uxml-preview 0.2.0 → 0.4.0

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.
package/CHANGELOG.md CHANGED
@@ -1,5 +1,147 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.4.0 — 2026-08-12
4
+
5
+ ### Added
6
+
7
+ - The Unity golden dumper can run from the command line, records the Unity
8
+ revision, editor scale, accessible font settings, and timestamp beside each
9
+ measurement, rejects `-nographics` and `-quit` up front, and exits non-zero
10
+ unless it writes exactly one JSON file per discovered case.
11
+
12
+ ### Changed
13
+
14
+ - Relative `@import` requests are deduplicated by `(url, from)`, matching Unity's
15
+ parent-relative loading. Root-fixed imports (`project://…` or a path beginning
16
+ with `/`) are globally deduplicated, and nested `<Style>` attachments now keep
17
+ their Unity subtree scope. The rule comes from five Unity measurement cases,
18
+ including the otherwise surprising split between `Packages/...` (relative)
19
+ and `/Packages/...` (root-fixed).
20
+
21
+ - **Issue #1: `resolveImport` now receives a second argument,
22
+ `from: string | null`** —
23
+ the URL of the stylesheet containing the import being resolved, `null` for
24
+ a `<Style src="…">` reference. Without it, a relative `@import` inside an
25
+ imported sheet was unresolvable in principle: `a.uss` importing `"b.uss"`
26
+ means a path relative to `a.uss`, and the host had no way to learn `a.uss`'s
27
+ own URL at the point it resolved `b.uss`. `from` is always the exact string
28
+ this hook received as `url` for the containing sheet, never reconstructed —
29
+ matters most for nested imports, where it is the immediate parent, not the
30
+ original sheet. **Non-breaking**: existing one-argument callbacks keep
31
+ working unchanged, same as `resolveAsset`'s `form` argument in 0.3.0.
32
+ - With this, `resolveImport` joins `resolveAsset` at two positional arguments.
33
+ Per the "코어에 무언가를 더하기 전에" rule in CLAUDE.md, the next argument
34
+ either hook would need is the point to switch to an options object instead
35
+ of adding a third positional one.
36
+
37
+ ### Fixed
38
+
39
+ - **Issue #2: inline `style="…"` asset URLs reach `resolveAsset`
40
+ XML-entity-decoded**,
41
+ matching what a `<Style src>`-loaded `.uss` file's URLs already got (that
42
+ file was never XML, so it never had entities to decode; an inline attribute
43
+ is XML text and did). `background-image: url('…&guid=…')` written inline
44
+ used to reach the hook as `&apos;…&amp;guid=…&apos;` — wrapper quotes and
45
+ ampersands both literal — because `inlineDeclarations` (style/resolve.ts)
46
+ parses straight from the raw XML source. Fixed at `assetPath()`
47
+ (render/css-map.ts), immediately before the `url()`/`resource()` wrapper is
48
+ parsed and right before the value leaves the model for the hook — the same
49
+ point `<Style src>` and painted text already decode at, not at parse time
50
+ (decoding then would corrupt the byte-exact round-trip source slices).
51
+ Confirmed against two real URLs from the external corpus in
52
+ `ReuHomi/vscode-uxml-preview` (`examples/external`, commit `dbda387`) where
53
+ this caused `asset-unresolved` for 7 URLs across 2 files. Scope: only the
54
+ asset-URL value reaching `resolveAsset`; other inline-style property values
55
+ still carry undecoded entities in the model (unaffected here, since no
56
+ other current consumer inspects their text — `-unity-font-definition`'s
57
+ `resource(...)` value is only checked for presence, never read).
58
+
59
+ ### Docs
60
+
61
+ - **`docs/accuracy.md`'s "which controls this covers" table now has a stated,
62
+ computed rule** instead of being left unverified. It never added up on its
63
+ own terms (17+1+6+5+1 = 30, not the stated 31) and the original counting
64
+ rule was never written down, so it is now defined explicitly — classify
65
+ `inventory` as the representative screen, then each other case by the first
66
+ of `<ui:ScrollView`, `<ui:Button`, `<ui:Label` its UXML contains, else
67
+ `VisualElement` only — and computed from `tests/golden/cases.ts` rather than
68
+ hand-counted. Updated numbers: `Button` 6→10, `ScrollView` 5→4 (`VisualElement`
69
+ only, `Label`, and the representative screen are unchanged). Mirrored into
70
+ `docs/accuracy.en.md`, which is not machine-checked.
