, `TS2415` fires on a redeclared private, `TS2416` on a differing signature.
1741
2058
  **`TableComponent` is by far the largest**: `lazy`, `lazyLoad`, `showLabel`, `showPageSize`,
1742
2059
  `filterMode`, `columnFilterOptions`, `sortAnnouncement`, `rangeAnnouncement` and `exportFailed`,
1743
2060
  plus a dozen protected members. **And one signature changed rather than appeared:** `isStriped()`
@@ -1746,8 +2063,10 @@ genuine gap worth reporting.
1746
2063
  bases gained `hidden` and `focus()` (inherited by every field control); `ds-input` / `ds-textarea`
1747
2064
  gained `readonlyBinding`, `isReadOnly`, `minLength`, `pattern`; `ds-modal` and `ds-modal-card`
1748
2065
  gained `subtitle` and `dividers`; `ds-autocomplete` gained `maxVisibleChips`, `showChevron` and
1749
- `clear()`. One member was also REMOVED: `TableComponent`'s private `distinctValuesOf()`, now `optionsFor()`, which prefers the new `columnFilterOptions` input. The full per-class list is in `migrations/0.9.0.json`. **Using the `<ds-…>` elements in
1750
- a template needs no change** — this only reaches you through `extends`.
2066
+ `clear()`. One member was also REMOVED: `TableComponent`'s private `distinctValuesOf()`, now
2067
+ `optionsFor()`, which prefers the new `columnFilterOptions` input. The full per-class list is in
2068
+ `migrations/0.9.0.json`. **Using the `<ds-…>` elements in a template needs no change** — this only
2069
+ reaches you through `extends`.
1751
2070
 
1752
2071
  - **`ds-table` `sortMode` and `filterMode` default to `'server'`, were `'client'`.** Server-paged is the
1753
2072
  normal case, and there the old default was actively wrong: the table sorted the current page, so the
@@ -1781,59 +2100,6 @@ genuine gap worth reporting.
1781
2100
  - **The global `--search-h` token is gone.** It existed only to hold that 36px and nothing reads it.
1782
2101
  Use `--ds-search-height` to retune search alone, or `--ds-field-height` for the family.
1783
2102
 
1784
- ### Changed
1785
-
1786
- - **`--ds-radius-inner` now reaches the small affordances it was written for.** Thirty-three reads across
1787
- fifteen component stylesheets took `var(--radius-sm)` directly — dropdown options, field clear buttons,
1788
- calendar days, chips, tag and toast close buttons — so the role a consumer is told to retune reached
1789
- exactly one component. Nothing moves (the role's default IS `--radius-sm`, verified by measuring a page
1790
- before and after), but it changes what a square brand costs: zeroing the three radius roles used to leave
1791
- five surfaces rounded and now squares every one. Five focus-ring radii inside `:focus-visible` keep
1792
- following the raw scale deliberately — a ring traces a box, it is not a surface a brand designs.
1793
-
1794
- - **`ds-auth-layout` compresses its own chrome on a short viewport, so the layout is never the reason a
1795
- card scrolls.** Block padding, the column gap, the letter-mark tile and the masthead's inner gap clamp
1796
- against `svh`. At ~330px of viewport — a landscape phone — that furniture costs ~131px instead of ~219px.
1797
- Every multiplier reaches its maximum at a viewport **800px tall or more** — a viewport height, not a
1798
- screen size: a 1366x768 laptop sits near 630px once browser and OS chrome come off, and a 1080p screen at
1799
- 150% display scaling near 585px, so a good many desktops render the compressed furniture too.
1800
- **Type is deliberately excluded.** An earlier draft clamped the wordmark and the tile's glyph as well;
1801
- both were measured to buy zero vertical space — the wordmark shares a flex row with the tile, whose height
1802
- exceeds the wordmark's line box at every viewport height, and the glyph sits in a fixed-size grid box —
1803
- while shrinking the brand against a fixed 14px tagline, worst for the users who scale their display up.
1804
- Both are back to the fixed sizes 0.8.0 ships.
1805
- `svh` rather than `dvh` because the dynamic unit changes as a mobile URL bar hides, which would re-space
1806
- the column mid-scroll; `min-height` stays on `dvh` so no strip of body background shows when the bar
1807
- retracts. New opt-outs: `--ds-auth-layout-pad-block` and `--ds-auth-layout-mark-size` (set
1808
- either, or the existing `-gap`, to a fixed value and that dimension stops compressing). A card taller than
1809
- the space left still scrolls — deliberately, because the alternative (pinning the masthead and footer out
1810
- of the flow) frees more room but lets chrome overlap a tall form.
1811
-
1812
- - **`ds-auth-layout` — a round of fixes from an audit of the above.** All additive or corrective; nothing
1813
- to migrate. The letter-mark is now `aria-hidden`, so a screen reader announces "MyAdmin" rather than
1814
- "A MyAdmin"; it also carries `flex-shrink: 0`, so a long wordmark can no longer squash the square tile
1815
- into a rectangle. `logoLetter` on its own renders its tile instead of nothing (the outer gate never
1816
- tested it). The glyph is taken with a code-point split rather than `charAt(0)`, which returned half a
1817
- surrogate pair for a brand starting with an emoji. `box-sizing: border-box` is declared on the host
1818
- rather than inherited, because the documented à-la-carte style path omits `styles/reset` and without it
1819
- `min-height: 100dvh` plus the block padding overflows by exactly that padding — the scrollbar this
1820
- component exists to avoid. The four `env(safe-area-inset-*)` calls gained `0px` fallbacks and moved to
1821
- physical `padding-left`/`padding-right`, since the safe area belongs to the device and must not swap
1822
- sides with text direction; note the insets are inert unless the **host app** sets `viewport-fit=cover`
1823
- in its viewport meta, which the library cannot do and did not previously say. New tokens, both
1824
- defaulting to today's behaviour: `--ds-auth-layout-mark-radius` (the tile's radius was the one value
1825
- with no override hop) and `--ds-auth-layout-brand-max-height` (`none` by default — set it to cap a
1826
- projected `[brand]` logo, which nothing constrained). New input `landmark` (`'none'` by default): the
1827
- card holds the whole content of an auth screen but `ds-card` renders an `<article>`, which is not a
1828
- landmark, so `landmark="main"` gives a screen-reader user something to jump to on a standalone auth
1829
- page. It is opt-in because the layout is also embedded inside pages that already own a `main`.
1830
-
1831
- - **BREAKING: `ds-auth-shell` is renamed `ds-auth-layout`** (`AuthShellComponent` → `AuthLayoutComponent`,
1832
- and `--ds-auth-shell-*` → `--ds-auth-layout-*`). "Shell" is not this library's word for a layout — the
1833
- app-level one was renamed `ds-app-shell` → `ds-app-layout` back in 0.7.0, and shipping an auth counterpart
1834
- called a shell in 0.8.0 reintroduced the split vocabulary that rename existed to remove. Mechanical:
1835
- rename the tag, the class and the six tokens. Nothing about the behaviour or the slots changes.
1836
-
1837
2103
  ### Added
1838
2104
 
1839
2105
  - **Exit motion is swappable, the way entrances already were** — `--ds-modal-leave-keyframes`,
@@ -1849,8 +2115,8 @@ genuine gap worth reporting.
1849
2115
  - **Signal Forms bind to every DS control, with no wrapper and no adapter** — `<ds-input [formField]="f.email" />`
1850
2116
  and the form owns the value, the validation, `touched`, `disabled` and the error text. This was very nearly
1851
2117
  true already and nobody knew: `FormValueControl` requires exactly one member, a `value` `model()`, and all
1852
- fifteen controls had it. Measured across all of them — thirteen bound with no library change at all. What
1853
- was fixed here: `min`/`max`/`maxLength` on `ds-input`, `ds-textarea` and `ds-number-stepper` now accept
2118
+ fifteen controls had it — thirteen of them bound with no library change at all. What was fixed here:
2119
+ `min`/`max`/`maxLength` on `ds-input`, `ds-textarea` and `ds-number-stepper` now accept
1854
2120
  `undefined` on the write side (the binder sends it for an unset bound; read types are untouched);
1855
2121
  `readonly` joins `ds-input`, `ds-textarea` and `ds-select` as the contract spelling of the existing
1856
2122
  `readOnly`, because signal forms binds by NAME and a `readonly()` rule was reaching nothing; `minLength`
@@ -1859,22 +2125,21 @@ genuine gap worth reporting.
1859
2125
  `focus()` resolves the element that holds the value, so `focusBoundControl()` reaches the stepper's number
1860
2126
  box rather than its minus button and the tabbable radio rather than the first in the DOM. Not implemented:
1861
2127
  `dirty`, `pending`, `disabledReasons`. Reactive and template-driven forms are untouched. Docs: **Practice →
1862
- Signal Forms**, with a live form and a contract table stating the measured state.
2128
+ Signal Forms**, with a live form and a contract table stating what each control implements.
1863
2129
 
1864
- - **`ds-modal` takes a `[title-badge]` slot and a `subtitle`.** Two consumer screens had abandoned
1865
- `ds-modal` and hand-rolled a `.scrim` + `.panel` because the header could only hold one string — losing
1866
- the focus trap, focus restore, the scroll lock, Escape, the enter/leave animation and, in one case, the
1867
- footer, which put a destructive action beside the close button. Same shape as `ds-page-header`; `title`
1868
- stays the accessible name, so projected markup never enters `aria-labelledby`. Two tokens:
1869
- `--ds-modal-subtitle-text-color` and `-font-size`.
2130
+ - **`ds-modal` takes a `[title-badge]` slot and a `subtitle`.** For a header that needs more than one
2131
+ string, without giving up what a hand-rolled `.scrim` + `.panel` dialog would otherwise lose: the focus
2132
+ trap, focus restore, the scroll lock, Escape and the enter/leave animation. Same shape as
2133
+ `ds-page-header`; `title` stays the accessible name, so projected markup never enters `aria-labelledby`.
2134
+ Two tokens: `--ds-modal-subtitle-text-color` and `-font-size`.
1870
2135
 
1871
2136
  - **`[dividers]` on `ds-modal` / `ds-modal-card`, default `true`.** `[dividers]="false"` drops the header
1872
2137
  and actions rules AND the vertical padding that existed to clear them, because those are one design
1873
- decision. Two consumers had set `--ds-modal-divider: transparent` and were left with a 35px gap where 16
1874
- was wanted; one needed `!important` to win it back.
2138
+ decision — setting `--ds-modal-divider: transparent` alone leaves the padding sized for a divider that
2139
+ is no longer there.
1875
2140
 
1876
- - **`ds-table`: the rest of what mm-broker's fork did that the base could not.** Keyboard column resizing —
1877
- the handle is a focusable `role="separator"` with Arrow (16px), PageUp/PageDown (64px) and Home to hand
2141
+ - **`ds-table`: keyboard column resizing, export failure handling, and two box tokens.** The resize
2142
+ handle is a focusable `role="separator"` with Arrow (16px), PageUp/PageDown (64px) and Home to hand
1878
2143
  the column back to automatic sizing; a drag handle that answered only to a pointer made a resizable column
1879
2144
  unusable without a mouse. `exportFetcher` may now resolve a `Blob` or a URL as well as rows, for a file
1880
2145
  the backend already formatted, and `exportFailed` emits the rejection — swallowing it made a failed export
@@ -1882,11 +2147,11 @@ genuine gap worth reporting.
1882
2147
  and `--ds-table-hscroll-bottom` (offsets the sticky scrollbar out of a padded container, default 0).
1883
2148
 
1884
2149
  - **The three `ds-modal` regions take their padding from tokens** — `--ds-modal-pad-x` for the gutter all
1885
- three share, plus `--ds-modal-header-pad-y`, `-body-pad-y` and `-actions-pad-y`. Two consumers were
1886
- rewriting these, one with `!important`, because the regions are padded for a divider-separated layout:
1887
- set `--ds-modal-divider: transparent` and you are left with spacing sized for dividers that are no
1888
- longer there. Nothing moves by default. Note the tokens go on `ds-modal` AND `ds-modal-card` — a dialog
1889
- opened through `ModalService` renders the card with no `ds-modal` ancestor.
2150
+ three share, plus `--ds-modal-header-pad-y`, `-body-pad-y` and `-actions-pad-y`. The regions are padded
2151
+ for a divider-separated layout: set `--ds-modal-divider: transparent` and you are left with spacing
2152
+ sized for dividers that are no longer there — these tokens let you retune it directly instead. Nothing
2153
+ moves by default. Note the tokens go on `ds-modal` AND `ds-modal-card` — a dialog opened through
2154
+ `ModalService` renders the card with no `ds-modal` ancestor.
1890
2155
 
