cg: enhance file I/O, layout engine, FBS decoder, and add new example fixtures#586
Merged
softmarshmallow merged 8 commits intomainfrom Mar 17, 2026
Merged
cg: enhance file I/O, layout engine, FBS decoder, and add new example fixtures#586softmarshmallow merged 8 commits intomainfrom
softmarshmallow merged 8 commits intomainfrom
Conversation
…alidation - Renamed the tool from "Grida File Validator" to "Grida File Inspector" to better reflect its functionality. - Expanded the tool's capabilities to inspect and validate both `.grida` and `.grida1` files in various formats (FlatBuffers, ZIP, JSON). - Added new command-line options for listing scenes, inspecting specific scenes, and running layout checks. - Updated documentation to include detailed usage instructions and feature descriptions. - Introduced a new module for file format detection and unified decoding, improving the overall architecture and usability of the tool.
- Updated the layout engine to extract layouts for all root nodes, including those that do not participate in Taffy (e.g., BooleanOperation, Group). - Added tests to verify that descendant containers receive layout results even when their parent is a non-Taffy node, preventing panics in GeometryCache. - Enhanced the handling of layout results to ensure all nodes have valid layout data during computation.
- Introduced a new SKILL.md file detailing the research skill for Grida, outlining its purpose and workflow for investigating upstream and peer projects. - Provided guidelines on when to use the skill, how to orient oneself in existing research, and key discovery queries for effective research practices. - Included references to relevant repositories and documentation to assist in the research process, enhancing the overall knowledge base for developers.
- Introduced a new example `tool_gen_cover.rs` that generates a cover image and a corresponding `.grida` file from a defined scene. - Added a `cover.rs` fixture module to build a visually rich scene showcasing the capabilities of the cg engine, including gradients, shadows, and typography. - Updated the `mod.rs` file to include the new `cover` module for better organization of example fixtures. - Added a new `cover.grida` file to store the generated scene data. - Updated the existing `cover.png` with a new version reflecting the latest design changes.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
2 Skipped Deployments
|
WalkthroughAdds a code-research SKILL doc; introduces multiple new example fixtures (cover, progressive blur, var-width strokes) and expanded transform examples; rewrites the CLI inspector into a format-aware inspector; adds a unified io_grida_file module; extends FlatBuffers encode/decode (progressive blur, varwidth strokes, richer text, diamond gradients, image filters, aspect-aware layout); and updates layout engine to compute layouts for all roots. Changes
Sequence Diagram(s)sequenceDiagram
participant CLI as "tool_io_grida (CLI)"
participant IO as "io_grida_file"
participant FBS as "io_grida_fbs"
participant ZIP as "ZIP extractor"
participant JSON as "JSON parser"
participant Layout as "Layout Engine"
rect rgba(220,240,255,0.5)
CLI->>IO: supply file bytes
IO->>IO: detect format (RawFbs / Zip / Json)
end
rect rgba(200,255,200,0.5)
alt RawFbs
IO->>FBS: decode bytes -> Scene(s) + id_map
else Zip
IO->>ZIP: extract document.grida -> fbs bytes
ZIP->>FBS: pass fbs bytes -> DecodeResult
else Json
IO->>JSON: parse JSON -> Canvas AST
JSON->>IO: convert -> Scene(s)
end
end
rect rgba(255,230,200,0.5)
CLI->>CLI: evaluate flags (--list-scenes/--scene/--layout-check)
CLI->>Layout: optionally run layout-check on Scene(s)
Layout->>Layout: compute layouts (includes extra roots)
CLI-->>CLI: present results (stats/tree/validation)
end
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
Suggested labels
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches
🧪 Generate unit tests (beta)
📝 Coding Plan
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
- Introduced a new scene `scene_progressive_blur_grid` showcasing progressive blur with varying gradient directions across a grid of rectangles. - Added `l0_effects_progressive_blur` fixture demonstrating layer and backdrop progressive blur effects. - Created `l0_strokes_varwidth` fixture to illustrate variable-width stroke profiles on vector paths with various taper and bulge patterns. - Updated `mod.rs` to include new fixture modules and enhanced the README with descriptions of the new examples. - Modified existing examples to incorporate the new features, ensuring comprehensive coverage of effects and strokes.
