@ids-group-ltd/ids-design-system 0.10.3 → 0.11.0

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.
@@ -6,7 +6,10 @@
6
6
  "kind": "removed-api",
7
7
  "old": "--opacity-muted",
8
8
  "new": null,
9
- "note": "The --opacity-muted custom property (0.6) is deleted from the token scale. No component has read it since 0.10.0 moved the disabled states onto --text-disabled + --opacity-disabled, so REMOVING IT CHANGES NOTHING THE LIBRARY RENDERS. Two consumer cases. (1) You SET it - most likely on ds-button or ds-icon-button, the override 0.10.0's DELETE-LIST already warned about: delete the declaration, it was a no-op then and is a no-op now (those two resolve their disabled state from oklch(from accent ...) colours, and --ds-button-label-color-disabled / --ds-icon-button-icon-color-disabled are the real hooks). (2) You READ it in your own CSS, i.e. var(--opacity-muted) appears in a value you wrote: that now resolves to nothing and the declaration is dropped, so pick the token that names the job instead - --opacity-disabled (0.5) where opacity is the ONLY disabled signal, --opacity-text-disabled (0.38) for disabled text - or inline 0.6 if what you actually wanted was de-emphasis rather than a disabled state. Do not re-declare --opacity-muted yourself to keep it alive: 'muted' is a de-emphasis level and disabled is a state, and sharing one number between them is what put this token on the DELETE-LIST in the first place.",
9
+ "action": "If you SET it, delete the declaration — it was a no-op then and is a no-op now; the real hooks are `--ds-button-label-color-disabled` and `--ds-icon-button-icon-color-disabled`. If you READ it, pick the token that names the job: `--opacity-disabled` (0.5) where opacity is the only disabled signal, `--opacity-text-disabled` (0.38) for disabled text, or an inline 0.6 if what you wanted was de-emphasis rather than a disabled state.",
10
+ "appliesWhen": "Nothing the library renders changes — no component has read it since 0.10.0 moved the disabled states onto `--text-disabled` plus `--opacity-disabled`.",
11
+ "pitfall": "Do NOT re-declare `--opacity-muted` yourself to keep it alive. \"Muted\" is a de-emphasis level and disabled is a state, and sharing one number between them is what put this token on the delete list in the first place. A read of it now resolves to nothing and the whole declaration is dropped, which is quieter than an error.",
12
+ "note": "The 0.6 custom property is deleted from the token scale.",
10
13
  "searchHints": [
11
14
  "--opacity-muted",
12
15
  "var(--opacity-muted",
@@ -20,7 +23,9 @@
20
23
  "kind": "behavior",
21
24
  "old": "ds-list-item ::ng-deep .activator",
22
25
  "new": null,
23
- "note": "ds-list-item's DOM changed: `[leading]` and `[trailing]` are now siblings of `.activator` rather than children of it, so a row can hold a checkbox or a kebab alongside a row-level action. NO INPUT OR OUTPUT CHANGED -- title, subtitle, clickable, selected, disabled and activated all behave as before, and a plain row looks identical. WHAT MOVED: the row's `display: flex`, its `--ds-list-item-pad-*` padding, its `--ds-list-item-gap`, the hover tint and the selected tint all sit on the HOST now; `.activator` is just the text button and has no padding of its own. ACTION: a `::ng-deep .activator` override that set padding, gap, background or hover now applies to the text only -- move it to `ds-list-item` itself. Also fixed alongside: `selected` set `aria-current` on the inner button, so it was inert whenever `clickable` was false; it is on the host now, which means a non-clickable row marked selected is finally announced as current.",
26
+ "action": "A `::ng-deep .activator` override that set padding, gap, background or hover now applies to the TEXT only — move it to `ds-list-item` itself.",
27
+ "appliesWhen": "No input or output changed: title, subtitle, clickable, selected, disabled and activated all behave as before, and a plain row looks identical.",
28
+ "note": "`[leading]` and `[trailing]` are siblings of `.activator` now rather than children, so a row can hold a checkbox or a kebab alongside a row-level action. The row's `display: flex`, its padding, its gap and both tints sit on the HOST; `.activator` is just the text button. Fixed alongside: `selected` set `aria-current` on the inner button, so it was inert whenever `clickable` was false — it is on the host now.",
24
29
  "searchHints": [
25
30
  "ds-list-item",
26
31
  ".activator",
@@ -44,7 +49,10 @@
44
49
  "kind": "behavior",
45
50
  "old": "ds-button[variant=\"ghost\"][disabled], ds-button[variant=\"text\"][disabled], ds-button[variant=\"outline\"][disabled]",
46
51
  "new": null,
47
- "note": "The DEFAULT disabled label colour changed for the three variants whose fill cannot carry the state. Was the shared `oklch(from <accent> 70% 0.035 h)`; is now `color-mix(in srgb, <label> 40%, var(--surface-default))` -- the lightness now moves toward the surface instead of being pinned at 70%. WHY: pinned at 70% the disabled label was two points of OKLCH lightness from the enabled one in the DARK theme (--primary-strong is L 67.9% there), a 1.11:1 gap, so a disabled ghost was indistinguishable from an active one. A fixed lightness cannot serve both themes -- in light disabled is lighter than the accent, in dark the accent is already light so disabled must go darker; only the surface knows that direction. Measured after: light 2.54 -> 3.41:1, dark 1.11 -> 2.91:1. `variant=\"outline\"` is included even though it has a border, because its DISABLED border resolves to only 1.30:1 against the page -- lower than its own enabled border at 1.49:1 -- so the label was carrying the state there too. NO ACTION NEEDED unless you relied on the exact old colour: `--ds-button-label-color-disabled` is the same hook and still wins, so an app that already sets it sees no change. If you matched the old #949EB5 by hand anywhere (a mock, a screenshot test, a hand-written CSS twin of the disabled state), for the default primary it is now #A5B4F8 in light and #3D4979 in dark. The label stays tinted rather than going grey -- chroma rises with the mix (0.036 -> 0.099 in light) -- which was chosen deliberately. LIMITATION: the mix target is --surface-default, a stand-in for the real backdrop, so a borderless button sitting on a tinted ground gets a label mixed against the page colour instead. `variant=\"solid\"` and `variant=\"tonal\"` are NOT affected and keep the shared recipe.",
52
+ "action": "None unless you relied on the exact old colour — `--ds-button-label-color-disabled` is the same hook and still wins. If you matched the old `#949EB5` by hand in a mock, a screenshot test or a hand-written twin of the disabled state, the default primary is now `#A5B4F8` in light and `#3D4979` in dark.",
53
+ "appliesWhen": "`variant=\"solid\"` and `variant=\"tonal\"` are NOT affected and keep the shared recipe.",
54
+ "pitfall": "The mix target is `--surface-default`, a stand-in for the real backdrop — so a borderless button sitting on a TINTED ground gets a label mixed against the page colour instead of the ground it is actually on. The label also stays tinted rather than going grey (chroma rises with the mix), which was chosen deliberately.",
55
+ "note": "The disabled label for ghost, text and outline moved from a pinned `oklch(from <accent> 70% …)` to `color-mix(in srgb, <label> 40%, var(--surface-default))`, so lightness moves toward the surface. Pinned at 70% the dark theme measured 1.11:1 — a disabled ghost was indistinguishable from an active one. After: light 2.54 → 3.41:1, dark 1.11 → 2.91:1. Outline is included because its disabled border resolves to only 1.30:1.",
48
56
  "searchHints": [
49
57
  "ds-button",
50
58
  "variant=\"ghost\"",
@@ -65,7 +73,9 @@
65
73
  "kind": "behavior",
66
74
  "old": "ds-icon-button[variant=\"standard\"][disabled]",
67
75
  "new": null,
68