1891
2156
  - **`maxItems` on `ds-breadcrumbs`** — collapses the middle of a deep trail into a crumb that says how
1892
2157
  many it stands for, keeping the first and, crucially, the LAST. Without a cap the row is `nowrap` with
@@ -1897,14 +2162,13 @@ genuine gap worth reporting.
1897
2162
  it — `isDevMode`, an `effect`, a condition, a `console.warn` — and each copy shared the same flaw: the
1898
2163
  effect re-runs when ANY signal it read changes, so a warning about a static misconfiguration could be
1899
2164
  logged again on every keystroke. They now go through one internal helper that fires at most once per
1900
- message. Eleven hand-written warnings across `ds-button`, `ds-stat` and `ds-table` moved onto it —
1901
- ten call sites at HEAD, since the table gained more with lazy loading. `ds-tabs` keeps its own, because
1902
- its effect also owns registry bookkeeping that a plain warning cannot express.
2165
+ message. Warnings on `ds-button`, `ds-stat` and `ds-table` moved onto it. `ds-tabs` keeps its own,
2166
+ because its effect also owns registry bookkeeping that a plain warning cannot express.
1903
2167
 
1904
2168
  - **`maxVisibleChips` on `ds-autocomplete`** — caps the chips in the trigger and collapses the rest into
1905
2169
  "+N more". Without it a multi-select trigger grows a line per selection and pushes the rest of the form
1906
- down the page, which is how it was reported. Same input name, same transform and same wording as
1907
- `ds-multi-select`, which already had it — the two multi-value pickers should not diverge on this.
2170
+ down the page. Same input name, same transform and same wording as `ds-multi-select`, which
2171
+ already had it — the two multi-value pickers should not diverge on this.
1908
2172
  Backspace still removes the last SELECTED chip, not the last visible one.
1909
2173
 
1910
2174
  - **`ds-stat` takes a projected value** — `[dsStatValue]` for a tile whose headline is a badge, a status
@@ -1917,7 +2181,7 @@ genuine gap worth reporting.
1917
2181
  - **The package now ships `design-system.md` and `components.md`.** `design-system.md` is no longer a
1918
2182
  catalogue: it carries the RULES — which spacing token goes between what, which type role a piece of
1919
2183
  text takes, the field-height contract, who owns a gap, and how to override a component (including the
1920
- two things that break it: reaching for an internal class, and portaled panels). The 491-line inventory
2184
+ two things that break it: reaching for an internal class, and portaled panels). The inventory
1921
2185
  it used to be moved to `components.md`. Both are readable from `node_modules`, so a consumer's agents
1922
2186
  and tooling can follow the system instead of guessing plausible token values.
1923
2187
 
@@ -1983,7 +2247,7 @@ genuine gap worth reporting.
1983
2247
  change unless set.
1984
2248
 
1985
2249
  - **`--ds-table-head-pad-top` / `--ds-table-head-pad-bottom`.** Header cells were the one padding in the
1986
- table that stayed symmetric — body cells have had per-side hooks since 07-31 — so a header wanting its
2250
+ table that stayed symmetric — body cells have had per-side hooks since 0.7.1 — so a header wanting its
1987
2251
  label tight to the top with air above the rule could only be reached with `!important`. Both fall back to
1988
2252
  the density value, so nothing moves until one is set.
1989
2253
 
@@ -2008,23 +2272,74 @@ genuine gap worth reporting.
2008
2272
  app with one rounded panel and no hook to reach it. Nothing moves by default — the role still resolves to
2009
2273
  `--radius-lg` — and the token is forwarded onto the portaled pane on open like the other panel tokens.
2010
2274
 
2011
- - **`ds-app-layout` takes a `scrollReset` strategy, and `LayoutNavigationService` is public** (both asked
2012
- for by the ai-sdlc starter after their `ds-app-layout` migration). Owning the scroll CONTAINER had
2013
- quietly meant owning the POLICY: `scrollToTop()` was public so a screen could ADD a reset, but nothing
2014
- could suppress one the layout decided to do. `scrollReset: 'path' | 'never' | 'always'` defaults to
2015
- `'path'`, which is exactly today's behaviour — no consumer sees a change unless they set it. `'always'`
2016
- also resets on a query- or fragment-only navigation (what a server-paged list wants); `'never'` hands the
2017
- decision to the screen.
2275
+ - **`ds-app-layout` takes a `scrollReset` strategy, and `LayoutNavigationService` is public.** Owning the
2276
+ scroll CONTAINER had quietly meant owning the POLICY: `scrollToTop()` was public so a screen could ADD a
2277
+ reset, but nothing could suppress one the layout decided to do. `scrollReset: 'path' | 'never' | 'always'`
2278
+ defaults to `'path'`, which is exactly today's behaviour — no consumer sees a change unless they set it.
2279
+ `'always'` also resets on a query- or fragment-only navigation (what a server-paged list wants); `'never'`
2280
+ hands the decision to the screen.
2018
2281
  — Exporting the navigation service removes a duplication the layout was causing: "did the path change?"
2019
- was being re-derived per consumer, and one of them had THREE answers that disagreed, because a
2020
- `split('?')` keeps the fragment `pathChanged` strips. It is root-provided and stateless, so exposing it
2021
- costs nothing.
2282
+ was being re-derived ad hoc wherever it was needed, and a naive `split('?')` check keeps the fragment
2283
+ that `pathChanged` strips, so two implementations could disagree on the answer. It is root-provided and
2284
+ stateless, so exposing it costs nothing.
2022
2285
 
2023
- - **`ds-icon` warns in dev when a name resolves to nothing** (asked for by the ai-sdlc starter after an audit
2024
- found misspelled names on three screens). An unknown name rendered an empty `<svg>` in silence: the types
2025
- cannot catch it, because `IconName` stays open (`| (string & {})`) so a consumer can register their own
2026
- glyphs through `provideDsIcons()`, and closing the union would reject every registered name. The warning
2027
- names the offending value and sits on the one branch that knows — production pays a single falsy check.
2286
+ - **`ds-icon` warns in dev when a name resolves to nothing.** An unknown name rendered an empty
2287
+ `<svg>` in silence: the types cannot catch it, because `IconName` stays open (`| (string & {})`)
2288
+ so a consumer can register their own glyphs through `provideDsIcons()`, and closing the union
2289
+ would reject every registered name. The warning names the offending value and sits on the one
2290
+ branch that knows — production pays a single falsy check.
2291
+
2292
+ ### Changed
2293
+
2294
+ - **`--ds-radius-inner` now reaches the small affordances it was written for.** Dropdown options, field
2295
+ clear buttons, calendar days, chips, and tag and toast close buttons took `var(--radius-sm)` directly
2296
+ instead of the role, so the role a consumer is told to retune reached exactly one component. Nothing
2297
+ moves (the role's default IS `--radius-sm`), but it changes what a square brand costs: zeroing the three
2298
+ radius roles used to leave five surfaces rounded and now squares every one. Five focus-ring radii inside
2299
+ `:focus-visible` keep following the raw scale deliberately — a ring traces a box, it is not a surface a
2300
+ brand designs.
2301
+
2302
+ - **`ds-auth-layout` compresses its own chrome on a short viewport, so the layout is never the reason a
2303
+ card scrolls.** Block padding, the column gap, the letter-mark tile and the masthead's inner gap clamp
2304
+ against `svh`. At ~330px of viewport — a landscape phone — that furniture costs ~131px instead of ~219px.
2305
+ Every multiplier reaches its maximum at a viewport **800px tall or more** — a viewport height, not a
2306
+ screen size: a 1366x768 laptop sits near 630px once browser and OS chrome come off, and a 1080p screen at
2307
+ 150% display scaling near 585px, so a good many desktops render the compressed furniture too.
2308
+ **Type is deliberately excluded**, because clamping it buys no vertical space — the wordmark shares a flex row with the tile, whose height
2309
+ exceeds the wordmark's line box at every viewport height, and the glyph sits in a fixed-size grid box —
2310
+ while shrinking the brand against a fixed 14px tagline, worst for the users who scale their display up.
2311
+ Both are back to the fixed sizes 0.8.0 ships.
2312
+ `svh` rather than `dvh` because the dynamic unit changes as a mobile URL bar hides, which would re-space
2313
+ the column mid-scroll; `min-height` stays on `dvh` so no strip of body background shows when the bar
2314
+ retracts. New opt-outs: `--ds-auth-layout-pad-block` and `--ds-auth-layout-mark-size` (set
2315
+ either, or the existing `-gap`, to a fixed value and that dimension stops compressing). A card taller than
2316
+ the space left still scrolls — deliberately, because the alternative (pinning the masthead and footer out
2317
+ of the flow) frees more room but lets chrome overlap a tall form.
2318
+
2319
+ - **`ds-auth-layout` — a round of additional fixes.** All additive or corrective; nothing to migrate.
2320
+ The letter-mark is now `aria-hidden`, so a screen reader announces "MyAdmin" rather than
2321
+ "A MyAdmin"; it also carries `flex-shrink: 0`, so a long wordmark can no longer squash the square tile
2322
+ into a rectangle. `logoLetter` on its own renders its tile instead of nothing (the outer gate never
2323
+ tested it). The glyph is taken with a code-point split rather than `charAt(0)`, which returned half a
2324
+ surrogate pair for a brand starting with an emoji. `box-sizing: border-box` is declared on the host
2325
+ rather than inherited, because the documented à-la-carte style path omits `styles/reset` and without it
2326
+ `min-height: 100dvh` plus the block padding overflows by exactly that padding — the scrollbar this
2327
+ component exists to avoid. The four `env(safe-area-inset-*)` calls gained `0px` fallbacks and moved to
2328
+ physical `padding-left`/`padding-right`, since the safe area belongs to the device and must not swap
2329
+ sides with text direction; note the insets are inert unless the **host app** sets `viewport-fit=cover`
2330
+ in its viewport meta, which the library cannot do and did not previously say. New tokens, both
2331
+ defaulting to today's behaviour: `--ds-auth-layout-mark-radius` (the tile's radius was the one value
2332
+ with no override hop) and `--ds-auth-layout-brand-max-height` (`none` by default — set it to cap a
2333
+ projected `[brand]` logo, which nothing constrained). New input `landmark` (`'none'` by default): the
2334
+ card holds the whole content of an auth screen but `ds-card` renders an `<article>`, which is not a
2335
+ landmark, so `landmark="main"` gives a screen-reader user something to jump to on a standalone auth
2336
+ page. It is opt-in because the layout is also embedded inside pages that already own a `main`.
2337
+
2338
+ - **BREAKING: `ds-auth-shell` is renamed `ds-auth-layout`** (`AuthShellComponent` → `AuthLayoutComponent`,
2339
+ and `--ds-auth-shell-*` → `--ds-auth-layout-*`). "Shell" is not this library's word for a layout — the
2340
+ app-level one was renamed `ds-app-shell` → `ds-app-layout` back in 0.7.0, and shipping an auth counterpart
2341
+ called a shell in 0.8.0 reintroduced the split vocabulary that rename existed to remove. Mechanical:
2342
+ rename the tag, the class and the six tokens. Nothing about the behaviour or the slots changes.
2028
2343
 
2029
2344
  ### Fixed
2030
2345
 
@@ -2078,8 +2393,7 @@ genuine gap worth reporting.
2078
2393
  focus only, and opening from the calendar glyph moves focus into the day grid — so the trigger went dark
2079
2394
  while its panel was open, and the refocus on close made it read as a flicker.
2080
2395
 
2081
- - **`ds-table`: the three things that still forced a consumer wrapper.** All from mm-broker folding its
2082
- 2160-line `ids-data-table` fork onto the base.
2396
+ - **`ds-table` gains a header filter pipeline and a live region for sort and row range.**
2083
2397
  — **`filterMode: 'client' | 'server' | 'none'`**, mirroring `sortMode` — it shipped defaulting to
2084
2398
  `'client'` and the default was flipped to `'server'` later in this same release, above — plus
2085
2399
  **`columnFilterOptions`** for the real value domain. A header filter on a server-paged table was wrong
@@ -2097,11 +2411,9 @@ genuine gap worth reporting.
2097
2411
  and `rangeAnnouncement` take the facts and return the sentence, so translation and word order stay with
2098
2412
  the consumer.
2099
2413
 