…ion options - Added support for decoding `DiamondGradientPaint` in the paint item processing. - Updated tile mode decoding to utilize the correct values from the FlatBuffers schema. - Introduced new text decoration options including `TextDecorationLine` and `TextDecorationStyle` for improved text styling capabilities. - Enhanced existing paint and layout decoding functions to accommodate new features and ensure consistency across the rendering pipeline. - Updated related tests and documentation to reflect these changes.
softmarshmallow
changed the title
Member
Author
|
@coderabbitai review |
✅ Actions performedReview triggered.
|
- Added references to the Adobe Photoshop File Format Specification and Figma .fig format documentation. - Enhanced contributor workflow section to guide users on regenerating bindings after editing `grida.fbs`. - Improved overall clarity and resource accessibility for developers working with file formats.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 7
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
crates/grida-canvas/src/io/io_grida_fbs.rs (1)
629-660:⚠️ Potential issue | 🟠 MajorPreserve tiled image paints in the decode path.
While expanding
ImagePainthere,decode_image_paint_fit()still has noImagePaintFitTilebranch, even though the encoder writes that variant at Line 2416. Tiled assets will silently decode asFit(Cover), which breaks rendering and round-trip fidelity.
🧹 Nitpick comments (2)
crates/grida-canvas/examples/fixtures/l0_layout_transform.rs (1)
214-249: Replace the duplicated numeric IDs with namedNodeIds.
build_scene()already acceptsNodeId-typed collections (crates/grida-canvas/examples/fixture_helpers/mod.rs:22-35), but this fixture repeats the same bare integers acrosspairs,links, androots. That makes later edits brittle: inserting or renumbering one node can silently break the parent/child wiring. PreferNodeIdconstants or a tiny allocator so each relationship is declared once.As per coding guidelines, use
NodeId(u64) for internal counter-based IDs in structs like NodeRecs, SceneGraph, and caches for high-performance operations.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/examples/fixtures/l0_layout_transform.rs` around lines 214 - 249, The fixture uses raw numeric IDs in pairs, links, and the roots vector which duplicates and risks breakage; replace these bare integers with named NodeId values (e.g., consts or a small NodeId allocator) and use the NodeId type everywhere build_scene is called so the same identifier is declared once and reused; update the pairs Vec<(u64, Node)> to Vec<(NodeId, Node)> (or cast/construct NodeId) and change the HashMap keys and roots Vec to the NodeId type, and ensure container keys (13,15,17) and their child references (14,16,18) are replaced by those named NodeId bindings so parent/child wiring is only defined once and remains stable when nodes are inserted or renumbered while still calling build_scene("L0 Layout Transform", None, pairs, links, roots).crates/grida-canvas/src/io/io_grida_file.rs (1)
91-134: Implementstd::error::ErrorforDecodeError.Right now callers like
crates/grida-dev/src/grida_file.rshave to stringify and re-wrap these errors, which drops theFbsDecodeErrorsource chain. AddingError::source()here keeps propagation and diagnostics intact.♻️ Proposed error-chain support
impl From<io_grida_fbs::FbsDecodeError> for DecodeError { fn from(e: io_grida_fbs::FbsDecodeError) -> Self { DecodeError::Fbs(e) } } + +impl std::error::Error for DecodeError { + fn source(&self) -> Option<&(dyn std::error::Error + 'static)> { + match self { + DecodeError::Fbs(e) => Some(e), + _ => None, + } + } +}🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/src/io/io_grida_file.rs` around lines 91 - 134, Add an impl of std::error::Error for DecodeError so callers can access error sources; implement fn source(&self) -> Option<&(dyn std::error::Error + 'static)> that returns Some(e) for the Fbs variant (DecodeError::Fbs(e)) and None for the other variants (UnrecognizedFormat, Zip, Json, Conversion, JsonNotSupported, NoScenes); ensure io_grida_fbs::FbsDecodeError implements Error (it already should) and import std::error::Error if needed.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.agents/skills/research/SKILL.md:
- Around line 33-44: The fenced code block in SKILL.md showing the directory
tree lacks a language specifier; update the opening fence (the triple backticks
that start the tree block) to include a language tag (e.g., "text") so the block
becomes ```text and satisfies linting and consistent rendering for the directory
tree visualization.