- "note": "The DEFAULT disabled glyph colour changed for variant=\"standard\" ONLY -- the one variant with no fill and no border, so the glyph colour is the whole signal. Was the shared `oklch(from <accent> 70% 0.035 h)`; is now `color-mix(in srgb, <icon colour> 40%, var(--surface-default))`. WHY: pinned at L 70% the disabled glyph sat only ten points of OKLCH lightness from the enabled one in the DARK theme (the enabled glyph is --text-secondary at L 80% there), a 1.46:1 gap. Light was already fine at 4.17:1 because the enabled glyph is L 35% there -- this is a dark-theme fix. Measured after: light 4.17 -> 5.19:1, dark 1.46 -> 3.58:1. NO ACTION NEEDED unless you relied on the exact old colour: `--ds-icon-button-icon-color-disabled` is the same hook and still wins. If you matched the old value by hand, for the default it was #969EB5 (light) / #949EB5 (dark) and is now #AEB1B8 (light) / #575D69 (dark). variant=\"filled\", \"tonal\" and \"outlined\" are NOT affected and keep the shared recipe -- filled deliberately so, because its glyph is --c-on (white) and mixing white into a white surface would not move it.",
76
+ "action": "None unless you relied on the exact old colour — `--ds-icon-button-icon-color-disabled` is the same hook and still wins. Hand-matched values move from `#969EB5` / `#949EB5` to `#AEB1B8` (light) / `#575D69` (dark).",
77
+ "appliesWhen": "`variant=\"standard\"` only — the one variant with no fill and no border, so the glyph colour is the whole signal. Filled, tonal and outlined keep the shared recipe; filled deliberately, because its glyph is white and mixing white into a white surface would not move it.",
78
+ "note": "A dark-theme fix: pinned at L 70% the disabled glyph sat ten points of OKLCH lightness from the enabled one, a 1.46:1 gap. After: light 4.17 → 5.19:1, dark 1.46 → 3.58:1.",
69
79
  "searchHints": [
70
80
  "ds-icon-button",
71
81
  "variant=\"standard\"",
@@ -84,7 +94,10 @@
84
94
  "kind": "behavior",
85
95
  "old": "ds-input [success]=\"true\"",
86
96
  "new": null,
87
- "note": "TWO changes in one, and only the first needs edits. (1) BREAKING on `ds-input`: `success` was `input(false, { transform: booleanAttribute })` and is now `input<string>('')`. A `[success]=\"true\"` binding now sets the literal text \"true\" as the message. Replace `[success]=\"true\" hint=\"Available\"` with `success=\"Available\"` -- the text moves out of `hint` and into `success`. A `[success]=\"someBoolean\"` binding must become `[success]=\"someBoolean ? 'Looks good' : ''\"`, since an empty string is now what means \"no success state\". (2) ADDITIVE: the same `success` input now also exists on ds-textarea, ds-select, ds-multi-select, ds-autocomplete, ds-date-picker, ds-time-picker, ds-otp-input and ds-color-picker, which had no success state at all. WHY THE TYPE CHANGED: `error` and `warning` already carried their own text, so `success` was the only state that could not say anything -- it was a flag that recoloured `hint` green, which meant the message had to be smuggled through a different input. It is now the fourth rung of the same ladder: error, errors, warning, success, hint. CONSEQUENCE: a success border with NO message is no longer expressible, exactly as it never was for `warning`. NOT added to ds-checkbox, ds-radio-group, ds-switch or ds-slider -- they have no bordered box, so the state would appear in the message line alone. `invalid` is unchanged and remains the flag partner of `error`. NOTE ON KIND: this is not a value remap even though it looks like one — there is no old value that maps to a new one, because the replacement TEXT did not exist before and has to be authored (usually by moving it out of `hint`).",
97
+ "action": "Replace `[success]=\"true\" hint=\"Available\"` with `success=\"Available\"` — the text moves out of `hint`. A `[success]=\"someBoolean\"` becomes `[success]=\"someBoolean ? 'Looks good' : ''\"`, since an empty string is what means \"no success state\" now.",
98
+ "appliesWhen": "Breaking on ds-input only; the same input is ADDITIVE on ds-textarea, ds-select, ds-multi-select, ds-autocomplete, ds-date-picker, ds-time-picker, ds-otp-input and ds-color-picker, which had no success state at all. Not added to ds-checkbox, ds-radio-group, ds-switch or ds-slider — no bordered box to paint. `invalid` is unchanged and remains the flag partner of `error`.",
99
+ "pitfall": "A leftover `[success]=\"true\"` does not error — it sets the literal text \"true\" as the message, so the field renders a green line reading `true`. And a success border with NO message is no longer expressible, exactly as it never was for `warning`.",
100
+ "note": "`success` went from `input(false, { transform: booleanAttribute })` to `input<string>('')`. It was the only state that could not say anything — a flag that recoloured `hint` green, which meant the message had to be smuggled through a different input. It is the fourth rung of the same ladder now: error, errors, warning, success, hint.",
88
101
  "searchHints": [
89
102
  "success",
90
103
  "[success]",
@@ -112,7 +125,10 @@
112
125
  "kind": "behavior",
113
126
  "old": "class MyField extends TextareaComponent (or any DS field)",
114
127
  "new": null,
115
- "note": "INTERNAL REFACTOR, no public API change -- listed only because it matters to SUBCLASSERS. The seven inputs every field shares (`hint`, `error`, `warning`, `errors`, `invalid`, `touched`, `reserveMessage`) plus the `reserveMessageResolved` helper are now declared once on a new `DsFieldBase` instead of being hand-copied into each of the fourteen fields. Every input keeps its name, type, default and documentation; the generated docs API reference was diffed and lost nothing. `DsControlValueAccessor` and `DsBooleanControlValueAccessor` now extend `DsFieldBase`, and `ds-radio` extends it directly. ACTION only if you SUBCLASS a DS field and declare a member with one of those seven names: it is now an inherited member rather than an own member, so an incompatible re-declaration fails to compile where it may previously have shadowed. Redeclaring with the same shape still works. `success`, `label`, `required` and `optional` were deliberately NOT lifted and remain per-component. SEPARATELY, a genuine new collision: `success` is a NEW member on the eight bordered fields that lacked it (ds-autocomplete, ds-color-picker, ds-date-picker, ds-multi-select, ds-otp-input, ds-select, ds-textarea, ds-time-picker), so a subclass of any of those declaring `success` now fails with TS4114. Beware the report itself here: `npm run class-surface` lists those seven moved names as LOST from fourteen field components, which is an artefact of counting declarations per FILE — they moved UP into DsFieldBase and every field still has them by inheritance. AUTHORING already flags a member that moved to a base class as something the tool cannot see.",
128
+ "action": "Only if you SUBCLASS a DS field and declare a member named `hint`, `error`, `warning`, `errors`, `invalid`, `touched` or `reserveMessage`: it is an inherited member now rather than an own member, so an incompatible re-declaration fails where it may previously have shadowed. Redeclaring with the same shape still works.",
129
+ "appliesWhen": "No public API changed — every input keeps its name, type, default and documentation, and the generated docs reference lost nothing. `success`, `label`, `required` and `optional` were deliberately NOT lifted and remain per-component.",
130
+ "pitfall": "Beware the REPORT here, not just the code. `npm run class-surface` lists those seven names as LOST from fourteen field components, which is an artefact of counting declarations per FILE — they moved UP into DsFieldBase and every field still has them by inheritance. AUTHORING already flags a member that moved to a base class as something the tool cannot see. The genuine new collision is different: `success` is a NEW member on the eight bordered fields that lacked it, so a subclass of any of those declaring `success` fails with TS4114.",
131
+ "note": "The seven shared inputs and the `reserveMessageResolved` helper are declared once on a new `DsFieldBase` instead of being hand-copied into each of the fourteen fields. Both CVA bases extend it, and ds-radio extends it directly.",