2100
- - **`ds-tabs` keeps the active tab in the URL: `urlParam="tab"`** (asked for by the ai-sdlc starter). The
2101
- input both switches it on and names the parameter; leave it out and the URL is untouched. Names come from
2102
- each `ds-tab`'s `value` — the identifier `tabClosed` already emits — so there is no second list to keep in
2103
- step. The evidence for shipping it: in one admin port THREE of the four tabbed screens never got the
2104
- hand-wiring at all, so only the screen whose author reached for a helper worked.
2414
+ - **`ds-tabs` keeps the active tab in the URL: `urlParam="tab"`.** The input both switches it on and
2415
+ names the parameter; leave it out and the URL is untouched. Names come from each `ds-tab`'s `value` —
2416
+ the identifier `tabClosed` already emits — so there is no second list to keep in step.
2105
2417
  — Names, not indices (`?tab=1` would repoint every old link the moment a tab is inserted); the first tab
2106
2418
  writes no parameter and every write replaces (clean address, and flipping tabs does not fill the Back
2107
2419
  button); an unknown or missing name lands on tab 0 rather than throwing. Outside a routed context the
@@ -2110,33 +2422,72 @@ genuine gap worth reporting.
2110
2422
  - **Options in the picker panels had no pressed state.** Hover only — so on a phone, where there is no
2111
2423
  hover, tapping a row in `ds-select`, `ds-autocomplete` or `ds-multi-select` gave no feedback that it had
2112
2424
  been hit. The equivalent rows elsewhere (`ds-menu-item`, `ds-list-item`) have had one since 0.7.x. Fixed
2113
- once in the shared dropdown-overlay partial, so all three pickers get it, with a selected-row variant that
2114
- stays in the brand family rather than stranding the selected label colour on a neutral fill.
2425
+ once, so all three pickers get it consistently, with a selected-row variant that stays in the brand
2426
+ family rather than stranding the selected label colour on a neutral fill.
2115
2427
 
2116
2428
  - **`ds-textarea size="sm"` and `="lg"` got the md vertical padding.** The size variants set the inner
2117
2429
  `<textarea>`'s vertical pad (`--space-2` / `--space-4`), but the `:host`-prefixed pin at the end of the
2118
- file — added so the docs' unencapsulated `.field .control` specimen could not zero it — out-ranks them
2430
+ file out-ranks them
2119
2431
  once emulated encapsulation shims both selectors: `[_nghost] .field[_ngcontent] .control[_ngcontent]
2120
2432
  textarea[_ngcontent]` carries one more attribute than `.field.sm[_ngcontent] …`. So every textarea
2121
- rendered with 12px top and bottom regardless of `size`, and retuning
2122
- `--ds-input-pad-y` moved all three. There are now three pins, one per size, and the size blocks keep
2433
+ rendered with 12px top and bottom regardless of `size`, and retuning `--ds-input-pad-y` moved all
2434
+ three. There are now three pins, one per size, and the size blocks keep
2123
2435
  only the horizontal padding they own. Consumers see it only on a non-md textarea; the md default is
2124
2436
  unchanged.
2125
2437
 
2438
+ ### What you can now DELETE
2439
+
2440
+ The half a release note usually leaves out: workarounds you may have written against hooks that
2441
+ already existed, with no signal that anything had been tokenised since. None of this is required;
2442
+ skipping it leaves working code that is merely redundant.
2443
+
2444
+ | If you have | Delete it, because |
2445
+ | --- | --- |
2446
+ | `.button.button.button { … }` or rules against `.v-outline` / `.c-secondary` | `ds-button` reflects `[data-variant]` and `[data-color]` on the HOST. Use `ds-button[data-variant="outline"]` |
2447
+ | `--radius-sm: 0` to square a checkbox | `--ds-checkbox-radius` |
2448
+ | `.title.title { font-size }` inside a modal | `--ds-modal-title-font-size` |
2449
+ | Pagination label type overrides — a rule on `ds-pagination` setting `--ds-pagination-label-*` | `--ds-pagination-label-*` |
2450
+ | `.size { display: none }` on a pager | `[showPageSize]="false"` — and it is now forwarded through `ds-table` too |
2451
+ | `.actions:empty { display: none }` on a modal | The library already ships it |
2452
+ | `--ds-modal-divider: transparent`, plus whatever you rewrote the modal padding to afterwards | `[dividers]="false"` — it drops the rules *and* the padding that existed to clear them |
2453
+ | A defensive `:hover` rule guarding the selected dropdown row — a selector naming `is-selected` | The `:not(.is-selected)` branch already handles it |
2454
+ | `ds-menu-divider { padding: 0 }` | `--ds-menu-divider-spacing`, which already defaults to 0 |
2455
+ | A dozen scattered `--ds-*-radius: 0` | For a fully square brand, four lines: `--radius-sm/md/lg/xl: 0` (leave `--radius-pill`). Measured — the three ROLES alone leave options, clear buttons and dropdown panels rounded; they are the tool for squaring one family and keeping another. Worked example on the Theming page |
2456
+ | A global rule squaring `.ds-overlay-pane` | Radius and elevation roles now travel onto portaled panes when the host overrides them |
2457
+ | Absolutely positioning a search glyph over an autocomplete, or hiding `.autocomplete-chevron` | `dsAutocompletePrefix` and `[showChevron]="false"` |
2458
+ | Querying the autocomplete's inner `<input>` and dispatching a synthetic `input` event to clear it | `clear()` |
2459
+ | A hand-built KPI tile — a `kpi` block pairing your own markup with `ds-badge`, where `ds-stat` now fits | `ds-stat` takes a projected `[dsStatValue]` |
2460
+ | A pinned `--ds-search-height` (or any sibling) to get a shorter control row | `density="compact"` on the container — every field control is 32px there, measured |
2461
+ | A hand-rolled `.scrim` + `.panel` dialog, written because the modal header could not hold a badge or a second line | `ds-modal` projects `[title-badge]` and takes a `subtitle`. You get back the focus trap, focus restore, the scroll lock, Escape and the footer |
2462
+ | `[value]="f.x().value()" (valueChange)="f.x().value.set($event)"` on every field of a signal form | `[formField]="f.x"` — the control implements the contract already |
2463
+ | `white-space: normal` forced onto a badge in a table cell | `--ds-badge-white-space` |
2464
+ | `::ng-deep .card { height: 100% }` for equal-height cards | The card's box fills a stretched host on its own |
2465
+ | `!important` on a `--ds-table-cell-pad` override under `density="compact"` | Compact no longer out-specifies the padding hooks |
2466
+ | A wrapper around `ds-table` for sort/range announcements, a `role="separator"` resize handle, or a header filter the checklist cannot express | All three ship: the live region, the `role="separator"` handle, and `dsHeaderFilter` |
2467
+ | `new Event('input'`, `dispatchEvent`, `querySelector` | Querying the autocomplete's inner `<input>` and dispatching a synthetic `input` event to clear it Replace with: `clear()`. |
2468
+ | `--ds-autocomplete-input-minw` | A --ds-autocomplete-input-minw override, written to hide the empty filter box that sat beside a chosen value in single mode. The input collapses on its own once a value is picked and the panel closes. Replace with: nothing - unless you genuinely want a wider typing area. |
2469
+ | `.trigger`, `:focus-within`, `ds-date-picker` | A :focus-within rule on the date-picker trigger, written because the trigger went dark while its panel was open. The component keeps an "open" class alongside the focus rules now. Replace with: nothing - the component holds the indicator itself. |
2470
+ | A wrapper around the export button catching a rejected `exportFetcher` — a `try {` / `catch` around it | ds-table emits `(exportFailed)` now. Replace with: bind (exportFailed) on ds-table. |
2471
+ | Your own live region around `ds-table` — an `aria-live` or `role="status"` element announcing sort state or row range | The table announces both through its own role="status" region, so keeping yours announces every change twice. Replace with: nothing - retune the wording through sortAnnouncement and rangeAnnouncement. |
2472
+
2473
+ **To sweep for the rest:** grep your styles for `::ng-deep`, `!important` and any selector naming a
2474
+ DS internal class (`.button`, `.trigger`, `.option`, `.modal`). Each hit is either a row above or a
2475
+ genuine gap worth reporting.
2476
+
2126
2477
  ## 0.8.0 — 2026-08-05
2127
2478
 
2128
2479
  ### Added
2129
2480
 
2130
2481
  - **`ds-icon-button` forwards `ariaExpanded` and `ariaControls` to its inner `<button>`, and `ds-button` gains `ariaControls`.** A trigger that opens a panel has to announce state on the element a screen reader focuses; set on `<ds-icon-button>` itself those attributes land on the wrapper, which is not focusable and is never read. `ds-button` already forwarded `ariaExpanded`; the pair is now complete on both.
2131
2482
  - **`LayoutScrollService` exposes its scroller and an absolute `scrollTo(top)`.** A scroll-spy or a virtualizer inside `ds-app-layout` needs the content region to measure against, and the only way to get it was `document.querySelector` against the layout's internal markup — a consumer reaching into DS internals for something the DS already holds. `scrollTo` assigns `scrollTop` rather than calling `scrollTo({ behavior })`, so a themed `scroll-behavior: smooth` cannot turn a positioning jump into an animation.
2132
- - **`dsRailTrigger` wires a consumer's own burger to the overlay rail.** The responsive rail shipped with `role="dialog"`, `aria-modal` and a focus trap, but nothing connected it to the button that opens it — the burger lives in the consumer's projected topbar, so it announced no state and no relationship. Put the directive on the burger and it toggles `[(railOpen)]` and keeps `aria-expanded` and `aria-controls` correct. Both attributes are dropped while the rail is not an overlay, where `aria-expanded="false"` would contradict what the user sees. A directive rather than a documentation note because these are exactly the attributes a consumer forgets — the DS itself forgot them. `ds-app-layout` also exposes `railId` and the read-only `railOverlay` signal for a consumer that builds its own trigger.
2483
+ - **`dsRailTrigger` wires a consumer's own burger to the overlay rail.** The responsive rail shipped with `role="dialog"`, `aria-modal` and a focus trap, but nothing connected it to the button that opens it — the burger lives in the consumer's projected topbar, so it announced no state and no relationship. Put the directive on the burger and it toggles `[(railOpen)]` and keeps `aria-expanded` and `aria-controls` correct. Both attributes are dropped while the rail is not an overlay, where `aria-expanded="false"` would contradict what the user sees. A directive rather than a documentation note because these are exactly the attributes a consumer forgets. `ds-app-layout` also exposes `railId` and the read-only `railOverlay` signal for a consumer that builds its own trigger.
2133
2484
 