71
+ - **The drift guard now covers this table's five numbers too**, in
72
+ `docs/accuracy.md` only. Same shape as the three headline figures: recomputed
73
+ live and failing (not skipping) if it can't find a row to check. Confirmed by
74
+ breaking a count and by removing a row's label, both on purpose.
75
+
76
+ ## 0.3.0 — 2026-08-09
77
+
78
+ Four fixes the VSCode extension (`vscode-uxml-preview`) hit while consuming
79
+ this package as a dependency — not speculative additions.
80
+
81
+ ### Added
82
+
83
+ - **`KNOWN_DIVERGENCES`** gives a host something to tell apart "a limit we
84
+ know about" from "a bug". Three entries, and the kinds carry the
85
+ distinction that matters. Font metrics move with the platform, so the
86
+ difference cannot be reproduced here at all. Unity's rule for a wrap
87
+ container's height is simply not specified, so there is nothing to match
88
+ yet. And the Yoga vendored in UI Toolkit resolves a main-axis percentage
89
+ against an indefinite parent differently from the one this package depends
90
+ on — that one reproduces perfectly every time; it just cannot be fixed
91
+ from inside this repository. The 1px case is deliberately absent: that is
92
+ the golden tests' judgement parameter, not a limit of the renderer, and a
93
+ host showing it would read as a defect.
94
+ - **`uxml-preview/unity-project`**, a Node-only subpath exporting
95
+ `buildGuidIndex(projectRoot)`. Builds a GUID → absolute-path index from a
96
+ Unity project's `.meta` files, for the case where an asset reference's path
97
+ has gone stale but its GUID is still good. Skips `Library/`, `Temp/`,
98
+ `obj/`, `.git/`, and any `.meta` file it cannot read. Kept out of the main
99
+ entry point because it uses `node:fs` and the main entry must stay
100
+ browser-safe.
101
+
102
+ ### Changed
103
+
104
+ - **`resolveAsset` now receives a second argument, `form: 'url' | 'resource'`**,
105
+ telling `url("...")` and `resource("...")` background-image references
106
+ apart. They resolve differently — `url()` is a path, `resource()` is a
107
+ Unity Resources-folder name — and a host that resolved both the same way
108
+ before this had no way to know which one it was looking at.
109
+ **Non-breaking**: existing `(path) => ...` implementations keep working
110
+ unchanged: `form` is an additional argument, not a replacement one.
111
+
112
+ ### Fixed
113
+
114
+ - `UxmlDocument.warnings` and `RenderResult.warnings` doc comments corrected.
115
+ The former said it covers "malformed input" only; it also carries
116
+ `'import-unresolved'` for an unreadable `<Style src="…">` or `@import`. The
117
+ latter said it is distinct from the former on that same wrong premise.
118
+ Comment-only — no behavior changed.
119
+
120
+ ### Docs
121
+
122
+ - **`docs/accuracy.md`'s published figures were stale**: two comparable cases
123
+ (`state-vs-id`, `state-vs-inline`) were merged after the doc's numbers were
124
+ last written, and nothing caught it. Recomputed from a live run rather than
125
+ trusted: **548 / 564 values (97.2%)**, **31 / 33 cases matching**, **33 / 33**
126
+ Unity baseline coverage — up from the stale 532/548, 29/31, 31/31. Mirrored
127
+ into `README.md` (both languages) and `docs/uss-vs-css.md`/`.ko.md`.
128
+ `docs/accuracy.en.md` was updated too but is not machine-checked (see next).
129
+ - **Added a regression guard**: `tests/golden/golden.test.ts` now has
130
+ `matches the figures published in docs/accuracy.md`, which recomputes the
131
+ three headline numbers from the actual case set and fails if
132
+ `docs/accuracy.md` says anything else — including failing (not skipping) if
133
+ it cannot find a row to check, so the guard can't be silently defeated by a
134
+ table-format change. This is the third time this specific number went stale
135
+ publicly (`docs/accuracy.md` itself already records the first two); the doc
136
+ fix alone would only prevent a third, not a fourth.