116
132
  "searchHints": [
117
133
  "extends TextareaComponent",
118
134
  "extends InputComponent",
@@ -135,7 +151,9 @@
135
151
  "kind": "behavior",
136
152
  "old": "[invalid]=\"true\" hint=\"<message>\"",
137
153
  "new": null,
138
- "note": "BREAKING. `invalid` used to promote `hint` to the ERROR state -- a red message rendered through role=\"alert\". It now paints the control and sets aria-invalid only, and `hint` stays a neutral grey hint. ACTION: wherever a field binds `[invalid]` together with `hint` and expected the hint to read as the error, move the text to `error` and drop the `[invalid]` binding -- `error` implies the invalid state. So `[invalid]=\"true\" hint=\"Must be before end date\"` becomes `error=\"Must be before end date\"`. A dynamic form is `[error]=\"showError() ? 'Must be before end date' : ''\"`. WHY: the idiom lost the real hint (with both set, the help text was replaced rather than shown alongside), announced help text as an alert, and made a template unreadable because whether `hint` was a hint or an error depended on another attribute. It was the last input borrowing another's text. NO ACTION when `[invalid]` is used WITHOUT `hint` -- that is still supported and is exactly what the input is for: the error state with the message living elsewhere (a form-level error summary, one message for a whole ds-field-group), plus aria-invalid. `invalid` is not deprecated.",
154
+ "action": "Wherever a field binds `[invalid]` together with `hint` and expected the hint to read as the error, move the text to `error` and drop the `[invalid]` binding — `error` implies the invalid state. `[invalid]=\"true\" hint=\"Must be before end date\"` becomes `error=\"Must be before end date\"`, and a dynamic form becomes `[error]=\"showError() ? 'Must be before end date' : ''\"`.",
155
+ "appliesWhen": "No action when `[invalid]` is used WITHOUT `hint` — that is still supported and exactly what the input is for: the error state with the message living elsewhere, such as a form-level summary or one message for a whole ds-field-group. `invalid` is not deprecated.",
156
+ "note": "It used to promote `hint` to the ERROR state — a red message rendered through `role=\"alert\"`. It paints the control and sets aria-invalid only now. The old idiom lost the real hint (with both set, the help text was replaced rather than shown alongside), announced help text as an alert, and made a template unreadable because whether `hint` was a hint depended on another attribute.",
139
157
  "searchHints": [
140
158
  "[invalid]",
141
159
  "invalid",
@@ -158,7 +176,10 @@
158
176
  "kind": "behavior",
159
177
  "old": "a validator returning { warning: ... }",
160
178
  "new": null,
161
- "note": "The DsFieldError kind `warning` is now RESERVED, exported as DS_FIELD_WARNING_KIND. An error carrying it paints the field's warning state and shows its message, but leaves the control valid: aria-invalid stays off and a submit guard reading validity is not blocked. MOSTLY ADDITIVE -- it gives a form model a way to raise a non-blocking remark, which Angular's AbstractControl has no channel for. ACTION ONLY IF you already have a validator using the key `warning` (reactive/template-driven) or a signal-forms ValidationError with kind 'warning' for a BLOCKING error: it now reads as advisory, so the field stops being invalid and the submit stops being blocked. Rename that key to anything else -- e.g. `warningLimit`, `softLimit` -- and the old behaviour returns. Ranked below the explicit `warning` input, and shown only after interaction like any other validation message. DsFieldError and DsFieldErrorsInput are now exported as well, for writing a typed validator.",
179
+ "action": "Only if you already have a validator using the key `warning` — reactive, template-driven, or a signal-forms ValidationError with kind `warning` — for a BLOCKING error. Rename that key to anything else, such as `warningLimit` or `softLimit`, and the old behaviour returns.",
180
+ "appliesWhen": "Otherwise additive: it gives a form model a way to raise a non-blocking remark, which Angular's AbstractControl has no channel for. Ranked below the explicit `warning` input, and shown only after interaction.",
181
+ "pitfall": "This one changes VALIDITY, silently. An existing `warning` key that used to block now reads as advisory: aria-invalid stays off, the field stops being invalid, and a submit guard reading validity is no longer blocked. A form that refused to submit will now submit.",
182
+ "note": "The DsFieldError kind `warning` is reserved and exported as `DS_FIELD_WARNING_KIND`. DsFieldError and DsFieldErrorsInput are exported too, for writing a typed validator.",
162
183
  "searchHints": [
163
184
  "warning",
164
185
  "DS_FIELD_WARNING_KIND",
@@ -182,7 +203,10 @@
182
203
  "kind": "behavior",
183
204
  "old": "ds-segmented",
184
205
  "new": null,
185
- "note": "ds-segmented now stacks a label, the pill and a message row, so the PILL moved off :host onto an inner .track. The pill's own custom properties stayed on :host and inherit down, so a token override (--ds-segmented-*) is unaffected and needs nothing. ACTION only for a ::ng-deep rule that treated the HOST as the pill — a background, a border-radius, a display:grid, a border — which now paints the whole stack instead of the control: retarget it at `ds-segmented .track`. ALSO NEW, and additive: the component provides NG_VALUE_ACCESSOR and resolves its message through resolveFieldMessage, so [(ngModel)], formControlName, signal forms, and the label / hint / error / warning / errors inputs all work as they do on the other fields. Nothing that worked before stops working.",
206
+ "action": "Only for a `::ng-deep` rule that treated the HOST as the pill — a background, a border-radius, a `display: grid`, a border. Retarget it at `ds-segmented .track`.",
207
+ "appliesWhen": "A token override (`--ds-segmented-*`) is unaffected and needs nothing: the pill's custom properties stayed on `:host` and inherit down.",
208
+ "pitfall": "Left alone, such a rule now paints the whole STACK — label, pill and message row — instead of the control.",
209
+ "note": "ds-segmented stacks a label, the pill and a message row now, so the pill moved off `:host` onto an inner `.track`. Additive alongside: the component provides NG_VALUE_ACCESSOR and resolves its message through resolveFieldMessage, so both form systems and the label / hint / error / warning / errors inputs work as on the other fields.",
186
210
  "searchHints": [
187
211
  "ds-segmented",
188
212
  "::ng-deep",
@@ -6,7 +6,9 @@
6
6
  "kind": "behavior",
7
7
  "old": "class MyThing extends ButtonComponent (or any DS component)",
8
8
  "new": null,
9
- "note": "Every readback that resolves through DS_CONFIG changed from `protected` to public: `sizeResolved` on fourteen controls, `elevationResolved` and `paddingResolved` (ds-card), `withDotResolved` (ds-badge), `borderedResolved` and `stickyScrollbarResolved` (ds-table), `dividersResolved` (ds-modal and ds-modal-card), `reserveMessageResolved` (every field, via DsFieldBase), the date/time picker locale and format resolvers, and `brandName` / `letterGlyph` / `taglineText` (ds-sidenav, ds-auth-layout). WHY: 0.10.0 told consumers to read `sizeResolved()` now that `size()` returns null when unbound, and that did not compile -- reading it from a host, a test or a viewChild was TS2445. NO ACTION for a consumer that composes. ACTION only if you SUBCLASS a DS component and declare a member with one of those names: it is now a public member of the base, so an incompatible re-declaration needs `override` or a rename. Members a host can derive from its own bindings stay protected on purpose: `invalidResolved`, `railResolved`, `sortModeResolved`, `filterModeResolved`.",
9
+ "action": "Only if you SUBCLASS a DS component and declare a member with one of those names: it is a public member of the base now, so an incompatible re-declaration needs `override` or a rename.",
10