2134
2485
  - **Overlay panels fade out when DISMISSED, and still close instantly when you choose something.** Floating
2135
2486
  panels (select, autocomplete, multi-select, both pickers, colour picker, popover) have closed instantly
2136
2487
  since the unconditional leave animation was cut, because ~120ms of fade after picking an option reads as the
2137
2488
  panel lagging behind a decision already made. A dismissal is the opposite: Escape, a press outside or the
2138
2489
  page scrolling away produced no result, so the fade reads as the panel backing out rather than as delay.
2139
- `disposeWithLeave()` now takes that cause — `'select'` (the default, instant) or `'dismiss'` — and each
2490
+ Teardown now takes that cause — `'select'` (the default, instant) or `'dismiss'` — and each
2140
2491
  panel passes `'dismiss'` from exactly its Escape / outside-press / scroll paths. Tab still closes instantly,
2141
2492
  because a fading panel must not linger behind moved focus.
2142
2493
  — Tunable through `--ds-overlay-leave-duration` / `-easing`, defaulting to the `--ds-motion-exit-*` role.
@@ -2155,8 +2506,7 @@ genuine gap worth reporting.
2155
2506
  class, so setting it pins the caption regardless of `size`. These are deliberately NOT part of the
2156
2507
  `--ds-field-label-*` contract: a label above an input is a heading for the field and reads secondary, while
2157
2508
  a caption beside a checkbox is the content being agreed to and reads primary — routing them together would
2158
- put "Remember me" in uppercase in any brand that sets `--ds-field-label-transform`. Recorded in the token
2159
- reference so it reads as a decision rather than an omission.
2509
+ put "Remember me" in uppercase in any brand that sets `--ds-field-label-transform`.
2160
2510
  - **`ds-table` renders only the rows in view: `[virtualScroll]="true"` + `virtualRowSize`.** The body used to
2161
2511
  `@for` over every row of `data()`, so 10 000 rows tanked scroll and paint. Virtual mode keeps ~25 rows in
2162
2512
  the DOM regardless of the row count, padding above and below with two spacer rows that stand in for the
@@ -2177,8 +2527,8 @@ genuine gap worth reporting.
2177
2527
  parity depends on its real position, and the spacer rows are children too.
2178
2528
  - **`ds-filter-bar` — a Priority+ filter row: as many filters as fit, the rest behind "+N more".**
2179
2529
  Every admin list past three filters hits the same wall — the row wraps to a second line on a laptop, or
2180
- the consumer hand-rolls an overflow menu. Each filter goes in an
2181
- `<ng-template [dsFilterItem]="180">`, and the split is decided from those declared widths plus one
2530
+ the consumer hand-rolls an overflow menu. Each filter goes in an `<ng-template
2531
+ [dsFilterItem]="180">`, and the split is decided from those declared widths plus one
2182
2532
  `ResizeObserver` on the row. Never by measuring the filters: they would have to be in the DOM to be
2183
2533
  measured, so anything that did not fit would flash into view before being taken away.
2184
2534
  — A template rather than plain projection because the bar has to MOVE what does not fit into the panel,
@@ -2201,24 +2551,24 @@ genuine gap worth reporting.
2201
2551
  Consumers hand-rolled the unauthenticated scaffold every time, and it is the same four parts in each: a
2202
2552
  brand lockup, an optional tagline, one card holding the form, and a support line under it. The brand row
2203
2553
  is the sidenav's row — same `brand` / `logoLetter` / `[dsBrandName]` contract and the same letter-mark
2204
- tile, so the mark on the login screen is the mark in the rail afterwards. Deliberately NOT the string
2205
- splitting the ports used to tint one slice of a wordmark: a split has to guess where the accent belongs,
2206
- and the projection already says.
2554
+ tile, so the mark on the login screen is the mark in the rail afterwards. The accent colour on the
2555
+ wordmark comes from the same projection rather than from splitting a string, so there is no guessing
2556
+ where the accent belongs.
2207
2557
  — The card body is the default slot, the support line is `[dsAuthSupport]` (a real `<footer>`, so it
2208
2558
  carries the `contentinfo` landmark), and both the tagline and the support line collapse when absent
2209
2559
  rather than leaving an empty flex child that would take the column's gap and push the card off centre.
2210
2560
  — `cardElevation` / `cardPadding` pass through to `ds-card`; `--ds-auth-shell-min-height`,
2211
2561
  `-card-max-width`, `-gap`, `-bg-color` and the two `-mark-*` colours are the public hooks.
2212
- — The host is the ONLY full-height box, which is what the ports got wrong: nesting a `100vh` wrapper
2213
- inside a `100dvh` one scrolls a page whose content fits. It also carries `env(safe-area-inset-*)` so a
2214
- notched phone in landscape does not put the card under the cutout. Inside it the column centres with auto
2215
- margins rather than centre alignment — identical while the host can grow, but not when it cannot: embed
2216
- the shell at `--ds-auth-shell-min-height: 100%` in a fixed-height parent and centre alignment puts a form
2217
- taller than the box above the scroll origin where no scrolling reaches it (measured at −115px), while
2218
- auto margins collapse to zero and pin it to the padding edge.
2219
- - **`ds-password-rules` — the live requirement checklist under a new-password field.** One consumer had
2220
- written it twice (set-password and registration) and the two copies had already drifted: different labels
2221
- for the same rule, a different order, different state class names. Each rule carries its own `test`,
2562
+ — The host is the ONLY full-height box: nesting a `100vh` wrapper inside a `100dvh` one scrolls a page
2563
+ whose content fits. It also carries `env(safe-area-inset-*)` so a notched phone in landscape does not
2564
+ put the card under the cutout. Inside it the column centres with auto margins rather than centre
2565
+ alignment — identical while the host can grow, but not when it cannot: embed the shell at
2566
+ `--ds-auth-shell-min-height: 100%` in a fixed-height parent and centre alignment puts a form taller than
2567
+ the box above the scroll origin where no scrolling reaches it, while auto margins collapse to zero and
2568
+ pin it to the padding edge.
2569
+ - **`ds-password-rules` — the live requirement checklist under a new-password field.** A password policy
2570
+ screen tends to get hand-rolled more than once, and the copies drift: different labels for the same
2571
+ rule, a different order, different state class names. Each rule carries its own `test`,
2222
2572
  because the POLICY has to mirror the server and only the consumer knows it; the component owns the
2223
2573
  tri-state glyph, the met/failed colours and the type scale. A rule that has not passed reads as PENDING
2224
2574
  and only becomes a failure once `[showUnmet]` is true — bind that to "the user has tried to submit", or a
@@ -2233,11 +2583,10 @@ genuine gap worth reporting.
2233
2583
  screen reader reads, and replacing the attribute would have dropped the field's own hint/error
2234
2584
  association. The input joins both ids instead.
2235
2585
 
2236
- - **`ds-table` persists the column layout and can render its own export button.** These were the last two
2237
- things keeping a consumer wrapper alive around the table: after 0.6.3/0.7.0 it owned every other feature
2238
- mm-broker forked it for, but the three column models were two-way with nothing storing them, so every
2239
- consumer wrote the same `localStorage` effect keyed by a screen id, and `exportCsv()` was a public method
2240
- with no affordance.
2586
+ - **`ds-table` persists the column layout and can render its own export button.** The three column models
2587
+ — order, widths, hidden columns — were two-way with nothing storing them, so every consumer wrote the
2588
+ same `localStorage` effect keyed by a screen id, and `exportCsv()` was a public method with no
2589
+ affordance to trigger it from.
2241
2590
  — `[persistLayout]="true"` + `storageKey` round-trip `columnOrder`, `columnWidths` and `hiddenKeys`
2242
2591
  through `localStorage`. Both are required: without a key there is nothing to identify a screen by and two
2243
2592
  tables would overwrite each other, so a missing key stores nothing rather than guessing. Only what was
@@ -2272,10 +2621,11 @@ genuine gap worth reporting.
2272
2621
  `--ds-radius-pill`. No `field` role: fields share `--radius-md` with buttons, so one would either
2273
2622
  duplicate `control` or silently re-round every field — scope the role instead
2274
2623
  (`ds-input { --ds-radius-control: 4px; }`).
2275
- — **Elevation:** `--ds-elevation-raised` / `-overlay` / `-floating` / `-modal`. Four, not the three first
2276
- sketched: the library genuinely uses four heights, and folding toast into `overlay` would have changed its
2277
- shadow. The older `--shadow-popover` / `-modal` / `-toast` / `-card` aliases now point AT the roles, so
2278
- either name works — a role moves every surface at that height, an alias retunes one family.
2624
+ — **Elevation:** `--ds-elevation-raised` / `-overlay` / `-floating` / `-modal`. Four roles: the
2625
+ library genuinely uses four heights, and folding toast into `overlay` would have changed its
2626
+ shadow. The older `--shadow-popover` / `-modal` / `-toast` / `-card` aliases now point AT the
2627
+ roles, so either name works — a role moves every surface at that height, an alias retunes one
2628
+ family.
2279
2629
  — **Motion:** `--ds-motion-enter-panel-*`, `-enter-surface-*`, `-enter-large-*`, `-exit-*`,
2280
2630
  `-exit-large-*` (duration + easing each). Five roles, not four: panel LEAVES run at `--duration-base`,
2281
2631
  not `fast`, so folding them into `exit` would have sped them up. `ds-loading-overlay` and the toast leave
@@ -2324,8 +2674,8 @@ genuine gap worth reporting.
2324
2674
  box, and retuning them from an app needed `.field.field .label.label, .select-label.select-label` to
2325
2675
  out-rank the shimmed component rule. `input`, `textarea`, `select`, `date-picker`, `time-picker` and
2326
2676
  `radio-group` now resolve through one mixin and one token set. No per-component token set is needed —
2327
- custom-property scoping already narrows a retune to one component
2328
- (`ds-select { --ds-field-label-transform: uppercase; }`), and `-transform`/`-tracking` make the
2677
+ custom-property scoping already narrows a retune to one component (`ds-select {
2678
+ --ds-field-label-transform: uppercase; }`), and `-transform`/`-tracking` make the
2329
2679
  uppercase tracked editorial label reachable through tokens for the first time. `ds-input`’s older
2330
2680
  `--ds-input-label-color`/`-fs`/`-fw` still win where set, and now fall through to the family token
2331
2681
  instead of straight to the raw scale.
@@ -2358,7 +2708,20 @@ genuine gap worth reporting.
2358
2708
 
2359
2709
  ### Changed
2360
2710
 
2361
- - **`ds-table`’s `clientSort` is replaced by `sortMode: 'client' | 'server' | 'none'`, and the default flips to client sorting.** The boolean could not express "nothing sorts", so a column marked `sortable: true` on a table with no sorting drew a header that looked live, showed a caret on hover and did nothing — an invisible failure that survived review and was caught by opening the page. An enum makes that state unrepresentable: `'none'` now renders the header as plain text, with no button, caret or `aria-sort`. Unset resolves to `'client'`, so `sortable: true` always does something; a server-sorted table declares `'server'`, which is the side that actually knows it wants server sorting. **Migration is mechanical except for one case:** a table that previously bound no `clientSort` and reordered in its own `(sortChange)` handler must now declare `sortMode="server"`, or it will sort locally as well. For a server-PAGED table (one that binds `total`) that is worse than a dead header, because only the current page reorders and the rows look sorted without being sorted — so dev builds warn on exactly that combination: `total` bound, sortable columns, `sortMode` never declared. Rejected alternative: reading `OutputEmitterRef.listeners` to detect an unbound `(sortChange)`. It is private, and a rename would kill the warning SILENTLY — `?.length` on `undefined` just means it never fires again. The union is exported as `TableSortMode`, so a host that keeps the mode in a signal or computed can type it instead of repeating the literals.
2711
+ - **`ds-table`’s `clientSort` is replaced by `sortMode: 'client' | 'server' | 'none'`, and the
2712
+ default flips to client sorting.** The boolean could not express "nothing sorts", so a column
2713
+ marked `sortable: true` on a table with no sorting drew a header that looked live, showed a caret
2714
+ on hover and did nothing when clicked. An enum makes that state unrepresentable: `'none'` now
2715
+ renders the header as plain text, with no button, caret or `aria-sort`. Unset resolves to
2716
+ `'client'`, so `sortable: true` always does something; a server-sorted table declares `'server'`,
2717
+ which is the side that actually knows it wants server sorting. **Migration is mechanical except
2718
+ for one case:** a table that previously bound no `clientSort` and reordered in its own
2719
+ `(sortChange)` handler must now declare `sortMode="server"`, or it will sort locally as well. For
2720
+ a server-PAGED table (one that binds `total`) that is worse than a dead header, because only the
2721
+ current page reorders and the rows look sorted without being sorted — so dev builds warn on
2722
+ exactly that combination: `total` bound, sortable columns, `sortMode` never declared. The union is
2723
+ exported as `TableSortMode`, so a host that keeps the mode in a signal or computed can type it
2724
+ instead of repeating the literals.
2362
2725
 