In `@crates/grida-canvas/examples/fixtures/l0_effects_progressive_blur.rs`:
- Around line 68-84: The backdrop blur card (the variable backdrop created with
effect_rect and LayerEffects.backdrop_blur) has no pixels behind it because
flat_scene sets background_color to None and the card doesn't overlap earlier
nodes; make the test meaningful by placing patterned content or another node
directly behind or overlapping the backdrop card. Modify the scene setup used by
flat_scene (or the vec passed into flat_scene) so that the backdrop node named
backdrop overlaps an earlier node (e.g., move or duplicate
layer_tl_br/layer_top_bottom/layer_both behind it) or add a patterned background
node before backdrop so FeBackdropBlur on backdrop actually samples pixels
behind it. Ensure the change keeps the backdrop variable and effect_rect usage
so the progressive-blur path is exercised.
In `@crates/grida-canvas/examples/tool_gen_cover.rs`:
- Around line 29-31: The test currently writes the serialized scene with
fixture_helpers::write_multi_fixture(&scenes, "cover") but then renders the
original scene variable, which misses encode/decode regressions; after calling
fixture_helpers::write_multi_fixture (or using the bytes it returns if
available), read back the just-written .grida data and decode it into a scene
instance and use that decoded scene for subsequent rendering (replace uses of
the original scene variable when generating cover.png), and make the same change
for the second occurrence around the area noted (line ~46) so both renders use
the round-tripped scene.
- Around line 48-50: Refactor the unsafe borrow of the surface out of the
example by adding a safe helper on Renderer (e.g., a with_surface_canvas or
with_surface<R, F>(&mut self, f: F) -> R where F: FnOnce(&mut Canvas) -> R) that
internally does the backend.get_surface() unsafe borrow, calls the closure with
the surface.canvas(), and releases the borrow; then update the example to call
renderer.with_surface_canvas(|canvas| renderer.render_to_canvas(canvas, width,
height)) (or alternatively move render_to_canvas into a &mut self API that uses
the same helper). Follow the pattern already used by Renderer::snapshot() and
Renderer::canvas() to own the surface-borrow lifecycle and avoid exposing unsafe
backend.get_surface() in examples.
In `@crates/grida-canvas/examples/tool_io_grida.rs`:
- Around line 120-184: The inspect_json function does not handle CLI flags
related to scene selection like --scene or --list-scenes, causing inconsistent
CLI behavior with legacy JSON input. To fix this, add checks for these flags in
inspect_json and return an explicit error message if they are used with legacy
JSON input to prevent silent ignoring. This ensures the CLI behavior remains
predictable when processing legacy grida1 files.
In `@crates/grida-canvas/src/io/io_grida_file.rs`:
- Around line 149-154: The code currently trusts zip::ZipFile::size() and
allocates/reads unbounded (doc, fbs_bytes, read_to_end), so add a hard
extraction limit (e.g. MAX_UNCOMPRESSED_SIZE constant) and enforce it before
allocation and while reading: check doc.size() (and treat missing/zero/too-large
as error), cap allocation to min(doc.size(), MAX_UNCOMPRESSED_SIZE) and replace
the unbounded read_to_end with a bounded reader (e.g.
doc.take(MAX_UNCOMPRESSED_SIZE) or iterative reads that track total bytes) and
return a DecodeError::Zip if the file exceeds MAX_UNCOMPRESSED_SIZE or the
announced size is absurd; reference the variables/functions doc, fbs_bytes,
archive, ZipFile::size(), and read_to_end() when making the changes.