+ "appliesWhen": "No action for a consumer that composes. Members a host can derive from its own bindings stay protected on purpose: `invalidResolved`, `railResolved`, `sortModeResolved`, `filterModeResolved`.",
11
+ "note": "Every readback that resolves through DS_CONFIG went from `protected` to public — `sizeResolved` on fourteen controls, the card, badge, table and modal resolvers, `reserveMessageResolved` on every field, the picker locale and format resolvers, and the sidenav / auth-layout brand resolvers. 0.10.0 told consumers to read `sizeResolved()` once `size()` returned null when unbound, and that did not compile: reading it from a host, a test or a viewChild was TS2445.",
10
12
  "searchHints": [
11
13
  "sizeResolved",
12
14
  "elevationResolved",
@@ -23,6 +25,12 @@
23
25
  "verify": "No consumer subclass redeclares a `*Resolved` readback as protected, and a host that reads `sizeResolved()` off a DS component compiles.",
24
26
  "supersedes": [
25
27
  "0.10.0#control-size-input-becomes-nullable"
28
+ ],
29
+ "detect": [
30
+ "this.sizeResolved",
31
+ ".sizeResolved()",
32
+ "extends ButtonComponent",
33
+ "TS2415"
26
34
  ]
27
35
  },
28
36
  {
@@ -30,7 +38,9 @@
30
38
  "kind": "behavior",
31
39
  "old": "--focus-ring-inner: var(--surface-canvas)",
32
40
  "new": null,
33
- "note": "The 2px gap between an element and its focus ring no longer paints a named surface; it is transparent, so it shows whatever is actually behind. NO ACTION on a canvas-coloured ground -- the same pixels show through. It CHANGES what you see where the focused element sits on something else: a white card in a dark app used to get a dark rim around it, which read as a rendering artifact. If you deliberately relied on an opaque gap -- e.g. to hold the ring off a busy photo background -- set `--focus-ring-inner` back to a colour on that scope. Also new and unrelated to this remap: `data-focus=\"inset\"` draws the same ring INSIDE the border box, for a subtree under an `overflow: hidden` ancestor that would otherwise clip the ring away entirely. Unlike the other focus styles it is not pinned to :root -- set it on the clipping element.",
41
+ "action": "If you deliberately relied on an opaque gap — to hold the ring off a busy photo background, say — set `--focus-ring-inner` back to a colour on that scope.",
42
+ "appliesWhen": "No action on a canvas-coloured ground: the same pixels show through.",
43
+ "note": "The 2px gap between an element and its focus ring shows whatever is actually behind it now. A white card in a dark app used to get a dark rim around it, which read as a rendering artifact.",
34
44
  "searchHints": [
35
45
  "--focus-ring-inner",
36
46
  "focus-ring",
@@ -48,7 +58,10 @@
48
58
  "kind": "value-remap",
49
59
  "old": "ds-multi-select[maxVisibleChips]",
50
60
  "new": null,
51
- "note": "BREAKING VALUE REMAP. `null` used to be the LOUDEST setting - render every chip and let the trigger wrap - and it now means 'unset', resolving to 'fit' (measure the trigger, show what fits) like every other nullable input since 0.10.0. Bind the new 'all' mode to keep the old behaviour. This covers the ATTRIBUTE and the BINDING form alike: [maxVisibleChips]=\"null\", [maxVisibleChips]=\"undefined\", and any expression, field or method that can EVALUATE to null feeding that binding - retype those to `number | 'fit' | 'all' | null` and return 'all' where they meant show-them-all. A numeric value is unchanged (still a ceiling, still measured), 'fit' is unchanged, and leaving the input unbound is unchanged because 'fit' was already the default. SUPERSEDES the 0.10.0 guidance: that release told you to bind [maxVisibleChips]=\"null\" to keep every chip. If you did, that binding now means \"unset\" and resolves to \"fit\" — change it to \"all\".",
61
+ "action": "Bind the new `all` mode to keep the old behaviour. This covers the attribute and the binding alike: `[maxVisibleChips]=\"null\"`, `[maxVisibleChips]=\"undefined\"`, and any expression, field or method that can EVALUATE to null — retype those to `number | 'fit' | 'all' | null` and return `'all'` where they meant show-them-all.",
62
+ "appliesWhen": "A numeric value is unchanged (still a ceiling, still measured), `fit` is unchanged, and leaving the input unbound is unchanged because `fit` was already the default.",
63
+ "pitfall": "This SUPERSEDES the 0.10.0 guidance, which told you to bind `[maxVisibleChips]=\"null\"` to keep every chip. If you followed it, that binding now means \"unset\" and resolves to `fit` — so the release that told you to write it is the reason you have to change it.",
64
+ "note": "`null` used to be the LOUDEST setting — render every chip and let the trigger wrap — and it means \"unset\" now, resolving to `fit` like every other nullable input since 0.10.0.",
52
65
  "searchHints": [
53
66
  "maxVisibleChips",
54
67
  "[maxVisibleChips]",
@@ -68,15 +81,20 @@
68
81
  "'fit'"
69
82
  ],
70
83
  "supersedes": [
71
- "0.10.0#chips-cap-by-width-by-default"
72
- ]
84
+ "0.10.0#chips-cap-by-width-by-default",
85
+ "0.10.0#max-visible-chips-zero-inverted"
86
+ ],
87
+ "verify": "A ds-multi-select left without maxVisibleChips caps its chips to the trigger width and shows \"+N more\"; one bound to \"all\" shows every chip and wraps; a numeric cap shows at most that many. No trigger renders an empty chip row."
73
88
  },
74
89
  {
75
90
  "id": "max-visible-chips-null-is-unset-autocomplete",
76
91
  "kind": "value-remap",
77
92
  "old": "ds-autocomplete[maxVisibleChips]",
78
93
  "new": null,
79
- "note": "BREAKING VALUE REMAP. `null` used to be the LOUDEST setting - render every chip and let the trigger wrap - and it now means 'unset', resolving to 'fit' (measure the trigger, show what fits) like every other nullable input since 0.10.0. Bind the new 'all' mode to keep the old behaviour. This covers the ATTRIBUTE and the BINDING form alike: [maxVisibleChips]=\"null\", [maxVisibleChips]=\"undefined\", and any expression, field or method that can EVALUATE to null feeding that binding - retype those to `number | 'fit' | 'all' | null` and return 'all' where they meant show-them-all. A numeric value is unchanged (still a ceiling, still measured), 'fit' is unchanged, and leaving the input unbound is unchanged because 'fit' was already the default. SUPERSEDES the 0.10.0 guidance: that release told you to bind [maxVisibleChips]=\"null\" to keep every chip. If you did, that binding now means \"unset\" and resolves to \"fit\" — change it to \"all\".",
94
+ "action": "Bind the new `all` mode to keep the old behaviour. This covers the attribute and the binding alike: `[maxVisibleChips]=\"null\"`, `[maxVisibleChips]=\"undefined\"`, and any expression, field or method that can EVALUATE to null — retype those to `number | 'fit' | 'all' | null` and return `'all'` where they meant show-them-all.",
95
+ "appliesWhen": "A numeric value is unchanged (still a ceiling, still measured), `fit` is unchanged, and leaving the input unbound is unchanged because `fit` was already the default.",
96
+ "pitfall": "This SUPERSEDES the 0.10.0 guidance, which told you to bind `[maxVisibleChips]=\"null\"` to keep every chip. If you followed it, that binding now means \"unset\" and resolves to `fit` — so the release that told you to write it is the reason you have to change it.",
97
+ "note": "`null` used to be the LOUDEST setting — render every chip and let the trigger wrap — and it means \"unset\" now, resolving to `fit` like every other nullable input since 0.10.0.",
80
98
  "searchHints": [
81
99
  "maxVisibleChips",
82
100
  "[maxVisibleChips]",
@@ -96,15 +114,20 @@
96
114
  "'fit'"
97
115
  ],
98
116
  "supersedes": [
99
- "0.10.0#chips-cap-by-width-by-default"
100
- ]
117
+ "0.10.0#chips-cap-by-width-by-default",
118
+ "0.10.0#max-visible-chips-zero-inverted"
119
+ ],
120
+ "verify": "A ds-autocomplete left without maxVisibleChips caps its chips to the trigger width and shows \"+N more\"; one bound to \"all\" shows every chip and wraps; a numeric cap shows at most that many. No trigger renders an empty chip row."