137
+ - Left unfixed and flagged in `docs/accuracy.md`/`accuracy.en.md`: the
138
+ "which controls this covers" breakdown table (`Button`/`ScrollView`/etc.
139
+ row counts) was already internally inconsistent before this change
140
+ (17+1+6+5+1 = 30, not the stated 31) and is not machine-checked. Recomputing
141
+ it requires a counting rule that is not written down anywhere; a guess
142
+ risked swapping one wrong number for another, so it was left alone and
143
+ called out inline instead.
144
+
3
145
  ## 0.2.0 — 2026-08-06
4
146
 
5
147
  Five controls instead of three, a representative screen measured end to end
package/README.md CHANGED
@@ -25,12 +25,14 @@ stylesheet comes out wrong.</sub>
25
25
  npm install uxml-preview
26
26
  ```
27
27
 
28
- > **v0.1.0 — early release.** Three controls (`VisualElement`, `Label`,
29
- > `Button`) and a support matrix that says plainly what has not been measured.
28
+ > **v0.2.0 — early release.** Five controls (`VisualElement`, `Label`, `Button`,
29
+ > `Image`, `ScrollView`), everything else drawn as a plain box with a warning,
30
+ > and a support matrix that says plainly what has not been measured.
30
31
  > Read [what is not verified yet](#what-is-not-verified-yet) before relying on it.
31
32
  >
32
- > **v0.1.0 — 초기 릴리스입니다.** 컨트롤 3종(`VisualElement`, `Label`, `Button`)이고,
33
- > 검증되지 않은 부분을 지원 범위표에 그대로 적어뒀습니다. 쓰시기 전에
33
+ > **v0.2.0 — 초기 릴리스입니다.** 컨트롤 5종(`VisualElement`, `Label`, `Button`,
34
+ > `Image`, `ScrollView`)이고, 나머지는 경고와 함께 일반 박스로 그립니다.
35
+ > 검증되지 않은 부분은 지원 범위표에 그대로 적어뒀습니다. 쓰시기 전에
34
36
  > [아직 검증되지 않은 것](#아직-검증되지-않은-것)을 먼저 봐주세요.
35
37
 
36
38
  ---
@@ -45,8 +47,8 @@ directly in a web browser. It uses **Yoga** — the same layout engine Unity UI
45
47
  itself uses — compiled to WebAssembly, so results match the Unity Editor rather than
46
48
  approximating it.
47
49
 
48
- Measured against Unity 6000.0.40f1: **532 of 548 element coordinates identical**
49
- across 31 layout cases, including a full working screen. Ten of the sixteen
50
+ Measured against Unity 6000.0.40f1: **660 of 676 element coordinates identical**
51
+ across 40 layout cases, including a full working screen. Ten of the sixteen
50
52
  differences are font metrics rather than layout — a browser cannot measure
51
53
  Unity's font asset. All sixteen are named in
52
54
  [`docs/accuracy.en.md`](docs/accuracy.en.md).
@@ -95,8 +97,9 @@ That broken feedback loop is what this library fixes.
95
97
  | 3 | USS style resolver | ✅ |
96
98
  | 4 | Yoga layout + DOM painting | ✅ |
97
99
  | 5 | Golden tests against Unity | ✅ |
98
- | 6 | **v0.1 release** | 🔨 |
99
- | 7+ | More controls, editing layer, tool integration | ⬜ |
100
+ | 6 | **v0.1 release** | ✅ |
101
+ | 7 | **v0.2 — ScrollView, Image, states, a measured screen** | ✅ |
102
+ | 8+ | Editing layer, tool integration | ⬜ |
100
103
 
101
104
  Full plan: [`docs/ROADMAP.md`](docs/ROADMAP.md)
102
105
 
@@ -207,12 +210,12 @@ Checked by comparing element geometry against the Unity Editor, case by case.