2363
2726
  - **A `ds-menu` row's hover fill is square and runs edge to edge.** The row carried `--radius-md`, the
2364
2727
  same 8px the standalone menu panel uses, while sitting flush inside it — and inside a `ds-popover`,
@@ -2382,33 +2745,28 @@ genuine gap worth reporting.
2382
2745
  `--combobox-maxh` size the listbox panel shared by `ds-autocomplete` and `ds-multi-select`, and survived the
2383
2746
  `ds-combobox` → `ds-autocomplete` rename carrying the name of a component that no longer exists. Now
2384
2747
  `--picker-offset` / `--picker-maxh`. These are global scale tokens rather than `--ds-*` component tokens, so
2385
- the blast radius is anyone who happened to override them — rename the override. The docs also now state the
2386
- autocomplete-vs-multi-select boundary in both sections: the LIST decides, not the count — typing is how you
2748
+ the blast radius is anyone who happened to override them — rename the override. Choosing between the two: the LIST decides, not the count — typing is how you
2387
2749
  narrow a set too large to scan, while a checkbox list shows the whole choice and what is NOT selected.
2388
2750
  - **A `required` `ds-time-picker` in `mode="range"` never announced as required.** `aria-required` was on the
2389
2751
  single-mode trigger only, so a screen-reader user on a range field was not told either segment was
2390
2752
  mandatory. Both segments now carry it. (Their accessible NAME was already correct — each announces the
2391
- field's own label plus "(from)" / "(to)" — and the docs gained a labelled, required range demo, whose
2392
- absence is why the gap went unnoticed.)
2753
+ field's own label plus "(from)" / "(to)".)
2393
2754
  - **`ds-table`: with `stickyHeader` and `stickyFirstColumn` together, the first body row painted OVER the
2394
2755
  header.** `--table-sticky-corner-z` exists so the corner where both sticky axes meet sits above the body's
2395
2756
  pinned cells, but the sticky-header rule out-specified it, leaving the pinned header cells at the same
2396
2757
  z-index as the pinned body cells — and on a tie DOM order wins, so the body won. The header row read as
2397
- the first row's data in the pinned columns. Found while building virtual scroll, and it was never specific
2398
- to it: any tall table with both flags had it. The sticky-header rule now leaves the pinned cells' z-index
2399
- to the corner token.
2758
+ the first row's data in the pinned columns, on any tall table with both flags set. The sticky-header rule
2759
+ now leaves the pinned cells' z-index to the corner token.
2400
2760
  - **`ng add` never ran its schematic — in any released version.** The package is `"type": "module"` while
2401
2761
  the compiled schematic is CommonJS, so Node refused to load it (`exports is not defined in ES module
2402
2762
  scope`) and the DS global styles were never wired automatically; every consumer either added the `@use`
2403
- line by hand or shipped without the styles. The build already wrote a nested
2404
- `schematics/package.json` with `{"type":"commonjs"}` to fix exactly this, and **`npm pack` silently drops
2405
- it** — npm treats a directory holding a `package.json` as a separate package root — so the marker existed
2406
- in `dist/` and was absent from the tarball. The schematic is now authored as `.cts` and ships as `.cjs`,
2407
- which survives packing, and the build fails loudly if that emit is missing.
2408
- - **`ds-menu-divider` added a gap on top of the menu's own row rhythm, so a separated group read as a hole.** The divider carried `padding: var(--space-1) 0` while `ds-menu` already spaces its rows with `gap: var(--space-0-5)`. The two stack: measured in a consumer's account menu, adjacent items sat 2px apart while the divider produced 6px + line + 6px, tripling the local rhythm exactly where the eye reads a boundary. Menu items already carry their own `8px 12px` padding, so the line needs no help to separate. The default padding is now **0** — the line takes one row's place instead of adding a second gap — and a menu that genuinely wants divider *plus* air opts in with the new **`--ds-menu-divider-spacing`** token. This follows the direction the platforms took: Material 3 Expressive treats "with gap" as a device SEPARATE from "with divider" rather than bundling them, and Angular Material exposes the same spacing as tokens (`--mat-menu-divider-top-spacing` / `-bottom-spacing`) after shipping it baked in. Visual-only, and only for menus that use the divider.
2409
- - **`ds-table` gave the page a second scrollbar when it sat below the fold.** Its host was `display: block` with no `position`, while the `.visually-hidden` nodes it renders — the caption, the two `role="status"` announcers and the per-header sort-state spans — are `position: absolute`. With no positioned ancestor they resolved against the initial containing block, so each one contributed its layout offset to the DOCUMENT's root scroller: a consumer whose shell is `100vh` + `overflow: hidden` with one inner scrolling region got a phantom second scrollbar beside it, growing with how far down the page the table sat. Measured in a consumer with three tables on one tab — the root scroller reported 1902px against a 728px viewport, 1174px of overflow that no visible content occupied; with the host positioned it is 0 and the app's own region still scrolls. The host is now `position: relative`. Nothing else moves: the SR nodes stay 1×1 and clipped, and the overlays that must escape the table are portaled rather than absolutely positioned inside it.
2763
+ line by hand or shipped without the styles. The schematic is now authored as `.cts` and ships as
2764
+ `.cjs`, which Node loads correctly under `"type": "module"`, so `ng add` wires up the global styles
2765
+ as documented.
2766
+ - **`ds-menu-divider` added a gap on top of the menu's own row rhythm, so a separated group read as a hole.** The divider carried `padding: var(--space-1) 0` while `ds-menu` already spaces its rows with `gap: var(--space-0-5)`. The two stack: adjacent menu items sat 2px apart while the divider produced 6px + line + 6px, tripling the local rhythm exactly where the eye reads a boundary. Menu items already carry their own `8px 12px` padding, so the line needs no help to separate. The default padding is now **0** — the line takes one row's place instead of adding a second gap — and a menu that genuinely wants divider *plus* air opts in with the new **`--ds-menu-divider-spacing`** token. This follows the direction the platforms took: Material 3 Expressive treats "with gap" as a device SEPARATE from "with divider" rather than bundling them, and Angular Material exposes the same spacing as tokens (`--mat-menu-divider-top-spacing` / `-bottom-spacing`) after shipping it baked in. Visual-only, and only for menus that use the divider.
2767
+ - **`ds-table` gave the page a second scrollbar when it sat below the fold.** Its host was `display: block` with no `position`, while the `.visually-hidden` nodes it renders — the caption, the two `role="status"` announcers and the per-header sort-state spans — are `position: absolute`. With no positioned ancestor they resolved against the initial containing block, so each one contributed its layout offset to the DOCUMENT's root scroller: a consumer whose shell is `100vh` + `overflow: hidden` with one inner scrolling region got a phantom second scrollbar beside it, growing with how far down the page the table sat. The host is now `position: relative`, which brings the root scroller's overflow back to 0 and leaves only the app's own region scrolling. Nothing else moves: the SR nodes stay 1×1 and clipped, and the overlays that must escape the table are portaled rather than absolutely positioned inside it.
2410
2768
  - **A nested `ds-app-layout` opened its rail whenever an OUTER one was open, and its scrim swallowed the clicks meant for the outer.** The open state was written as `:host(.rail-open) &`, which Angular's emulated encapsulation compiles to a DESCENDANT selector — and every instance of a component shares one encapsulation attribute, so `.rail-open[_nghost-x] .rail[_ngcontent-x]` matched the rail of every layout nested inside an open one. Two layouts on a page (a docs shell holding a demo, an app holding a preview pane) opened and closed in lockstep, and clicking the backdrop hit the inner layout's scrim, whose `closeRail()` was a no-op. The state now lives on the elements it styles (`.rail.open` / `.scrim.open`), where a selector cannot reach across host boundaries.
2411
- - **The overlay rail escaped its layout and covered the whole window.** Below `lg` the rail is `position: fixed`, which resolves against the viewport, so a layout embedded in a page threw its rail over the top-left of the window instead of its own box — three shells on one docs page stacked three rails and three full-page scrims. The rail and scrim are now `absolute` against a `position: relative` host, which is identical for a full-viewport shell and correct for an embedded one.
2769
+ - **The overlay rail escaped its layout and covered the whole window.** Below `lg` the rail is `position: fixed`, which resolves against the viewport, so a layout embedded in a page threw its rail over the top-left of the window instead of its own box, and multiple embedded layouts on one page each stacked a full-page scrim. The rail and scrim are now `absolute` against a `position: relative` host, which is identical for a full-viewport shell and correct for an embedded one.
2412
2770
  - **A nested `ds-app-layout` could paint its rail OVER the outer shell's.** The rail sits at
2413
2771
  `--layer-drawer` and its scrim one below; `position: relative` alone still left every layout resolving
2414
2772
  that layer in the one root stacking context, so an inner layout's rail could stack above an outer
@@ -2427,27 +2785,26 @@ genuine gap worth reporting.
2427
2785
  trigger's radius is untouched.
2428
2786
  - **A path change that only mirrors the reader's position no longer resets the content scroll.** A docs shell whose scroll spy rewrites `/area/section` as the user scrolls hit a loop: the rewrite looked like a page change, the layout scrolled to top, the spy saw the first section again and rewrote the path back. Such a navigation must now carry `state: { dsSamePage: true }`, which the layout skips for both the scroll reset and the rail close. Read from `Router.lastSuccessfulNavigation()` — `getCurrentNavigation()` is already cleared when `NavigationEnd` reaches a subscriber.
2429
2787
  - **Closing the rail by NAVIGATING no longer pulls focus back to the burger.** Restoring focus to the trigger is right for a dismissal — Escape, the scrim — and wrong for a link: the reader activated it, so focus belongs wherever the router put it, and handing it back announced "Open navigation" instead of the page they asked for. Dismissal keeps the restore; the navigation and breakpoint paths now clear `railOpen` directly. The breakpoint path had also depended on `compact` being written one line earlier to avoid focusing a trigger that is hidden on desktop — that ordering trap is gone with it.
2430
- - **A closed compact `ds-app-layout` rail no longer holds focus, or 112 tab stops, behind the screen.**
2431
- Below the breakpoint the rail is translated off-canvas, which hides it visually and not at all from
2432
- assistive tech: it kept `role="dialog"` + `aria-modal="true"`, every nav link stayed tabbable, and the
2433
- caret stayed inside the hidden panel after Escape or a scrim click — `cdkTrapFocusAutoCapture` hands
2434
- focus back only when the directive is DESTROYED, not when it flips off — so the next Tab continued from
2435
- invisible navigation with no visible focus ring. The closed overlay rail is now `inert`, and closing it
2436
- returns focus to whatever control claims the rail through `aria-controls`, which both `dsRailTrigger`
2437
- and the DS-button forwarding (`[ariaControls]="shell.railId"`) already set. Verified in Chrome at 604px.
2438
- At/above the breakpoint nothing changes — the rail stays in the layout and interactive.
2788
+ - **A closed compact `ds-app-layout` rail no longer holds focus behind the screen.** Below the
2789
+ breakpoint the rail is translated off-canvas, which hides it visually and not at all from
2790
+ assistive tech: it kept `role="dialog"` + `aria-modal="true"`, every nav link stayed tabbable, and
2791
+ the caret stayed inside the hidden panel after Escape or a scrim click — `cdkTrapFocusAutoCapture`
2792
+ hands focus back only when the directive is DESTROYED, not when it flips off — so the next Tab
2793
+ continued from invisible navigation with no visible focus ring. The closed overlay rail is now
2794
+ `inert`, and closing it returns focus to whatever control claims the rail through `aria-controls`,
2795
+ which both `dsRailTrigger` and the DS-button forwarding (`[ariaControls]="shell.railId"`) already
2796
+ set. At/above the breakpoint nothing changes — the rail stays in the layout and interactive.
2439
2797
 
2440
2798
  - **`--ds-overlay-enter-scale` now reaches the portaled panel, so `ds-scale-in` can be retuned per
2441
- instance.** The token was declared and resolved on the pane but missing from the set
2442
- `prepareOverlayEnter()` forwards, so `ds-select { --ds-overlay-enter-scale: 0.9; }` did nothing while
2443
- the other four enter tokens worked — 0.7.1 shipped it as a known gap. Covers every overlay panel
2799
+ instance.** The token was declared and resolved on the pane but missing from the set forwarded onto it, so `ds-select { --ds-overlay-enter-scale: 0.9; }` did nothing while
2800
+ the other four enter tokens worked. Covers every overlay panel
2444
2801
  (select, multi-select, autocomplete, date / time / colour picker, popover, tooltip, menu submenus). A
2445
2802
  global override always worked and is unaffected.
2446
2803
 
2447
2804
  ### What you can now DELETE
2448
2805
 
2449
2806
  None of this is required — skipping it leaves working code that is merely redundant. Each row is a
2450
- workaround this release made unnecessary. Generated from the release manifest’s `deleteList`.
2807
+ workaround this release made unnecessary. Generated from the `deleteList` in `migrations/0.8.0.json`.
2451
2808
 
2452
2809
  | If you have | Delete it, because |
2453
2810
  | --- | --- |
@@ -2462,17 +2819,16 @@ workaround this release made unnecessary. Generated from the release manifest’
2462
2819
  character and `Oct 09, 2023 / Feb 27, 2024` sorted by month NAME: the table looked sorted and was not,
2463
2820
  which is worse than leaving it unsorted. Pass an accessor per column —
2464
2821
  `{ key: 'amount', label: 'Amount', sortValue: (row) => row.amountCents }` — returning a number, `Date`
2465
- or plain string. A consumer that had to re-parse its own rendered output (mm-broker shipped a
2466
- `display-compare.ts` sniffing currency, percents and three date formats) can delete it. Only affects
2467
- sorting: the header filter still lists what the user can see. Columns with plain numeric text already
2468
- sorted correctly and need nothing.
2822
+ or plain string. A consumer that had to re-parse its own rendered output — sniffing currency,
2823
+ percents and date formats — can delete it. Only affects sorting: the header filter still lists
2824
+ what the user can see. Columns with plain numeric text already sorted correctly and need nothing.
2469
2825
 
2470
2826
  - **`ds-checkbox`, `ds-switch` and `ds-radio` warn in dev when a caption is projected instead of passed.**
2471
2827
  `<ds-checkbox>Accept the terms</ds-checkbox>` renders no caption — the text comes from `[label]` only —
2472
- and it failed silently: no error, no warning, nothing in the types, so it survived review and was caught
2473
- by opening the page. All three share that contract, so all three now log which component swallowed the
2474
- text. Production is untouched: the probe that detects it is rendered only under `isDevMode()`, so no
2475
- extra node ships. Real projection support is still open — see `docs/todo.md`.
2828
+ and it failed silently: no error, no warning, nothing in the types, so nothing caught it short of
2829
+ opening the rendered page. All three share that contract, so all three now log which component
2830
+ swallowed the text. Production is untouched: the probe that detects it is rendered only under
2831
+ `isDevMode()`, so no extra node ships. A projected caption still renders nothing.
2476
2832
 