101
121
  },
102
122
  {
103
123
  "id": "max-visible-chips-type-widened",
104
124
  "kind": "behavior",
105
125
  "old": "maxVisibleChips: number | 'fit' | null",
106
126
  "new": null,
107
- "note": "TYPE WIDENING on ds-autocomplete and ds-multi-select, breaking for READERS and OVERRIDERS, not for writers - the same asymmetry 0.10.0's two widenings had, where the consumer's type-checker was the only finder. A binding compiles as before; what stops compiling is an explicitly typed declaration that has to accept the value (`const cap: number | 'fit' | null = ...`, a helper parameter, a viewChild target) and a subclass override, because the `equal` type-parameter position of InputSignalWithTransform is invariant (TS2416). ACTION: retype to `number | 'fit' | 'all' | null`, or drop the annotation and let inference carry it. The VALUE change that comes with it is the separate max-visible-chips-null-is-unset-* pair; this entry is only the declaration.",
127
+ "action": "Retype to `number | 'fit' | 'all' | null`, or drop the annotation and let inference carry it.",
128
+ "appliesWhen": "A binding compiles as before. The VALUE change is the separate `max-visible-chips-null-is-unset-*` pair; this entry is only the declaration.",
129
+ "pitfall": "Breaking for READERS and OVERRIDERS and invisible to writers — the same asymmetry as 0.10.0's two widenings, where the consumer's type-checker was the only finder. A subclass override fails with TS2416 because the `equal` type-parameter position of InputSignalWithTransform is invariant.",
130
+ "note": "The declared type gained the `all` member on ds-autocomplete and ds-multi-select.",
108
131
  "searchHints": [
109
132
  "maxVisibleChips",
110
133
  "InputSignalWithTransform",
@@ -124,7 +147,10 @@
124
147
  "kind": "behavior",
125
148
  "old": "--ds-sidenav-item-bg-color-active",
126
149
  "new": null,
127
- "note": "VISUAL CHANGE with no required edit, plus one override that should now GO. The active ds-sidenav-item filled with --accent and labelled itself --text-on-primary; it is now a wash (--accent-subtitle) behind an --accent-strong label, and the active-hover mix goes toward the accent (--accent-muted) instead of 88% toward white, which was invisible on a wash. Material fills its drawer's active item with secondary-container and Carbon with $background-selected; neither uses the brand accent. NO TOKEN WAS RENAMED - every --ds-sidenav-item-*-active hook keeps its name, so an app that already retuned them looks exactly as it did and needs no change. ACTION only if you replaced the solid pill yourself (two apps did): compare your override against the new default and drop it if they now agree, since keeping it means maintaining a copy of the DS default. SEPARATELY: the edge rail is an input now - [rail]=\"false\" on ds-sidenav drops it for every item and [rail] on ds-sidenav-item overrides that per item (unset = follow the parent). If you hid the rail with `--ds-sidenav-item-rail: transparent` or a ::ng-deep box-shadow reset, both still work, but the input is the supported route. SidenavContract gained a required `rail: () => boolean` member - only relevant if you provide the SIDENAV token yourself, which nothing outside the library is expected to do. THE RAIL ALSO CHANGED COLOUR, and this one is a contrast fix rather than a taste call: it was --secondary (#FF7B2E) and is now --accent-strong, the active label's own colour. An orange bar across a monochrome-blue row measured 2.33:1 on the light wash, under the 3:1 floor a non-text indicator needs; the accent measures 6.16:1 light and 5.28:1 dark. RESTORE the old bar with `--ds-sidenav-item-rail: var(--secondary)` if the second hue was deliberate.",
150
+ "action": "Only if you replaced the solid pill yourself: compare your override against the new default and DROP it if they now agree, since keeping it means maintaining a copy of the DS default. The edge rail is an input now — `[rail]=\"false\"` on ds-sidenav drops it for every item, `[rail]` per item overrides that.",
151
+ "appliesWhen": "No token was renamed: every `--ds-sidenav-item-*-active` hook keeps its name, so an app that already retuned them looks exactly as it did. Hiding the rail through `--ds-sidenav-item-rail: transparent` or a `::ng-deep` box-shadow reset still works, though the input is the supported route. SidenavContract gained a required `rail: () => boolean`, which matters only if you provide the SIDENAV token yourself.",
152
+ "pitfall": "The rail also changed colour, and that one is a CONTRAST fix rather than a taste call: it was `--secondary` (orange) and is `--accent-strong` now. An orange bar across a monochrome-blue row measured 2.33:1 on the light wash, under the 3:1 floor a non-text indicator needs; the accent measures 6.16:1 light and 5.28:1 dark. Restore it with `--ds-sidenav-item-rail: var(--secondary)` only if the second hue was deliberate.",
153
+ "note": "The active item filled with `--accent` under a `--text-on-primary` label; it is a wash (`--accent-subtitle`) behind an `--accent-strong` label now, and the active-hover mix goes toward the accent instead of 88% toward white, which was invisible on a wash. Material fills its drawer active item with secondary-container and Carbon with $background-selected; neither uses the brand accent.",
128
154
  "searchHints": [
129
155
  "--ds-sidenav-item-bg-color-active",
130
156
  "--ds-sidenav-item-label-color-active",
@@ -141,7 +167,10 @@
141
167
  "kind": "behavior",
142
168
  "old": "extends TableComponent",
143
169
  "new": null,
144
- "note": "This release's class-member-collision entry, the counterpart of 0.8.0's, 0.9.0's and 0.10.0's: subclassing a DS component is unsupported (design-system.md), but where it happens a new base member with the same name stops the build - TS4114 demands `override`, TS2415 fires on a redeclared private. GAINED MEMBERS, per `npm run class-surface` against v0.10.1: AutocompleteComponent and MultiSelectComponent (chipsMode, showsEveryChip); MenuItemComponent (description, descriptionId); SearchComponent (disabled, isDisabled); SidenavComponent (rail); SidenavItemComponent (rail, railResolved); TableComponent (actionReason, actionsFor, hasActionsFor, visibleRowActions). NO members were lost. CHANGED SIGNATURES, which collide differently: SearchComponent.clear() went protected to public and SearchComponent.focus() now carries `override` (it overrides the CVA base) - a subclass that redeclared either needs `override` and a compatible signature. ACTION: rename the subclass's own member, or add `override` where the base's semantics genuinely match. The separate resolved-readbacks-are-public entry covers the protected-to-public sweep across fourteen classes, which collides the same way. Consumers who only use the <ds-...> elements in templates need no change. LATER THE SAME DAY, from a second `npm run class-surface` run after the overlay-guard fix: TimeColumnsComponent is unchanged but TimePickerComponent also gained the private `insideATrigger` - a redeclared private fails with TS2415 exactly like a public collision, which is why a private counts here at all. AND, from the state-glyph work: `messageIcon` on all FIFTEEN field components (AutocompleteComponent, CheckboxComponent, ColorPickerComponent, DatePickerComponent, InputComponent, MultiSelectComponent, OtpInputComponent, RadioComponent, RadioGroupComponent, SegmentedComponent, SelectComponent, SliderComponent, SwitchComponent, TextareaComponent, TimePickerComponent), plus `iconResolved` and `config` on FieldMessageComponent - which is internal and not exported, so it only matters to a fork.",
170
+ "action": "Run `npm run class-surface v0.10.1 v0.10.2` for the authoritative list, then per member: rename your own, or add `override` where the base semantics genuinely match.",
171
+ "appliesWhen": "Only a consumer that EXTENDS a DS component class. Subclassing is unsupported (design-system.md § Components); template users of the `<ds-…>` elements need nothing. No members were lost this release.",
172
+ "pitfall": "A redeclared PRIVATE fails with TS2415 exactly like a public collision, which is why `TimePickerComponent.insideATrigger` counts here at all. And `FieldMessageComponent` gained `iconResolved` and `config` but is internal and not exported, so those matter only to a fork.",
173
+ "note": "Gained: AutocompleteComponent and MultiSelectComponent (`chipsMode`, `showsEveryChip`); MenuItemComponent (`description`, `descriptionId`); SearchComponent (`bridge`, `disabled`, `formDisabled`, `hidden`, `isDisabled`, `markTouched`, `registerOnChange`, `registerOnTouched`, `setDisabledState`, `writeValue` — it declares the control-value-accessor surface itself now); SidenavComponent (`rail`); SidenavItemComponent (`rail`, `railResolved`); TableComponent (`actionReason`, `actionsFor`, `hasActionsFor`, `visibleRowActions`); TimePickerComponent (private `insideATrigger`); and `messageIcon` on all fifteen field components. TS4114 demands `override`, TS2415 fires on a redeclared private. The protected-to-public sweep across fourteen classes is the separate `resolved-readbacks-are-public` entry and collides the same way.",
145
174
  "searchHints": [
146
175
  "extends TableComponent",
147
176
  "extends SearchComponent",
@@ -153,7 +182,14 @@
153
182
  "TS4114",
154
183
  "extends TimePickerComponent",
155
184
  "messageIcon",
156
- "extends InputComponent"
185
+ "extends InputComponent",
186
+ "writeValue",
187
+ "registerOnChange",
188
+ "registerOnTouched",
189
+ "setDisabledState",
190
+ "markTouched",
191
+ "formDisabled",
192
+ "bridge"
157
193
  ],
158
194
  "surface": "typescript",
159
195
  "detect": [
@@ -168,7 +204,10 @@
168
204