207
210
  Geometry rather than pixels: Unity draws text with its own font asset and a
208
211
  browser does not, so a pixel diff would measure the font more than the layout.
209
212
 
210
- Measured against Unity on 2026-08-02, over 18 cases and 61 elements:
213
+ Measured against Unity on 2026-08-12, over 40 cases and 169 elements:
211
214
 
212
215
  | | |
213
216
  |---|---|
214
- | Cases matching exactly | **29 / 31** |
215
- | Values within 0.5px | **532 / 548 (97.1%)** |
217
+ | Cases matching exactly | **38 / 40** |
218
+ | Values within 0.5px | **660 / 676 (97.6%)** |
216
219
 
217
220
  The sixteen divergences are four different things, and the distinction matters
218
221
  more than the ratio: **ten are text metrics**, three are 1px, two are a
@@ -243,7 +246,7 @@ as nested flex, and computed values have to be resolved to fixed numbers or
243
246
  calculated in C#.
244
247
 
245
248
  **How close is the output to Unity?**
246
- 532 of 548 element coordinates identical against Unity 6000.0.40f1, across 31
249
+ 660 of 676 element coordinates identical against Unity 6000.0.40f1, across 40
247
250
  layout cases including a complete inventory screen. Geometry is compared rather
248
251
  than screenshots, because Unity draws text with its own font asset — which is
249
252
  also why ten of the sixteen differences are font metrics rather than layout.
@@ -277,8 +280,8 @@ Apache-2.0
277
280
  **Yoga**를 WebAssembly로 그대로 사용하기 때문에, 결과를 "비슷하게 흉내내는" 것이 아니라
278
281
  유니티 에디터와 일치시킵니다.
279
282
 
280
- Unity 6000.0.40f1과 대조한 결과, 실무형 화면을 포함한 레이아웃 케이스 31개에서
281
- **요소 좌표 548개 중 532개가 일치**합니다. 어긋난 16개 중 **10개는 레이아웃이 아니라
283
+ Unity 6000.0.40f1과 대조한 결과, 실무형 화면을 포함한 레이아웃 케이스 40개에서
284
+ **요소 좌표 676개 중 660개가 일치**합니다. 어긋난 16개 중 **10개는 레이아웃이 아니라
282
285
  폰트 메트릭** 차이입니다 — 브라우저는 유니티 폰트 에셋을 잴 수 없습니다.
283
286
  16개 전부 [`docs/accuracy.md`](docs/accuracy.md)에 이름과 사유를 적어뒀습니다.
284
287
 
@@ -325,8 +328,9 @@ Unity 6000.0.40f1과 대조한 결과, 실무형 화면을 포함한 레이아
325
328
  | 3 | USS 스타일 리졸버 | ✅ |
326
329
  | 4 | Yoga 레이아웃 + DOM 페인팅 | ✅ |
327
330
  | 5 | 유니티 대조 골든 테스트 | ✅ |
328
- | 6 | **v0.1 공개** | 🔨 |
329
- | 7+ | 컨트롤 확장, 편집 레이어, 도구 통합 | ⬜ |
331
+ | 6 | **v0.1 공개** | ✅ |
332
+ | 7 | **v0.2 — ScrollView·Image·상태, 대표 화면 대조** | ✅ |
333
+ | 8+ | 편집 레이어, 도구 통합 | ⬜ |
330
334
 
331
335
  전체 계획: [`docs/ROADMAP.md`](docs/ROADMAP.md)
332
336
 
@@ -410,8 +414,8 @@ pnpm build
410
414
  계산이 필요한 값은 고정값으로 풀거나 C#에서 계산해야 합니다.