In `@crates/grida-canvas/src/layout/engine.rs`:
- Around line 103-108: build_taffy_subtree currently returns None when
encountering non-Taffy roots (e.g., Group/BooleanOperation), which prevents
visiting children and denies Taffy layout to Taffy-capable descendants; change
build_taffy_subtree so that when a node is not itself Taffy-participating it
does not immediately return None but instead recurses into its children (calling
build_taffy_subtree for each child) to locate and construct Taffy subtrees for
descendants, merging any found subtrees into the result map; keep
extract_all_layouts usage for schema fallbacks but ensure build_taffy_subtree
still visits children of non-Taffy nodes so Containers and other Taffy-capable
nodes receive proper Taffy layout rather than only schema fallbacks.
---
Nitpick comments:
In `@crates/grida-canvas/examples/fixtures/l0_layout_transform.rs`:
- Around line 214-249: The fixture uses raw numeric IDs in pairs, links, and the
roots vector which duplicates and risks breakage; replace these bare integers
with named NodeId values (e.g., consts or a small NodeId allocator) and use the
NodeId type everywhere build_scene is called so the same identifier is declared
once and reused; update the pairs Vec<(u64, Node)> to Vec<(NodeId, Node)> (or
cast/construct NodeId) and change the HashMap keys and roots Vec to the NodeId
type, and ensure container keys (13,15,17) and their child references (14,16,18)
are replaced by those named NodeId bindings so parent/child wiring is only
defined once and remains stable when nodes are inserted or renumbered while
still calling build_scene("L0 Layout Transform", None, pairs, links, roots).
In `@crates/grida-canvas/src/io/io_grida_file.rs`:
- Around line 91-134: Add an impl of std::error::Error for DecodeError so
callers can access error sources; implement fn source(&self) -> Option<&(dyn
std::error::Error + 'static)> that returns Some(e) for the Fbs variant
(DecodeError::Fbs(e)) and None for the other variants (UnrecognizedFormat, Zip,
Json, Conversion, JsonNotSupported, NoScenes); ensure
io_grida_fbs::FbsDecodeError implements Error (it already should) and import
std::error::Error if needed.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 845495a5-8e62-4c5b-864d-7385a8ef21f9
⛔ Files ignored due to path filters (1)
crates/grida-canvas/cover.pngis excluded by!**/*.png
📒 Files selected for processing (21)
.agents/skills/research/SKILL.mdcrates/grida-canvas/AGENTS.mdcrates/grida-canvas/Cargo.tomlcrates/grida-canvas/examples/fixtures/README.mdcrates/grida-canvas/examples/fixtures/cover.rscrates/grida-canvas/examples/fixtures/l0_effects_progressive_blur.rscrates/grida-canvas/examples/fixtures/l0_layout_transform.rscrates/grida-canvas/examples/fixtures/l0_strokes_varwidth.rscrates/grida-canvas/examples/fixtures/mod.rscrates/grida-canvas/examples/tool_gen_bench_fixture.rscrates/grida-canvas/examples/tool_gen_cover.rscrates/grida-canvas/examples/tool_gen_fixtures.rscrates/grida-canvas/examples/tool_io_grida.rscrates/grida-canvas/src/io/io_grida_fbs.rscrates/grida-canvas/src/io/io_grida_file.rscrates/grida-canvas/src/io/mod.rscrates/grida-canvas/src/layout/engine.rscrates/grida-dev/src/grida_file.rsfixtures/test-grida/L0.gridafixtures/test-grida/bench.gridafixtures/test-grida/cover.grida
Comment on lines
+29
to
+31
| // ── 2. Write .grida file ──────────────────────────────────────────── | ||
| let scenes: Vec<(&str, _)> = vec![("cover", scene.clone())]; | ||
| fixture_helpers::write_multi_fixture(&scenes, "cover"); |
There was a problem hiding this comment.
Render the serialized round-trip, not the original scene.
fixture_helpers::write_multi_fixture() only verifies that decoding succeeds and that scene names match (crates/grida-canvas/examples/fixture_helpers/mod.rs:127-157). Because this tool renders the original scene afterward, encode/decode regressions can still produce a “good” cover.png while cover.grida is wrong. Load the just-written bytes/file back and render that scene instead.