  "kind": "behavior",
169
205
  "old": "SearchComponent.clear",
170
206
  "new": null,
171
- "note": "CHANGED TYPE, same name: SearchComponent.clear went from `protected clear(): void` to a public `clear(): void`. For a CALLER this is purely additive — the method is now reachable, which is the point (see the ds-search.clear() entry). It breaks an OVERRIDER: a subclass that redeclared `protected clear()` no longer satisfies the base, because a derived class cannot narrow a public member to protected (TS2445-class error). ACTION: drop the `protected` keyword on your override, or rename your own method. SearchComponent.focus separately picked up an `override` modifier, which changes nothing for a consumer.",
207
+ "action": "Drop the `protected` keyword on your override, or rename your own method.",
208
+ "appliesWhen": "For a CALLER this is purely additive — the method is reachable now, which is the point.",
209
+ "pitfall": "A derived class cannot narrow a public member to protected, so a subclass that redeclared `protected clear()` no longer satisfies the base.",
210
+ "note": "`SearchComponent.clear` went from `protected clear(): void` to a public `clear(): void`.",
172
211
  "searchHints": [
173
212
  "extends SearchComponent",
174
213
  "clear()",
@@ -187,7 +226,9 @@
187
226
  "kind": "behavior",
188
227
  "old": "TimeColumnsComponent.onKey",
189
228
  "new": null,
190
- "note": "CHANGED TYPE, same name: TimeColumnsComponent.onKey's parameter list changed. It is `protected`, so no consumer calls it and templates are unaffected; it is listed because an overriding subclass now has a signature that does not match the base (TS2416). ACTION: read the current declaration in time-columns.component.ts and match it, or drop the override — the component is internal to ds-time-picker and subclassing it was never supported.",
229
+ "action": "Read the current declaration in `time-columns.component.ts` and match it, or drop the override.",
230
+ "appliesWhen": "It is `protected`, so no consumer calls it and templates are unaffected. The component is internal to ds-time-picker and subclassing it was never supported.",
231
+ "note": "The parameter list changed, so an overriding subclass has a signature that no longer matches the base (TS2416).",
191
232
  "searchHints": [
192
233
  "extends TimeColumnsComponent",
193
234
  "onKey",
@@ -206,7 +247,10 @@
206
247
  "kind": "behavior",
207
248
  "old": "ds-select",
208
249
  "new": null,
209
- "note": "BEHAVIOUR CHANGE, narrow but test-visible. Five overlays now test their TRIGGER, not the whole host, in the outside-pointer guard: ds-select, ds-multi-select, ds-autocomplete, ds-time-picker and ds-color-picker. A press in the host's own empty width — the label, the message row, the slack beside a content-width trigger — used to be treated as INSIDE and left the panel open; it now dismisses, which is the ds-date-picker fix from 0.10.1 (MBP-225) applied to its siblings. NO ACTION in normal use: it removes a dead zone rather than adding one, and a press on the trigger itself still keeps the panel open, which is what that exception exists for (the caret, and re-focus). ACTION where a TEST or a scripted flow relied on the old dead zone — clicking a picker's LABEL while its panel is open, to focus or assert without closing it — that press now closes the panel, so the assertion after it sees a closed overlay. Retarget such a click at the trigger, or assert the closed state instead. ds-time-picker checks whichever of its three trigger boxes are present, since a range renders two.",
250
+ "action": "Only where a TEST or a scripted flow relied on the old dead zone — clicking a picker's LABEL while its panel is open, to focus or assert without closing it. Retarget such a click at the trigger, or assert the closed state instead.",
251
+ "appliesWhen": "No action in normal use: it removes a dead zone rather than adding one, and a press on the trigger itself still keeps the panel open. ds-time-picker checks whichever of its three trigger boxes are present, since a range renders two.",
252
+ "pitfall": "The press now closes the panel, so an assertion written after it sees a closed overlay rather than failing where the click was.",
253
+ "note": "Five overlays test their TRIGGER rather than the whole host in the outside-pointer guard — ds-select, ds-multi-select, ds-autocomplete, ds-time-picker and ds-color-picker. A press in the host's own empty width used to count as INSIDE. It is the ds-date-picker fix from 0.10.1 applied to its siblings.",
210
254
  "searchHints": [
211
255
  "ds-select",
212
256
  "ds-multi-select",
@@ -230,7 +274,10 @@
230
274
  "kind": "behavior",
231
275
  "old": "ds-stat[value]",
232
276
  "new": null,
233
- "note": "TYPE WIDENING, breaking for READERS and OVERRIDERS and invisible to writers: StatComponent.value went `string | number` to `string | number | null`. Every existing binding compiles unchanged and nothing renders differently (null renders as nothing, like the empty-string default). ACTION only where the type is written down: an explicitly typed declaration that has to accept it (`const shown: string | number = stat.value()`), a helper parameter, or a subclass override, which fails with TS2416 because the `equal` type-parameter position of InputSignal is invariant - retype to `string | number | null`, or drop the annotation. WHY it was widened: Angular's number / currency / percent / date pipes are all typed `string | null`, so `[value]=\"count | number\"` failed strictTemplates with TS2322 and pushed consumers to `count.toLocaleString('en-US')` in the template, which pins a locale instead of reading LOCALE_ID. If you have that workaround, the pipe is now the better form.",
277
+ "action": "Retype to `string | number | null`, or drop the annotation. If you worked around the old type with `count.toLocaleString('en-US')` in the template, the pipe is the better form now.",
278
+ "appliesWhen": "Every existing binding compiles unchanged and nothing renders differently — null renders as nothing, like the empty-string default.",
279
+ "pitfall": "Breaking for READERS and OVERRIDERS and invisible to writers. A subclass override fails with TS2416 because the `equal` type-parameter position of InputSignal is invariant.",
280
+ "note": "Angular's number, currency, percent and date pipes are all typed `string | null`, so `[value]=\"count | number\"` failed strictTemplates with TS2322 and pushed consumers into a template-side `toLocaleString`, which pins a locale instead of reading LOCALE_ID.",
234
281
  "searchHints": [
235
282
  "ds-stat",
236
283
  "toLocaleString",
@@ -251,7 +298,10 @@
251
298
  "kind": "behavior",
252
299
  "old": "ds-field-message",
253
300
  "new": null,
254
- "note": "APPEARANCE CHANGE on every field that shows a state: the message line now leads with a glyph for `error` (icon `error`), `warning` (`warn`) and `success` (`success`). The `hint` line is unchanged - no glyph - because it is the resting text rather than a state, and it is the majority of rows on a real form. NO CODE ACTION for anyone who wants it. THE ONE LINE BACK, if you do not: `provideDsConfig({ field: { stateIcons: false } })` at the root. Per state instead of all: `stateIcons: { warning: false }` turns one off, `stateIcons: { hint: 'info' }` turns the hint line on, and an IconName replaces a glyph rather than removing it - each key independent. Per FIELD: every one of the fifteen field components gained a `messageIcon` input which wins over the config - `[messageIcon]=\"false\"` drops the glyph on that field alone, `messageIcon=\"info\"` replaces it. TWO NEW CSS HOOKS: --ds-field-message-icon-gap and --ds-field-message-icon-offset. ALSO CHANGED in the same row: `align-items` went `center` to `flex-start`, so a message that WRAPS keeps its glyph beside the first line instead of centred on the block - if you had compensated for the old centring with your own margin on that row, drop it. The glyph size is not a token (ds-icon takes a number): it is 14px, matching --icon-sm.",
301
+ "action": "None if you want it. The one line back is `provideDsConfig({ field: { stateIcons: false } })` at the root. Per state: `stateIcons: { warning: false }` turns one off, `stateIcons: { hint: 'info' }` turns the hint line on, and an IconName replaces a glyph rather than removing it. Per FIELD: every one of the fifteen field components gained a `messageIcon` input that wins over the config — `[messageIcon]=\"false\"` drops the glyph there alone.",
302
+ "appliesWhen": "The `hint` line is unchanged and carries no glyph: it is resting text rather than a state, and it is the majority of rows on a real form.",
303
+ "pitfall": "One thing rides along: `align-items` on that row went `center` to `flex-start`, so a message that WRAPS keeps its glyph beside the first line instead of centred on the block. If you compensated for the old centring with your own margin on that row, drop it.",
304
+ "note": "The message line leads with a glyph for `error`, `warning` and `success`. Two new hooks — `--ds-field-message-icon-gap` and `--ds-field-message-icon-offset`. The glyph size is not a token (ds-icon takes a number): it is 14px, matching `--icon-sm`.",
255
305
  "searchHints": [
256
306
  "ds-field-message",
257
307
  "--ds-field-message",
@@ -270,11 +320,14 @@
270
320
  "kind": "removed-api",
271
321
  "old": "ds-search[hint]",
272
322
  "new": null,
323
+ "action": "Delete the binding. If the field genuinely needs a label or a validation line, use ds-input — ds-search is a toolbar filter and deliberately has no message row.",
324
+ "appliesWhen": "Not affected: `value`, `disabled`, `hidden`, `size`, `debounce`, `clearable`, `placeholder`, `shortcut`, `showIcon`, `ariaLabel`.",