2477
2833
  - **`ds-app-layout` collapses its rail to an off-canvas overlay below `lg`.** On narrow viewports the
2478
2834
  rail no longer occupies a grid column: below the `lg` breakpoint (read at runtime from `--bp-lg`) it
@@ -2506,9 +2862,8 @@ workaround this release made unnecessary. Generated from the release manifest’
2506
2862
  triggers 22.8px apart vertically. The label renders the same `.label` block with the same tokens as
2507
2863
  `ds-date-picker`, and drives the single-mode trigger's accessible name (`aria-labelledby` when a label
2508
2864
  is set, `aria-label` when it is not) plus `aria-required`; range mode's start/end segment triggers are
2509
- not wired to the label yet.
2510
- Measured on a two-column probe, date vs time now differ by 0px on label top/height, trigger
2511
- top/height/bottom, value inset and glyph top.
2865
+ not wired to the label yet. Measured on a two-column probe, date vs time now differ by 0px on label
2866
+ top/height, trigger top/height/bottom, value inset and glyph top.
2512
2867
 
2513
2868
  - **The value typography of both pickers is now tokenised.** It was hardcoded: the date value was pinned to
2514
2869
  `--font-mono` and the time trigger to `--font-weight-medium`, so a brand could not bring either in line
@@ -2585,7 +2940,7 @@ workaround this release made unnecessary. Generated from the release manifest’
2585
2940
  component stylesheet. `none` disables an entrance.
2586
2941
  All defaults are unchanged. Leave keyframes are deliberately not swappable — modal/drawer/bottom-sheet
2587
2942
  teardown matches them by name — and reduced motion still wins: the library zeroes the duration itself,
2588
- so a custom entrance carrying its own timing is silenced too. New: `src/lib/styles/_surface-motion.scss`
2943
+ so a custom entrance carrying its own timing is silenced too. New: `styles/_surface-motion.scss`
2589
2944
  (exported from `styles/ds`; add it alongside `styles/overlay-motion` if you import partials à la carte).
2590
2945
 
2591
2946
  ### Changed
@@ -2650,8 +3005,9 @@ workaround this release made unnecessary. Generated from the release manifest’
2650
3005
  `--ds-time-picker-value-weight: var(--font-weight-medium)` to keep the old weight.
2651
3006
 
2652
3007
  - **`fullWidth` on `ds-date-picker` now pushes the calendar glyph to the trailing edge.** Stretching the
2653
- trigger left the value + glyph packed at the left (the glyph sat 122px short of the right edge on a 274px
2654
- field), so a full-width date field did not line up with the time picker or a `ds-select` beside it.
3008
+ trigger left the value + glyph packed at the left (the glyph sat 122px short of the right edge on a
3009
+ 274px field), so a full-width date field did not line up with the time picker or a `ds-select` beside
3010
+ it.
2655
3011
 
2656
3012
  - **Overlay panels now enter FROM their trigger edge instead of always sliding downward.** A panel below
2657
3013
  its trigger still slides down into place; one placed above now rises, and a side-placed panel comes in
@@ -2716,9 +3072,9 @@ workaround this release made unnecessary. Generated from the release manifest’
2716
3072
  and padding even with nothing projected into `[actions]`; it now hides when empty, the same guard the
2717
3073
  card already used for `.impact`.
2718
3074
 
2719
- - **`ds-docs`-style à la carte consumers were missing the overlay-panel entrance.** `styles/overlay-motion`
2720
- is now imported by the docs app itself, so portaled panels animate there as they do for consumers of the
2721
- bundled `styles/ds` entry.
3075
+ - **À la carte imports were missing the overlay-panel entrance.** If you import partials individually
3076
+ rather than the bundled `styles/ds` entry, add `styles/overlay-motion` — without it, portaled panels
3077
+ (select, popover, tooltip, menu submenus, and more) will not animate.
2722
3078
 
2723
3079
  - **Dialogs opened through `ModalService` no longer carry the `ds-overlay-pane` panel class.**
2724
3080
  `ModalService.open()` — and therefore `confirm()` — tagged the CDK overlay pane with `ds-overlay-pane`,
@@ -2729,62 +3085,16 @@ workaround this release made unnecessary. Generated from the release manifest’
2729
3085
  ### What you can now DELETE
2730
3086
 
2731
3087
  None of this is required — skipping it leaves working code that is merely redundant. Each row is a
2732
- workaround this release made unnecessary. Generated from the release manifest’s `deleteList`.
3088
+ workaround this release made unnecessary. Generated from the `deleteList` in `migrations/0.7.1.json`.
2733
3089
 
2734
3090
  | If you have | Delete it, because |
2735
3091
  | --- | --- |
2736
- | `[reserveMessage]="false"` | The default is FALSE again, so an opt-out binding only restates it - unless the field sits under a `dsReserveMessages` host or an app config that turns it on, where the binding is still the way back out. mm-broker carried six, three of them dead (2026-08-21). Replace with: nothing - delete the binding where no host reserves. |
3092
+ | `[reserveMessage]="false"` | The default is FALSE again, so an opt-out binding only restates it - unless the field sits under a `dsReserveMessages` host or an app config that turns it on, where the binding is still the way back out. Replace with: nothing - delete the binding where no host reserves. |
2737
3093
  | `gap: var(--space-1)`, `margin-top: -` | A tighter sub-stack or a negative margin added to swallow the reserved message row is now over-tightening a form that no longer reserves it. Replace with: the normal 16px field stack. |
2738
3094
  | `ds-pagination`, `ds-field-message`, `display: none` | A rule hiding the empty message row inside the pagination rows-per-page select. The component pins that select to its own compact shape now, so the row is gone at the source. Replace with: nothing - delete the rule. |
2739
3095
 
2740
3096
  ## 0.7.0 — 2026-07-30
2741
3097
 
2742
- ### Changed
2743
-
2744
- - **BREAKING: the control `size` scale is unified into one shared `DsSize` type.** The per-component
2745
- size types (`InputSize`, `ButtonSize`, `SelectSize`, `SearchSize`, `CheckboxSize`, `RadioSize`,
2746
- `SwitchSize`, `SpinnerSize`, `TextareaSize`, `NumberStepperSize`, `DatePickerSize`, `TimePickerSize`,
2747
- `DrawerSize`, `IconButtonSize`, `AutocompleteSize`) are removed from the public API — import
2748
- `DsSize` instead (`type DsSize = 'sm' | 'md' | 'lg'`, exported from the package root). `AvatarSize`,
2749
- `ModalSize` and `ContainerSize` remain and now extend it (`'xs' | DsSize`, `DsSize | 'xl' | 'fullscreen'`,
2750
- `DsSize | 'xl' | 'full'`). No runtime or visual change — every component's `size` input behaves exactly as before.
2751
-
2752
- - **BREAKING: form fields now reserve one line for the message row by default.** Every field
2753
- (`ds-input`, `ds-textarea`, `ds-select`, `ds-autocomplete`, and the other message-bearing controls)
2754
- gains a `reserveMessage` input (default `true`). With it on, the field always keeps a message line's
2755
- worth of vertical space even when there's no hint or error, so a form no longer reflows when
2756
- validation messages appear — the row is there from the start. Set `[reserveMessage]="false"` to
2757
- collapse the space back (the previous behaviour) where the layout can absorb the shift.
2758
-
2759
- - **BREAKING: `ds-pagination` drops its manual `(pageChange)` and `(pageSizeChange)` outputs.** They
2760
- duplicated the two-way models — the manual `pageSizeChange` even name-collided with the `pageSize`
2761
- model's own implicit output, and one size change fired the change events three times over. Page and
2762
- size are now driven purely by the `[(currentPage)]` / `[(pageSize)]` two-way models. Migrate
2763
- `(pageChange)="…"` to `(currentPageChange)="…"`; `(pageSizeChange)` keeps working unchanged (it is now
2764
- the `pageSize` model's implicit output, same name and payload). `[(currentPage)]` / `[(pageSize)]`
2765
- bindings are unaffected. (`ds-table` already drove pagination via the models, so it is untouched.)
2766
-
2767
- - **BREAKING: `ds-input` drops the `withAddon` / `withAddonRight` inputs.** They hand-declared what the
2768
- DOM already states — and the doc line was wrong (the flag stripped the control's inset rather than
2769
- "reserving space"). The flush-addon layout is now derived automatically: a boxed `.addon` projected
2770
- into `[dsInputPrefix]`/`[dsInputSuffix]` (`<span dsInputPrefix class="addon">…</span>`) sits flush via
2771
- `:has()`, while a bare projected `<ds-icon dsInputPrefix>` correctly keeps its inset. Delete
2772
- `[withAddon]` / `[withAddonRight]` from your templates — a boxed addon renders identically, and a bare
2773
- icon that previously (mis)set the flag regains its proper inset (a bugfix).
2774
-
2775
- - **BREAKING: `ds-app-shell` is renamed to `ds-app-layout`.** The slotted fixed-frame application layout
2776
- shipped as `ds-app-shell` in the tagged 0.6.3 source; this renames the selector (`ds-app-shell` →
2777
- `ds-app-layout`), the component class (`AppShellComponent` → `AppLayoutComponent`), the exported scroll
2778
- service (`ShellScrollService` → `LayoutScrollService`), and its four tokens (`--ds-app-shell-pad-x` /
2779
- `-pad-y` / `-content-max-width` / `-height` → `--ds-app-layout-*`). Behaviour is unchanged — rename the
2780
- references (see `migrations/`). A consumer still on 0.6.3's `ds-app-shell` must migrate.
2781
-
2782
- - **Retroactive note: `TableComponent` gained four more members (`total`, `sortingOrder`,
2783
- `animateHeight`, `scrollToTopOnPageChange`), and most other component classes changed too**, so a
2784
- consumer that extends a DS class (unsupported — `design-system.md` § Components) can collide on a
2785
- name here as well, exactly as in 0.6.3. Same resolution: rename the subclass's own member, or add
2786
- `override` where the base's semantics genuinely match.
2787
-
2788
3098
  ### Added
2789
3099
 
2790
3100
  - **New override tokens for two baked-in weights: `--ds-sidenav-item-label-weight` and
@@ -2797,9 +3107,9 @@ workaround this release made unnecessary. Generated from the release manifest’
2797
3107
  Off by default.
2798
3108
 
2799
3109
  - **`ds-search` gains a `size` input (`sm`/`md`/`lg`, default `md`).** Search previously ignored `size`
2800
- and its size CSS never matched, so it sat 4px taller than sibling controls in dense toolbars. `sm` is
2801
- now 32px (matching `ds-input`/`ds-button` at `sm`); the default `md` keeps the prior 36px search-row
2802
- height, so existing consumers are unaffected.
3110
+ and its size CSS never matched, so it sat 4px taller than sibling controls in dense toolbars.
3111
+ `sm` is now 32px (matching `ds-input`/`ds-button` at `sm`); the default `md` keeps the prior 36px
3112
+ search-row height, so existing consumers are unaffected.
2803
3113
 
2804
3114
  - **`ds-table` gains a `total` input for server-side pagination.** Set `[total]` to the full row count
2805
3115
  when the data you pass in is already the current page (fetched from the server); the table then treats
@@ -2840,14 +3150,59 @@ workaround this release made unnecessary. Generated from the release manifest’
2840
3150
  `ds-drawer` / `ds-bottom-sheet` share is injectable, so a consumer's own full-screen overlay can
2841
3151
  participate in the same lock instead of fighting it (`lock()` / `unlock()`, paired).
2842
3152
 