Also applies to: 46-46
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@crates/grida-canvas/examples/tool_gen_cover.rs` around lines 29 - 31, The
test currently writes the serialized scene with
fixture_helpers::write_multi_fixture(&scenes, "cover") but then renders the
original scene variable, which misses encode/decode regressions; after calling
fixture_helpers::write_multi_fixture (or using the bytes it returns if
available), read back the just-written .grida data and decode it into a scene
instance and use that decoded scene for subsequent rendering (replace uses of
the original scene variable when generating cover.png), and make the same change
for the second occurrence around the area noted (line ~46) so both renders use
the round-tripped scene.
Comment on lines
+48
to
+50
| let surface = unsafe { &mut *renderer.backend.get_surface() }; | ||
| let canvas = surface.canvas(); | ||
| renderer.render_to_canvas(canvas, width, height); |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "Renderer / backend surface APIs"
rg -n -C3 '\bfn (render_to_canvas|get_surface|render)\b' crates/grida-canvas/src crates/grida-canvas/examples
echo
echo "Direct get_surface call sites"
rg -n -C2 '\bget_surface\s*\(' crates/grida-canvas/src crates/grida-canvas/examplesRepository: gridaco/grida
Length of output: 8542
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Get the full implementation of render_to_canvas
echo "=== render_to_canvas implementation ==="
sed -n '1468,1500p' crates/grida-canvas/src/runtime/scene.rs
echo
echo "=== Methods called by render_to_canvas (frame, paint_frame, etc.) ==="
# Look for the frame method
sed -n '1400,1468p' crates/grida-canvas/src/runtime/scene.rs | tail -n 70Repository: gridaco/grida
Length of output: 4203
Encapsulate the unsafe surface access into a safe Renderer API.
The unsafe pattern bypasses the type system unnecessarily. While render_to_canvas currently only takes an immutable self borrow and doesn't access backend, this API design is brittle—future changes to the method could reintroduce unsafe aliasing bugs. Since the same pattern is already used internally in Renderer::snapshot() and Renderer::canvas(), extract this into a safe helper method that owns the full surface borrow lifecycle. This allows the examples to remain safe without relying on external unsafe code and defensive understanding of internal implementation details.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@crates/grida-canvas/examples/tool_gen_cover.rs` around lines 48 - 50,
Refactor the unsafe borrow of the surface out of the example by adding a safe
helper on Renderer (e.g., a with_surface_canvas or with_surface<R, F>(&mut self,
f: F) -> R where F: FnOnce(&mut Canvas) -> R) that internally does the
backend.get_surface() unsafe borrow, calls the closure with the
surface.canvas(), and releases the borrow; then update the example to call
renderer.with_surface_canvas(|canvas| renderer.render_to_canvas(canvas, width,
height)) (or alternatively move render_to_canvas into a &mut self API that uses
the same helper). Follow the pattern already used by Renderer::snapshot() and
Renderer::canvas() to own the surface-borrow lifecycle and avoid exposing unsafe
backend.get_surface() in examples.
- Updated `inspect_json` function to reject unsupported options for legacy JSON files, preventing misuse. - Enhanced layout engine to compute layouts for Taffy subtrees discovered under non-Taffy parents, ensuring proper layout for nested structures. - Introduced a hard limit on the uncompressed size of `document.grida` in ZIP archives to protect against zip bombs and excessive memory allocation. - Improved documentation and comments for clarity on new features and error handling.
There was a problem hiding this comment.