325
+ "pitfall": "It used to compile and do NOTHING, and it is a template type error now — which is the whole point of the change, but it means the upgrade surfaces bindings that looked fine for releases.",
326
+ "note": "Removed from ds-search, which never rendered any of them: `hint`, `error`, `warning`, `errors`, `invalid`, `touched`, `reserveMessage`, `messageIcon` and the `reserveMessageResolved` readback. They arrived by inheritance when search gained form binding, so the API reference advertised eight inputs whose values nothing read.",
273
327
  "surface": [
274
328
  "template",
275
329
  "typescript"
276
330
  ],
277
- "note": "REMOVED from ds-search, which never rendered any of them: hint, error, warning, errors, invalid, touched, reserveMessage, messageIcon (and the reserveMessageResolved readback). They arrived by inheritance when search gained form binding - it extended DsControlValueAccessor, which sits on DsFieldBase - so the API reference advertised eight inputs whose values nothing read. ACTION: delete the binding. It used to compile and do nothing; it is now a template type error, which is the whole point of the change. If the field genuinely needs a label or a validation line, use ds-input - ds-search is a toolbar filter and deliberately has no message row. NOT affected: value, disabled, hidden, size, debounce, clearable, placeholder, shortcut, showIcon, ariaLabel, and the CVA itself - formControl / formControlName / ngModel / signal forms all still bind. INVISIBLE TO class-surface, recorded by hand: the tool reads per-file declarations, and these were never declared in search.component.ts.",
278
331
  "verify": "Binding hint/error/warning/errors/invalid/touched/reserveMessage/messageIcon on a ds-search fails to compile, while formControl binding and the disabled surface still work.",
279
332
  "searchHints": [
280
333
  "ds-search",
@@ -283,6 +336,12 @@
283
336
  "hint=",
284
337
  "reserveMessage",
285
338
  "messageIcon"
339
+ ],
340
+ "detect": [
341
+ "<ds-search",
342
+ "error=",
343
+ "hint=",
344
+ "[errors]"
286
345
  ]
287
346
  },
288
347
  {
@@ -290,8 +349,10 @@
290
349
  "kind": "behavior",
291
350
  "old": "ds-search, ds-autocomplete, ds-date-picker, ds-number-stepper, ds-field-group",
292
351
  "new": null,
352
+ "action": "Only if you wrote a workaround: a rule that killed one of the two rings — `box-shadow: none` on an inner input, or a re-declared `--ds-field-focus-shadow` meant to hide the doubling — can go.",
353
+ "appliesWhen": "Deliberately NOT suppressed: a ds-number-stepper step button and a chip close button keep their own ring, because both are `tabindex=0` tab stops the shell halo cannot distinguish.",
354
+ "note": "A focused field drew TWO rings — the tight one the universal `:where(:focus-visible)` box-shadow shipped in 0.10.0 puts on the inner control, and the halo the shell paints on `:focus-within`. The inner one is suppressed where the shell speaks for the control.",
293
355
  "surface": "style",
294
- "note": "APPEARANCE FIX, no action: a focused field drew TWO rings - a tight one around the inner control (the universal :where(:focus-visible) box-shadow shipped in 0.10.0) and the halo the shell paints on :focus-within. The inner one is now suppressed where the shell speaks for the control. ACTION only if you wrote a workaround: a rule that killed one of the two rings (box-shadow: none on an inner input, or a re-declared --ds-field-focus-shadow meant to hide the doubling) can go. Deliberately NOT suppressed: a ds-number-stepper step button and a chip close button keep their own ring, because both are tabindex=0 tab stops that the shell halo cannot distinguish.",
295
356
  "verify": "A focused ds-search, ds-autocomplete, ds-date-picker or ds-input inside a ds-field-group shows exactly one ring; a step button inside ds-number-stepper still shows its own when tabbed to.",
296
357
  "searchHints": [
297
358
  "--ds-field-focus-shadow",
@@ -306,8 +367,10 @@
306
367
  "kind": "behavior",
307
368
  "old": ".option-label, .cell-clip, .crumb, ds-breadcrumbs .label",
308
369
  "new": null,
370
+ "action": "Only if you worked around it: a `padding-block` or `line-height` override added to an option label, a table cell clip or a breadcrumb to get the descenders back can go.",
371
+ "appliesWhen": "No geometry change — option row 30px, multi-select 36px, table row 38px are all unchanged. For your own CSS of the same shape: `overflow: clip` keeps `text-overflow: ellipsis` working and keeps `scrollWidth > clientWidth` reporting true, both checked in Chromium, so a clip-detection test needs no change. Where `overflow-clip-margin` is unsupported, clip degrades to exactly the old shave.",
372
+ "note": "Every clipped label sat in a `line-height: 1` box with no padding, so 2.13px above the cap line and 1.47px below the baseline were cut (MBP-245). They use `overflow: clip` with a `--space-1` clip margin now.",
309
373
  "surface": "style",
310
- "note": "APPEARANCE FIX, no action, no geometry change (option row 30px, multi-select 36px, table row 38px are all unchanged): every clipped label sat in a line-height:1 box with no padding, so 2.13px above the cap line and 1.47px below the baseline were cut - MBP-245. They now use overflow: clip with a --space-1 clip margin. ACTION only if you worked around it: a padding-block or line-height override added to an option label, a table cell clip or a breadcrumb to get the descenders back can go. NOTE for a consumer with the same shape in their own CSS: overflow: clip keeps text-overflow: ellipsis working and keeps scrollWidth > clientWidth reporting true, both checked in Chromium, so a clip-detection test does not need changing. Where overflow-clip-margin is unsupported, clip degrades to exactly the old shave.",
311
374
  "verify": "An option list, a clipped table cell and a breadcrumb render full descenders (the tail of a g is complete) at unchanged row heights.",
312
375
  "searchHints": [
313
376
  "option-label",
@@ -329,7 +392,10 @@
329
392
  "ds-button variant=\"tonal\": the disabled label now uses the same 40% mix toward --surface-default as ghost / text / outline, instead of the shared oklch(from accent 70% ...) recipe. Visual only, no action. It fixes the dark theme, where the old recipe left the disabled label 1.11:1 from the enabled one (a hue going grey and nothing else); the state gap is now 2.90:1 dark and 3.39:1 light, and the label keeps its hue. --ds-button-label-color-disabled still overrides it.",
330
393
  "ds-table: TableColumn gained `interactive?: boolean`. The marked column's cell takes --ds-table-cell-link-color (default --accent) and --ds-table-cell-link-decoration (default underline) while its row is hovered, on a rowInteractive table only. Purely additive - an existing column config renders exactly as before. Reach for it on the cell whose TEXT names what the row opens; the row tint says the row is clickable and nothing else said which text was the target.",
331
394
  "ds fields on the SIGNAL FORMS path: they threw once per field (NgControl bridge has no events stream) and then displayed the first validation failure forever, because the frozen snapshot was the fallback whenever live errors went empty. Fixed - the NgControl path now bows out when it has no live stream. No consumer action: if you bind [field] on a DS control, upgrade and the console noise and the stale message both go. Reactive and template-driven forms are unchanged.",
332
- "DS_CONFIG: provideDsConfig() must be called ONCE per injector. Two calls in the same `providers` array do not merge - both resolve their parent with skipSelf, so the later wins and the earlier is silently lost. Merging happens across injector LEVELS (a route or component inheriting the app's config), not within one array. Documented on the function after it bit us in our own docs app."
395
+ "DS_CONFIG: provideDsConfig() must be called ONCE per injector. Two calls in the same `providers` array do not merge - both resolve their parent with skipSelf, so the later wins and the earlier is silently lost. Merging happens across injector LEVELS (a route or component inheriting the app's config), not within one array. Documented on the function after it bit us in our own docs app.",
396
+ "data-focus=\"inset\" draws the focus ring INSIDE the border box, for a subtree under an overflow:hidden ancestor that would otherwise clip it away and leave the element focusable with no indicator at all. Unlike the other focus styles it is NOT pinned to :root - set the attribute on the clipping element and the declarations inherit into the subtree.",
397
+ "migrations/<version>.json - the files dist/ds/migrations carries into node_modules - now carry `surface`, `detect`, `verify`, `restore`, `supersedes` and a structured `deleteList` on every published release. Read `surface` for which artifact to search, `detect` plus searchHints to find it, `verify` for the positive assertion a re-check can test, and deleteList[].find for code that is now redundant. `supersedes` points BACKWARD only: a manifest never references a later version, and the lint rejects one that tries.",
398
+ "components.md - the digest tools/build-schematics.mjs copies into dist/ds for a consumer's own agent to read - was missing ds-app-layout and ds-auth-layout entirely, and named --primary, --text-on-primary and --secondary as the active sidenav item defaults, which this release replaced with the wash. Both corrected. If you hand-rolled an app shell or a sign-in screen because the digest did not list one, or overrode those three tokens to restyle the active item, re-read those entries."