3153
+ ### Changed
3154
+
3155
+ - **BREAKING: the control `size` scale is unified into one shared `DsSize` type.** The per-component
3156
+ size types (`InputSize`, `ButtonSize`, `SelectSize`, `SearchSize`, `CheckboxSize`, `RadioSize`,
3157
+ `SwitchSize`, `SpinnerSize`, `TextareaSize`, `NumberStepperSize`, `DatePickerSize`, `TimePickerSize`,
3158
+ `DrawerSize`, `IconButtonSize`, `AutocompleteSize`) are removed from the public API — import
3159
+ `DsSize` instead (`type DsSize = 'sm' | 'md' | 'lg'`, exported from the package root). `AvatarSize`,
3160
+ `ModalSize` and `ContainerSize` remain and now extend it (`'xs' | DsSize`, `DsSize | 'xl' | 'fullscreen'`,
3161
+ `DsSize | 'xl' | 'full'`). No runtime or visual change — every component's `size` input behaves exactly as before.
3162
+
3163
+ - **BREAKING: form fields now reserve one line for the message row by default.** Every field
3164
+ (`ds-input`, `ds-textarea`, `ds-select`, `ds-autocomplete`, and the other message-bearing controls)
3165
+ gains a `reserveMessage` input (default `true`). With it on, the field always keeps a message line's
3166
+ worth of vertical space even when there's no hint or error, so a form no longer reflows when
3167
+ validation messages appear — the row is there from the start. Set `[reserveMessage]="false"` to
3168
+ collapse the space back (the previous behaviour) where the layout can absorb the shift.
3169
+
3170
+ - **BREAKING: `ds-pagination` drops its manual `(pageChange)` and `(pageSizeChange)` outputs.** They
3171
+ duplicated the two-way models — the manual `pageSizeChange` even name-collided with the `pageSize`
3172
+ model's own implicit output, and one size change fired the change events three times over. Page and
3173
+ size are now driven purely by the `[(currentPage)]` / `[(pageSize)]` two-way models. Migrate
3174
+ `(pageChange)="…"` to `(currentPageChange)="…"`; `(pageSizeChange)` keeps working unchanged (it is now
3175
+ the `pageSize` model's implicit output, same name and payload). `[(currentPage)]` / `[(pageSize)]`
3176
+ bindings are unaffected. (`ds-table` already drove pagination via the models, so it is untouched.)
3177
+
3178
+ - **BREAKING: `ds-input` drops the `withAddon` / `withAddonRight` inputs.** They hand-declared what the
3179
+ DOM already states. The flush-addon layout is now derived automatically: a boxed `.addon` projected
3180
+ into `[dsInputPrefix]`/`[dsInputSuffix]` (`<span dsInputPrefix class="addon">…</span>`) sits flush via
3181
+ `:has()`, while a bare projected `<ds-icon dsInputPrefix>` correctly keeps its inset. Delete
3182
+ `[withAddon]` / `[withAddonRight]` from your templates — a boxed addon renders identically, and a bare
3183
+ icon that previously (mis)set the flag regains its proper inset (a bugfix).
3184
+
3185
+ - **BREAKING: `ds-app-shell` is renamed to `ds-app-layout`.** The slotted fixed-frame application layout
3186
+ shipped as `ds-app-shell` in 0.6.3; this renames the selector (`ds-app-shell` →
3187
+ `ds-app-layout`), the component class (`AppShellComponent` → `AppLayoutComponent`), the exported scroll
3188
+ service (`ShellScrollService` → `LayoutScrollService`), and its four tokens (`--ds-app-shell-pad-x` /
3189
+ `-pad-y` / `-content-max-width` / `-height` → `--ds-app-layout-*`). Behaviour is unchanged — rename the
3190
+ references (see `migrations/`). A consumer still on 0.6.3's `ds-app-shell` must migrate.
3191
+
3192
+ - **`TableComponent` gained four more members (`total`, `sortingOrder`,
3193
+ `animateHeight`, `scrollToTopOnPageChange`), and most other component classes changed too**, so a
3194
+ consumer that extends a DS class (unsupported — `design-system.md` § Components) can collide on a
3195
+ name here as well, exactly as in 0.6.3. Same resolution: rename the subclass's own member, or add
3196
+ `override` where the base's semantics genuinely match.
3197
+
2843
3198
  ### Fixed
2844
3199
 
2845
3200
  - **`ds-menu`: disabled items are now keyboard-reachable.** A disabled `ds-menu-item` was dropped from the
2846
3201
  roving-tabindex nav (unreachable by keyboard) and rendered a native `disabled` button; it now uses
2847
3202
  `aria-disabled` and stays in the arrow / type-ahead order (ARIA menu practice), so a keyboard user can
2848
3203
  focus it — and a `[dsTooltip]` on it now shows on focus, surfacing a "why is this disabled" reason.
2849
- Activation stays blocked. (Reliable screen-reader announcement of that reason is a separate follow-up —
2850
- the tooltip's `aria-describedby` currently lands on the `role="none"` host, not the inner button.)
3204
+ Activation stays blocked. (That reason is not announced reliably yet: the tooltip's `aria-describedby` lands on the
3205
+ `role="none"` host, not the inner button.)
2851
3206
 
2852
3207
  - **Stacked overlays no longer unlock body scroll under each other.** `ds-modal`, `ds-drawer` and
2853
3208
  `ds-bottom-sheet` each wrote `body.style.overflow` directly, so closing the outer of two stacked
@@ -2864,8 +3219,8 @@ workaround this release made unnecessary. Generated from the release manifest’
2864
3219
 
2865
3220
  - **`ds-sidenav`: collapsed items no longer overflow by 1px.** A collapsed item was sized to the
2866
3221
  `--sidenav-collapsed-item` hit target, 1px wider than the rail's content box (border-box minus padding
2867
- and the right border), which produced a hairline horizontal scrollbar. Collapsed items now fill the
2868
- available width, and the rail clips any residual x-overflow.
3222
+ and the right border), which produced a hairline horizontal scrollbar. Collapsed items now fill
3223
+ the available width, and the rail clips any residual x-overflow.
2869
3224
 
2870
3225
  - **`dsTooltip`: keyboard focus now shows and hides the tip reliably, and panels no longer strand.** The
2871
3226
  directive listened on `focus`/`blur`, which don't fire for focus moving between descendants; it now uses
@@ -3014,11 +3369,11 @@ workaround this release made unnecessary. Generated from the release manifest’
3014
3369
  - **`ds-toast` now slides up from below into place as it enters** (previously it dropped in from
3015
3370
  slightly above). The exit animation is unchanged — the toast still slides up as it fades out.
3016
3371
  Visual/motion only; no API change.
3017
- - **Retroactive note: the new `ds-table` column API grew `TableComponent`'s member surface by ~76
3018
- names, which breaks a consumer that extends the class.** Everything above is additive for anyone
3019
- using the `<ds-table>` element. Subclassing a DS component is unsupported (`design-system.md`
3020
- § Components), but where it happens the new names collide — TS4114 demands an `override` modifier,
3021
- TS2416 fires where the signature differs. Collisions seen in the wild: `zebra`, `columnOrder`,
3372
+ - **The new `ds-table` column API grew `TableComponent`'s member surface by ~76 names, which breaks
3373
+ a consumer that extends the class.** Everything above is additive for anyone using the
3374
+ `<ds-table>` element. Subclassing a DS component is unsupported (`design-system.md` §
3375
+ Components), but where it happens the new names collide — TS4114 demands an `override` modifier,
3376
+ TS2416 fires where the signature differs. Collisions include: `zebra`, `columnOrder`,
3022
3377
  `orderedColumns`, `visibleColumns`, `toggleColumn`, `onColumnDrop`, `moveColumn`, `resetColumns`,
3023
3378
  `columnFilters`, `exportCsv`, `onResizeMove`, `onResizeEnd`. Rename the subclass's own member, or
3024
3379
  add `override` where the base's semantics genuinely match (`zebra` must then adopt the base's
@@ -3122,7 +3477,7 @@ None of this is required — skipping it leaves working code that is merely redu
3122
3477
  ### What you can now DELETE
3123
3478
 
3124
3479
  None of this is required — skipping it leaves working code that is merely redundant. Each row is a
3125
- workaround this release made unnecessary. Generated from the release manifest’s `deleteList`.
3480
+ workaround this release made unnecessary. Generated from the `deleteList` in `migrations/0.6.1.json`.
3126
3481
 
3127
3482
  | If you have | Delete it, because |
3128
3483
  | --- | --- |
@@ -3328,13 +3683,13 @@ workaround this release made unnecessary. Generated from the release manifest’
3328
3683
  ### Breaking
3329
3684
 
3330
3685
  - **`ds-combobox` renamed to `ds-autocomplete`.** The clearer industry name for the editable
3331
- type-to-pick control (MUI's term); "combobox" survives only as the ARIA role — which the
3332
- select-only `ds-select` trigger also carries. Renames: selector `ds-combobox` →
3333
- `ds-autocomplete`; `ComboboxComponent` → `AutocompleteComponent`; exported types
3334
- `ComboboxOption` / `ComboboxOptionGroup` / `ComboboxSize` / `ComboboxWidth` →
3335
- `AutocompleteOption` / `AutocompleteOptionGroup` / `AutocompleteSize` / `AutocompleteWidth`;
3336
- public tokens `--ds-combobox-focus-shadow` / `--ds-combobox-maxw` / `--ds-combobox-input-minw`
3337
- → `--ds-autocomplete-*`; docs anchor `#combobox` → `#autocomplete`. Behavior unchanged.
3686
+ type-to-pick control; "combobox" survives only as the ARIA role — which the select-only
3687
+ `ds-select` trigger also carries. Renames: selector `ds-combobox` → `ds-autocomplete`;
3688
+ `ComboboxComponent` → `AutocompleteComponent`; exported types `ComboboxOption` /
3689
+ `ComboboxOptionGroup` / `ComboboxSize` / `ComboboxWidth` → `AutocompleteOption` /
3690
+ `AutocompleteOptionGroup` / `AutocompleteSize` / `AutocompleteWidth`; public tokens
3691
+ `--ds-combobox-focus-shadow` / `--ds-combobox-maxw` / `--ds-combobox-input-minw` →
3692
+ `--ds-autocomplete-*`. Behavior unchanged.
3338
3693
  - **`ds-dropdown` removed — use `ds-select`.** The two were near-identical select-only single
3339
3694
  pickers (same trigger, same panel, same keyboard model); `ds-select` is the richer canonical
3340
3695
  form field (`ControlValueAccessor`, `invalid` + `hint` via `ds-field-message`, `required` /
@@ -3342,8 +3697,7 @@ workaround this release made unnecessary. Generated from the release manifest’
3342
3697
  `<ds-select>`; `[error]` → `[invalid]`; `[optionGroups]="groups"` → pass groups inline in
3343
3698
  `[options]` (`SelectItem = SelectOption | SelectOptionGroup`); `DropdownOption` /
3344
3699
  `DropdownOptionGroup` / `DropdownSize` → `SelectOption` / `SelectOptionGroup` / `SelectSize`.
3345
- `readOnly` moved to `ds-select` (see Added). Docs "Dropdown" section removed — the Select
3346
- section covers everything.
3700
+ `readOnly` moved to `ds-select` (see Added).
3347
3701
  - **Public token prefixes normalized to match the selector (not aliased).** Four components had
3348
3702
  their `--ds-*` token name fragments hyphenated to align with the Angular selector; the old
3349
3703
  concatenated names silently stop resolving, so consumers who override any of these must rename.
@@ -3453,7 +3807,7 @@ workaround this release made unnecessary. Generated from the release manifest’
3453
3807
 
3454
3808
  - **`ds-search` inner input class renamed `.input` → `.search-input`.** The over-generic name let
3455
3809
  consumer global `.input` styles/resets leak into the component through emulated encapsulation
3456
- (root cause of the PIN-475 double-focus-ring). Only consumer styling that targeted the
3810
+ (root cause of a duplicate focus ring). Only consumer styling that targeted the
3457
3811
  undocumented internal class is affected.
3458
3812
  - **`ds-date-picker` trigger is now a text input by default.** Manual masked entry is on
3459
3813
  (`allowInput` defaults to `true`), so the trigger renders as an editable field with a trailing
@@ -3492,6 +3846,32 @@ workaround this release made unnecessary. Generated from the release manifest’
3492
3846
 