🧹 Nitpick comments (1)
crates/grida-canvas/src/io/io_grida_file.rs (1)
204-211: Consider avoiding allocation for RawFbs path (optional optimization).
resolve_fbs_bytescopies the entire input forRawFbsformat (bytes.to_vec()), which is unnecessary since FlatBuffers can verify and access data directly from a borrowed slice.This could be optimized by returning
Cow<'_, [u8]>or having the FBS decoder work directly with the input slice. However, this is a minor performance consideration for file loading—not blocking.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/src/io/io_grida_file.rs` around lines 204 - 211, The function resolve_fbs_bytes currently clones the input slice for the RawFbs branch (bytes.to_vec()), which is unnecessary; update either resolve_fbs_bytes to return a borrowed-friendly type (e.g., Cow<'_, [u8]>) or change the FBS decoder to accept a borrowed &[u8] so you can return a borrowed slice for Format::RawFbs instead of allocating; modify resolve_fbs_bytes signature and callers accordingly (refer to resolve_fbs_bytes, detect, Format::RawFbs, and DecodeError) to avoid the allocation while preserving the existing Zip/Json/Unknown behavior.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@crates/grida-canvas/src/io/io_grida_file.rs`:
- Around line 204-211: The function resolve_fbs_bytes currently clones the input
slice for the RawFbs branch (bytes.to_vec()), which is unnecessary; update
either resolve_fbs_bytes to return a borrowed-friendly type (e.g., Cow<'_,
[u8]>) or change the FBS decoder to accept a borrowed &[u8] so you can return a
borrowed slice for Format::RawFbs instead of allocating; modify
resolve_fbs_bytes signature and callers accordingly (refer to resolve_fbs_bytes,
detect, Format::RawFbs, and DecodeError) to avoid the allocation while
preserving the existing Zip/Json/Unknown behavior.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 80c3c30e-d619-48df-9e10-323b388d89e8
📒 Files selected for processing (5)
.agents/skills/research/SKILL.mdcrates/grida-canvas/examples/fixtures/l0_effects_progressive_blur.rscrates/grida-canvas/examples/tool_io_grida.rscrates/grida-canvas/src/io/io_grida_file.rscrates/grida-canvas/src/layout/engine.rs
🚧 Files skipped from review as they are similar to previous changes (2)
- crates/grida-canvas/examples/fixtures/l0_effects_progressive_blur.rs
- .agents/skills/research/SKILL.md
This was referenced Mar 28, 2026
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
A batch of improvements to the
cg(grida-canvas) crate spanning file I/O, layout extraction, FlatBuffers decoding, and example fixtures.File I/O & Inspector (
tool_io_grida)tool_io_gridafrom a basic validator into a comprehensive file inspector supporting.grida(FlatBuffers),.grida(ZIP archive), and legacy.grida1(JSON) formats with auto-detection.--list-scenes,--scene <N>,--verbose(full node tree),--layout-check(diagnoses "Container must have layout result" panics).io_grida_filemodule for unified file format detection and decoding.zipcrate dependency (deflate-only, pure Rust, wasm-safe) for ZIP archive support.Layout Engine Fix
BooleanOperation,Group) — previously these could cause panics inGeometryCachebecause descendant containers had no layout results.FBS Decoder Enhancements
DiamondGradientPaintdecoding support.TextDecorationLineandTextDecorationStyledecoding for richer text styling.New Example Fixtures
cover.rs— 1600x900 hero-image scene showcasing gradients, shadows, blend modes, corner smoothing, and typography, with a correspondingtool_gen_coverexample andcover.gridaoutput.l0_effects_progressive_blur— Progressive blur (layer + backdrop) with varying gradient directions.l0_strokes_varwidth— Variable-width stroke profiles (taper, reverse taper, bulge, multi-stop, uniform).l0_layout_transform— Expanded from 3 to 16 test cases covering rotation at key angles, custom transform origins, nested container composition, and clipped rotated containers.README.mdwith descriptions for all new scenes.tool_gen_bench_fixturefor generating benchmark.gridafiles.Agent/Skill Documentation
.agents/skills/research/SKILL.md— research workflow for investigating upstream projects (Chromium, Skia, Servo) before writing code.crates/grida-canvas/AGENTS.mdto reflect the expandedtool_io_gridacapabilities.Housekeeping
.gridafile reading logic fromgrida-devinto the newio_grida_filemodule, reducing duplication..gridafixture files:L0.grida,bench.grida,cover.grida.Summary by CodeRabbit
New Features
Bug Fixes
Documentation
Chores