333
399
  ],
334
400
  "deleteList": [
335
401
  {
@@ -464,7 +530,7 @@
464
530
  "box-shadow: none",
465
531
  "--ds-field-focus-shadow"
466
532
  ],
467
- "why": "A rule that killed one of the two focus rings a field used to paint - on an inner input, or by re-declaring the field focus shadow.",
533
+ "why": "A rule that killed one of the two focus rings a field used to paint - on an inner input, or by re-declaring the field focus shadow. SCOPE, and it matters: this release suppresses the inner ring on ds-search, ds-autocomplete, ds-date-picker, ds-number-stepper and ds-field-group ONLY. A workaround on ds-input or ds-textarea is still load-bearing here - deleting it brings the double ring back. Leave those two until you are on the release that names them. Also leave any rule you wrote against the universal :where(:focus-visible) base indicator itself: that rule is still the one this library ships.",
468
534
  "replaceWith": "nothing: the field paints one ring now"
469
535
  },
470
536
  {
@@ -6,8 +6,10 @@
6
6
  "kind": "behavior",
7
7
  "old": "ds-input, ds-textarea",
8
8
  "new": null,
9
+ "action": "Only if you wrote a workaround: a rule that killed one of the two rings — `box-shadow: none` on an inner input, or a re-declared `--ds-input-focus-shadow` meant to hide the doubling — can go.",
10
+ "appliesWhen": "Deliberately NOT suppressed: the clear button and the password toggle inside the same box keep their own ring, because both are real tab stops the box halo cannot distinguish.",
11
+ "note": "The two shells 0.10.2's pass missed. A focused ds-input or ds-textarea drew the 4px control ring the universal `:where(:focus-visible)` rule puts on the inner element, plus the 3px field halo the box paints on `:focus-within`. The inner one is suppressed now.",
9
12
  "surface": "style",
10
- "note": "APPEARANCE FIX, no action: the two shells 0.10.2's one-focus-ring-per-field pass missed. A focused ds-input or ds-textarea drew TWO rings - the 4px control ring the universal :where(:focus-visible) rule puts on the inner input/textarea, plus the 3px field halo the box paints on :focus-within. The inner one is now suppressed. ACTION only if you wrote a workaround: a rule that killed one of the two (box-shadow: none on an inner input, a re-declared --ds-input-focus-shadow meant to hide the doubling) can go. Deliberately NOT suppressed: the clear button and the password toggle inside the same box keep their own ring, because both are real tab stops the box halo cannot distinguish.",
11
13
  "verify": "Tab into a ds-input and a ds-textarea: each shows exactly one ring, the halo around the box. Tab on to the clear button or password toggle inside that field and a second, tighter ring appears on the button - that one is intended.",
12
14
  "searchHints": [
13
15
  "--ds-input-focus-shadow",
@@ -22,8 +24,9 @@
22
24
  "kind": "behavior",
23
25
  "old": "ds-search",
24
26
  "new": null,
27
+ "action": "Only if you re-added a ring to that button yourself — that rule can go. (delete list: search-clear-button-focus-rule)",
28
+ "note": "0.10.2's ring fix silenced the base indicator with a shell-wide reset, which also hit `.search-clear`, a `tabindex=0` tab stop: tabbing from the input to the clear button changed nothing on screen — box-shadow, outline, border, background and colour all identical — so a keyboard user could not tell it had focus. A WCAG 2.4.7 failure. Every shell names the single control it silences now.",
25
29
  "surface": "style",
26
- "note": "ACCESSIBILITY FIX, no action: 0.10.2's ring fix silenced the base focus indicator with a shell-wide reset, which also hit .search-clear - a tabindex=0 tab stop. Tabbing from the input to the clear button changed nothing on screen (box-shadow, outline, border, background and colour all identical), so a keyboard user could not tell the button had focus: a WCAG 2.4.7 failure. Every shell now names the single control it silences, so tab stops inside a field box keep their own ring. ACTION only if you re-added a ring to that button yourself - that rule can go.",
27
30
  "verify": "Type into a ds-search so the clear button appears, then Tab: the × shows its own ring on top of the field halo. Same for a chip close button in ds-autocomplete or ds-multi-select, and for the arrows in ds-number-stepper.",
28
31
  "searchHints": [
29
32
  "search-clear",
@@ -37,8 +40,10 @@
37
40
  "kind": "behavior",
38
41
  "old": "ds-date-picker, ds-time-picker",
39
42
  "new": null,
43
+ "action": "Only if you worked around either: a rule adding a ring to `.period-trigger` or `.time-action`, or a `border-radius` forced onto `.time-panel`, can go.",
44
+ "pitfall": "The mechanism is worth knowing before you style a DS panel yourself: a private `--_*` token declared on `:host` does NOT reach a panel the CDK has portaled out of that host, and an unresolvable `var()` silently becomes the property's initial value — `none` for box-shadow, `0` for border-radius. Read the PUBLIC `--ds-*` token from outside instead.",
45
+ "note": "Two tabbable buttons inside the portaled panels had no focus indicator at all, and one panel had square corners: both read a private focus-shadow token declared on the host, and with `outline: none` beside it nothing was left.",
40
46
  "surface": "style",
41
- "note": "ACCESSIBILITY FIX, no action: two tabbable buttons inside the portaled panels had no focus indicator at all, and one panel had square corners. Both read a private --_*-focus-shadow token declared on :host while the panel is portaled out of the host by the CDK, so the var() resolved to nothing and box-shadow fell back to its initial value none - with outline: none beside it, nothing was left. ds-time-picker's panel border-radius read the same way and fell back to 0. ACTION only if you worked around either: a rule adding a ring to .period-trigger or .time-action, or a border-radius forced onto .time-panel, can go.",
42
47
  "verify": "Open the ds-date-picker calendar and Tab to the month/year header: it shows a focus ring. Open the ds-time-picker panel and Tab to its action button: same. The time-picker panel has rounded corners.",
43
48
  "searchHints": [
44
49
  "period-trigger",
@@ -54,7 +59,7 @@
54
59
  "The shell list in styles/_focus.scss is now maintained from a measurement, not from a read of the code: it had been wrong for two releases. If you add a component that paints its own ring on :focus-within, suppress the base rule on the control it speaks for - `outline: 0` cannot do it, because the base indicator is a box-shadow.",
55
60
  "Two authoring mixins are now documented for consumers building their own composite controls: `focus.owns-the-focus-ring($control)` in styles/focus, and `text-clip.single-line-ellipsis` in styles/text-clip. Both are opt-in - no existing code changes.",
56
61
  "A private --_* token declared on :host does not reach a CDK-portaled panel, and an unresolvable var() silently becomes the property's initial value - none for box-shadow, 0 for border-radius. If you style a DS panel from your own CSS, read the PUBLIC --ds-* token there, not the private one.",
57
- "Eleven public tokens were always overridable but never reached the token reference, because the extractor searched for the literal `var(--ds-` and a resolver wrapped across lines put a newline inside it: --ds-container-pad-x, --ds-multi-select-border-color (+-focus), --ds-segmented-selected-bg-color / -label-color / --ds-segmented-radius, --ds-time-picker-option-bg-color-selected / -text-color-selected / -bg-color-hover, --ds-time-picker-trigger-bg-color / -border / -color. Nothing changed in the components - if you hand-rolled around one of these because it looked unsupported, the override works and the workaround can go."
62
+ "Fourteen public tokens were always overridable but never reached the token reference, because the extractor searched for the literal `var(--ds-` and a resolver wrapped across lines put a newline inside it: --ds-container-pad-x, --ds-multi-select-border-color (+-focus), --ds-segmented-selected-bg-color / -label-color / --ds-segmented-radius, --ds-time-picker-option-bg-color-selected / -text-color-selected / -bg-color-hover, --ds-time-picker-trigger-bg-color / -border / -color, --ds-field-message-icon-offset, --ds-time-picker-focus-shadow. Nothing changed in the components - if you hand-rolled around one of these because it looked unsupported, the override is supported and is the way to do it."
58
63
  ],
59
64
  "deleteList": [
60
65
  {
@@ -68,6 +73,18 @@
68
73
  ],
69
74
  "why": "A rule killing one of the two focus rings a ds-input or ds-textarea used to paint - typically box-shadow: none on the inner input, or --ds-input-focus-shadow re-declared as none to hide the doubling.",
70
75
  "replaceWith": "nothing - the component now paints one ring on its own"
76
+ },
77
+ {
78
+ "id": "search-clear-button-focus-rule",
79
+ "surface": "style",
80
+ "find": [
81
+ ".search-clear",
82
+ "ds-search",
83
+ ":focus-visible",
84
+ "box-shadow"
85
+ ],
86
+ "why": "A focus rule you added to the ds-search clear button because it showed no indicator at all - the shell-wide reset 0.10.2 used silenced that tab stop too. Every shell names the one control it silences now, so the button keeps its own ring.",
87
+ "replaceWith": "nothing - the component paints it"
71
88
  }
72
89
  ]
73
90
  }