3493
3847
  ## 0.4.0 — 2026-06-30
3494
3848
 
3849
+ ### Breaking
3850
+
3851
+ - **`ds-button` now takes two orthogonal axes — `color` × `variant` — replacing the single
3852
+ conflated `variant` input.** `color` is the hue family (`primary` | `secondary` | `tertiary` |
3853
+ `neutral` | `error` | `success` | `warning`, default `primary`); `variant` is the fill style
3854
+ (`solid` | `tonal` | `ghost` | `outline` | `text`, default `solid`). The old single axis mixed hue
3855
+ and emphasis — migrate as: `primary` → default (drop the attr); `secondary` →
3856
+ `variant="tonal"`; `tertiary` → `variant="ghost"`; `outline` → `color="neutral"
3857
+ variant="outline"`; `error` / `success` / `warning` → `color="error|success|warning"`; `text` →
3858
+ `variant="text"`. New exported `ButtonColor` / `ButtonVariant` types. `text` still renders but is
3859
+ **discouraged** (dev-mode `console.warn` — prefer the global `.link` class); a small set of
3860
+ low-contrast `color × variant` combos (e.g. `neutral` solid, `success` / `warning` ghost/outline)
3861
+ render with a discouraged ⚠ marker rather than being banned. Each instance now sources its accent
3862
+ from the `--accent-*` pointer layer (see ds-color).
3863
+ - **Reactive-forms radio now uses `<ds-radio-group>` as the group container and form control.**
3864
+ Standalone `<ds-radio [(checked)]>` continues to work for non-group usage, but a group of
3865
+ radios that participates in a reactive form should now be wrapped in `<ds-radio-group>`.
3866
+ - **`ButtonSize` collapsed to `sm | md | lg`** (32 / 40 / 48 px), dropping the duplicate
3867
+ legacy `s | m | l` tier. Default is now `md` (was `l` — same 40 px). Migration:
3868
+ `m` → `sm` (both 32), `l` → `md` (both 40); the 24 px `s` is removed — its callers move to
3869
+ `sm` (32 px). Aligns button with input / select / date-picker / icon-button.
3870
+ - **Removed `--ds-input-hint-color` / `--ds-input-hint-fs`.** Field helper/error text is now
3871
+ rendered by the shared `ds-field-message`; tune its neutral colour and size via the new
3872
+ `--ds-field-message-color` / `--ds-field-message-font-size` hooks (these apply to every field's
3873
+ helper/error line, not just input).
3874
+
3495
3875
  ### Added
3496
3876
 
3497
3877
  #### New components
@@ -3530,19 +3910,17 @@ workaround this release made unnecessary. Generated from the release manifest’
3530
3910
  each item has `title` / `subtitle` text, `[leading]` / `[trailing]` slots, and optional
3531
3911
  `clickable` (emits `activated`) with `selected` / `disabled` states.
3532
3912
  - **`ds-timeline` / `ds-timeline-item`** — vertical activity feed; each item has a
3533
- `heading`, `time`, an optional node `icon`, and a `tone`
3534
- (`default` | `primary` | `success` | `warning` | `error` | `muted`). Exports
3535
- `TimelineTone`.
3913
+ `heading`, `time`, an optional node `icon`, and a `tone` (`default` | `primary` | `success` |
3914
+ `warning` | `error` | `muted`). Exports `TimelineTone`.
3536
3915
  - **`ds-description-list` / `ds-description-item`** — term/definition pairs with
3537
3916
  `horizontal` | `stacked` `orientation`; each item takes a `term` and optional `copyValue`
3538
3917
  (renders a copy-to-clipboard button). Exports `DescriptionListOrientation`.
3539
3918
  - **`ds-stack` / `ds-inline` / `ds-container` layout primitives.** `ds-stack` (column) and
3540
3919
  `ds-inline` (row, wraps by default) take a named `gap` (`none … xl`) + `align` /
3541
- `justify`; `ds-container` centres content at a named `size`
3542
- (`sm` | `md` | `lg` | `xl` | `full`). Adds semantic `--space-stack` / `--space-inline`
3543
- spacing aliases plus per-primitive `--ds-stack-gap` / `--ds-inline-gap` /
3544
- `--ds-container-max` / `--ds-container-pad-x` hooks. Exports `LayoutGap`, `LayoutAlign`,
3545
- `LayoutJustify`, `ContainerSize`.
3920
+ `justify`; `ds-container` centres content at a named `size` (`sm` | `md` | `lg` | `xl` | `full`).
3921
+ Adds semantic `--space-stack` / `--space-inline` spacing aliases plus per-primitive
3922
+ `--ds-stack-gap` / `--ds-inline-gap` / `--ds-container-max` / `--ds-container-pad-x` hooks.
3923
+ Exports `LayoutGap`, `LayoutAlign`, `LayoutJustify`, `ContainerSize`.
3546
3924
  - **`ds-loading-overlay`** — region scrim with a centred spinner + optional `label`,
3547
3925
  toggled by `loading`, with `blur`, `spinnerSize` and an `insetTop` offset (clears a
3548
3926
  sticky header). `ds-table` and `ds-empty-state` now render their loading state through it.
@@ -3620,8 +3998,7 @@ workaround this release made unnecessary. Generated from the release manifest’
3620
3998
  - **`ds-date-picker` month & year grid pickers** in the calendar header — click the
3621
3999
  month/year labels to pick from a compact grid (years in a paged 12-cell grid) instead of
3622
4000
  stepping month-by-month.
3623
- - **`ds-table` pagination, client-side sort & sticky first column** (absorbed from the
3624
- prototype data-grid): new `pageSize` / `page` / `pageSizeOptions`, `clientSort`,
4001
+ - **`ds-table` pagination, client-side sort & sticky first column**: new `pageSize` / `page` / `pageSizeOptions`, `clientSort`,
3625
4002
  `stickyFirstColumn`, plus `cellTemplateMap` / `emptyTemplate` for wrapper components.
3626
4003
  - **`ModalService.confirm(options): Promise<boolean>`** — promise-based confirmation dialog
3627
4004
  rendered as a modal card (`ModalConfirmOptions`: `title`, `message`, labels, `tone`, `icon`,
@@ -3637,9 +4014,9 @@ workaround this release made unnecessary. Generated from the release manifest’
3637
4014
  CSS custom properties can't be used in `@media` conditions, so the `--bp-*` scale was
3638
4015
  unusable there. A new partial is now the single source for both representations: it emits
3639
4016
  `--bp-*` (values unchanged) and exposes compile-time `$bp-xs … $bp-2xl` plus the mixins
3640
- `bp.up(md)` (`width >= 768px`) / `bp.down(sm)` (`width < 640px`), via
3641
- `@use 'breakpoints' as bp;`. Heads-up: advanced consumers importing `styles/tokens` à la
3642
- carte now also need `styles/breakpoints` for `--bp-*`; the `styles/ds` entry forwards both.
4017
+ `bp.up(md)` (`width >= 768px`) / `bp.down(sm)` (`width < 640px`), via `@use 'breakpoints' as bp;`.
4018
+ Heads-up: advanced consumers importing `styles/tokens` à la carte now also need
4019
+ `styles/breakpoints` for `--bp-*`; the `styles/ds` entry forwards both.
3643
4020
  - **`--ds-button-border-disabled`** and **`--ds-icon-button-bg-color-disabled` /
3644
4021
  `-icon-color-disabled` / `-border-disabled`** override tokens — the disabled
3645
4022
  fill/label/border are now individually themeable (previously only button had
@@ -3647,34 +4024,6 @@ workaround this release made unnecessary. Generated from the release manifest’
3647
4024
  - **`--ds-button-press-scale`** (default `0.98`) and **`--ds-icon-button-press-scale`**
3648
4025
  (default `0.96`) — tune the press scale-down (set `1` to disable).
3649
4026
 
3650
- ### Breaking
3651
-
3652
- - **`ds-button` now takes two orthogonal axes — `color` × `variant` — replacing the single
3653
- conflated `variant` input.** `color` is the hue family
3654
- (`primary` | `secondary` | `tertiary` | `neutral` | `error` | `success` | `warning`,
3655
- default `primary`); `variant` is the fill style
3656
- (`solid` | `tonal` | `ghost` | `outline` | `text`, default `solid`). The old single axis
3657
- mixed hue and emphasis — migrate as: `primary` → default (drop the attr); `secondary` →
3658
- `variant="tonal"`; `tertiary` → `variant="ghost"`; `outline` →
3659
- `color="neutral" variant="outline"`; `error` / `success` / `warning` →
3660
- `color="error|success|warning"`; `text` → `variant="text"`. New exported `ButtonColor` /
3661
- `ButtonVariant` types. `text` still renders but is **discouraged** (dev-mode `console.warn`
3662
- — prefer the global `.link` class); a small set of low-contrast `color × variant` combos
3663
- (e.g. `neutral` solid, `success` / `warning` ghost/outline) render with a discouraged ⚠
3664
- marker rather than being banned. Each instance flips the internal `--c-*` accent slots,
3665
- which now default to the `--accent-*` pointer layer (see ds-color).
3666
- - **Reactive-forms radio now uses `<ds-radio-group>` as the group container and form control.**
3667
- Standalone `<ds-radio [(checked)]>` continues to work for non-group usage, but a group of
3668
- radios that participates in a reactive form should now be wrapped in `<ds-radio-group>`.
3669
- - **`ButtonSize` collapsed to `sm | md | lg`** (32 / 40 / 48 px), dropping the duplicate
3670
- legacy `s | m | l` tier. Default is now `md` (was `l` — same 40 px). Migration:
3671
- `m` → `sm` (both 32), `l` → `md` (both 40); the 24 px `s` is removed — its callers move to
3672
- `sm` (32 px). Aligns button with input / select / date-picker / icon-button.
3673
- - **Removed `--ds-input-hint-color` / `--ds-input-hint-fs`.** Field helper/error text is now
3674
- rendered by the shared `ds-field-message`; tune its neutral colour and size via the new
3675
- `--ds-field-message-color` / `--ds-field-message-font-size` hooks (these apply to every field's
3676
- helper/error line, not just input).
3677
-
3678
4027
  ### Changed
3679
4028
 
3680
4029
  - **Button & icon-button press is now darken + a micro scale-down; the pressed halo is off by default.**
@@ -3747,12 +4096,11 @@ workaround this release made unnecessary. Generated from the release manifest’
3747
4096
 
3748
4097
  ### Added
3749
4098
 
3750
- - **Reactive-forms support on `ds-input`, `ds-textarea`, `ds-select`.** They now
3751
- implement `ControlValueAccessor`, so `formControlName` / `[formControl]` and template
3752
- `[(ngModel)]` bind to them out of the box, and `FormControl.disable()` is honoured. The
3753
- existing `[(value)]` signal API is unchanged — both work. Shared via an internal
3754
- `DsControlValueAccessor` base. Checkbox/switch/combobox/date-picker/radio are not yet
3755
- wired (tracked).
4099
+ - **Reactive-forms support on `ds-input`, `ds-textarea`, `ds-select`.** They now implement
4100
+ `ControlValueAccessor`, so `formControlName` / `[formControl]` and template `[(ngModel)]` bind to
4101
+ them out of the box, and `FormControl.disable()` is honoured. The existing `[(value)]` signal API
4102
+ is unchanged — both work. Checkbox / switch / combobox / date-picker / radio do not implement
4103
+ `ControlValueAccessor` yet.
3756
4104
  - **`ng add` support.** `ng add @ids-group-ltd/ids-design-system` wires the design
3757
4105
  system's global styles (`@use '@ids-group-ltd/ids-design-system/styles/ds';`) into the
3758
4106
  target project's global styles entry, idempotently. It targets the default/first
@@ -3793,7 +4141,7 @@ workaround this release made unnecessary. Generated from the release manifest’
3793
4141
  hard ring (`--focus-ring`), matching `ds-input` / `ds-select` / `ds-search`. Minor visual
3794
4142
  change to the combobox focus state.
3795
4143
  - Backgrounds and several control transitions switched from the `background` shorthand to
3796
- `background-color` for precision. Alongside it, a repo-wide, value-preserving CSS-syntax
4144
+ `background-color` for precision. Alongside it, a library-wide, value-preserving CSS-syntax
3797
4145
  modernization: legacy `hsl()/hsla()` comma syntax → modern space/slash form, `clip:
3798
4146
  rect(…)` → `clip-path: inset(50%)` on visually-hidden helpers, and `min-/max-width`
3799
4147
  media queries → range syntax (`width >= …`). No intended visual change.
@@ -3816,16 +4164,15 @@ rect(…)` → `clip-path: inset(50%)` on visually-hidden helpers, and `min-/max
3816
4164
 
3817
4165
  ### Changed
3818
4166
 
3819
- - Removed the repository link from the npm package: dropped the `repository` field
3820
- from `package.json` and moved the maintainer Development/Releasing docs out of the
3821
- published README into a repo-only `CONTRIBUTING.md`. The README License entry is
3822
- now plain text (no repo link), fixing the npm 404.
4167
+ - Removed the repository link from the npm package: dropped the `repository` field from
4168
+ `package.json`. The published README no longer includes its maintainer-only sections, and its
4169
+ License entry is now plain text (no repo link) — fixing the npm 404.
3823
4170
 
3824
4171
  ## 0.2.1 — 2026-06-16
3825
4172
 
3826
4173
  ### Added
3827
4174
 
3828
- - A `homepage` field linking the live docs/showcase, so the npm package page
4175
+ - A `homepage` field linking the live docs, so the npm package page
3829
4176
  points to it.
3830
4177
 
3831
4178
  ## 0.2.0 — 2026-06-16