411
415
 
412
416
  **유니티와 얼마나 같나요?**
413
- Unity 6000.0.40f1 기준, 인벤토리 화면 전체를 포함한 레이아웃 케이스 31개에서
414
- **요소 좌표 548개 중 532개 일치**입니다. 스크린샷이 아니라 좌표를 비교하는데,
417
+ Unity 6000.0.40f1 기준, 인벤토리 화면 전체를 포함한 레이아웃 케이스 40개에서
418
+ **요소 좌표 676개 중 660개 일치**입니다. 스크린샷이 아니라 좌표를 비교하는데,
415
419
  유니티가 자기 폰트 에셋으로 글자를 그리기 때문입니다 — 그리고 그 사실 때문에
416
420
  불일치 16개 중 **10개가 레이아웃이 아니라 폰트 메트릭**입니다. 전부
417
421
  [`docs/accuracy.md`](docs/accuracy.md)에 이름이 적혀 있습니다.
package/dist/index.d.ts CHANGED
@@ -17,6 +17,8 @@ export { resolveStyles, explainProperty } from './style/resolve';
17
17
  export type { Candidate, ComputedStyle, ComputedValue, ResolveOptions, ResolveResult, } from './style/resolve';
18
18
  export type { Specificity } from './style/specificity';
19
19
  export { isInherited } from './style/properties';
20
+ export { KNOWN_DIVERGENCES } from './known-divergences';
21
+ export type { KnownDivergence } from './known-divergences';
20
22
  /**
21
23
  * Parse UXML and USS text into a document model.
22
24
  *
@@ -35,8 +37,23 @@ export interface ParseOptions {
35
37
  * `project://database/Assets/UI/base.uss`, which only the host application
36
38
  * knows how to turn into file contents, and there is no reason for a host to
37
39
  * implement that lookup twice.
40
+ *
41
+ * `from` is the URL of the stylesheet containing this import — exactly the
42
+ * string this hook received as `url` when that containing sheet was itself
43
+ * resolved, unmodified. `null` for a `<Style src="…">` reference, which is
44
+ * not contained in any stylesheet. A relative `@import` cannot be resolved
45
+ * without this: `a.uss` importing `"b.uss"` means a path relative to `a.uss`,
46
+ * not to the UXML document, and there is no other way for the host to learn
47
+ * `a.uss`'s own URL at the point it resolves `b.uss`.
48
+ *
49
+ * Non-breaking: this is an added argument, not a replacement one, so an
50
+ * existing one-argument callback keeps working unchanged.
51
+ *
52
+ * Relative imports are deduplicated by `(url, from)`, so two parent sheets
53
+ * may resolve the same spelling to different files. Root-fixed URLs —
54
+ * `project://…` or a path beginning with `/` — are deduplicated by URL alone.
38
55
  */
39
- resolveImport?: (url: string) => string | null;
56
+ resolveImport?: (url: string, from: string | null) => string | null;
40
57
  }
41
58
  export declare function parse(uxml: string, uss?: string, options?: ParseOptions): UxmlDocument;
42
59
  /**
@@ -57,8 +74,15 @@ export interface RenderOptions {
57
74
  * USS background images look like `url("project://database/Assets/foo.png")`
58
75
  * or `resource("foo")`. Browsers cannot load these, so the host application
59
76
  * must map them to real URLs. Return `null` to draw a placeholder instead.
77
+ *
78
+ * `form` says which of the two `path` came from — `'url'` is a path
79
+ * (project-relative or absolute), `'resource'` is a Unity Resources-folder
80
+ * name (no extension, resolved by Unity's own lookup rules). They are not
81
+ * interchangeable: a host that ignores `form` and resolves both as file
82
+ * paths will resolve `resource("foo")` against whatever happens to sit at
83
+ * that path — the wrong image, silently, if one exists there at all.
60
84
  */
61
- resolveAsset?: (path: string) => string | null;
85
+ resolveAsset?: (path: string, form: 'url' | 'resource') => string | null;
62
86
  /** Panel size in pixels. Defaults to the container's client size. */
63
87
  size?: {
64
88
  width: number;
@@ -94,7 +118,8 @@ export interface RenderResult {
94
118
  /**
95
119
  * Everything the renderer could not honour: unsupported controls, properties,
96
120
  * selectors and units, plus unresolved assets. Distinct from
97
- * `UxmlDocument.warnings`, which only covers malformed input.
121
+ * `UxmlDocument.warnings`, which covers malformed input and unresolved
122
+ * `<Style src="…">` / `@import` references rather than render-time support.
98
123
  */
99
124
  warnings: readonly Warning[];
100
125
  /** Painted element per model node. Lets a host map a click back to the tree. */