Resync with graphify-py @ 92e682f (v0.9.0)#31
Merged
Conversation
Port the applicable feature, security, and bug-fix changes from
`graphify-py` v0.8.49 -> v0.9.0 (38 upstream commits) and bump the
workspace to `0.9.0`.
Ports:
- Node IDs are now the full repo-relative path with the extension
dropped and separators collapsed to `_`, so same-named files in
different directories stay distinct (#1504, #1509). Legacy graphs
read as-is with a one-line rebuild nudge.
- Cross-file type-reference stubs are sourceless and carry an
`origin_file` so same-label stubs from different files do not merge
(#1462, #1500).
- New C# cross-file type resolver plus `enum`/`struct`/`record` node
types (#1466).
- Java `enum`/annotation node types, generic-argument and field-type
references (#1512, #1510, #1485, #1487).
- Objective-C header routing, generic property types, class-method
selectors, and `@import` handling (#1475).
- New WPF/XAML extractor: `x:Class`, named controls, `{Binding}`
paths, and ViewModel / code-behind resolution (#1460, #1473).
- Vue single-file-component extraction and `.metal` routing through
the C++ extractor (#1468, #1480).
- Obsidian export only overwrites notes it owns via a manifest, the
Canvas uses a `ceil(sqrt(n))` grid, and note/canvas filenames are
de-duplicated case-insensitively (#1506, #1452, #1453).
- Wiki articles link with portable relative Markdown links instead of
`[[wikilinks]]`, with case-folded slug dedup (#1444, #1453).
- LLM requests send `stream: false`, and `kimi`/`gemini`/`deepseek`
honour a bare `*_BASE_URL` (#1223, #1458).
- CLI `--timing` on `extract`/`cluster-only`/`label`, and
`--missing-only` on `cluster-only`/`label` (#1490, #1481).
- `reflect --if-stale` also weighs the graph sidecars, the Read|Glob
hook matches a file's real trailing extension, and `affected` /
`explain` resolve a source-file path to its file-level node
(#1470, #1463, #1503).
Glory to the Omnissiah
Resolve the 24 findings from the first CodeRabbit pass on the resync.
Genuine divergences from graphify-py are fixed; findings that match
graphify-py are kept and the rationale recorded in an in-code comment
(CodeRabbit reads code, not commit messages).
Fixes:
- `extract()` now clears the thread-local XAML `ViewModel` class cache
per run (generation-gated across the rayon pool), matching graphify-py
`_XAML_CSHARP_CLASS_CACHE.clear()`; a repeated in-process run re-scans
`.cs` instead of serving stale members. Adds a regression test.
- Obsidian manifest write propagates IO errors instead of swallowing
them, like the sibling note writes and graphify-py `.write_text`.
- `ingest` C-family map includes `.cxx` (extracted as C++ and present in
`analyze`'s family map; graphify-py's build map omits it).
- Legacy-id alias registration is skipped unless the node id shares the
new stem, so a mismatched prefix no longer maps unrelated edges.
- `collect_cs_files` uses `entry.file_type()` so a symlinked directory
cannot loop forever.
- `is_objc_header` reads only the 256 KiB it inspects.
- Vue dynamic-import recovery scans the masked source, so an `import()`
in `<template>`/`<style>` is not edged.
Kept as graphify-py parity (documented in code):
- Sourceless stub `source_location` stays `Some("")` (graphify-py emits
`""`); wiki resolver keys by label and reserves `index` only in the
resolver; C# merges a file's namespaces into one scope; the direct
`<Binding Path>` element emits no generated-member edge; validation
runs once before the semantic re-key.
Also strengthens the affected test assertions (exact set/count checks,
`origin_file` and canonical-id checks, real absolute/relative paths, raw
event-edge count, loud directory-walk errors).
By the will of the Machine God
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (7)
🚧 Files skipped from review as they are similar to previous changes (7)
📝 WalkthroughWalkthroughThis release bumps the workspace to 0.9.0, changes node IDs to full repo-relative stems, adds ChangesGraphify 0.9.0 release
Sequence Diagram(s)sequenceDiagram
participant extract_xaml
participant code_behind_resolver
participant viewmodel_resolver
participant cs_class_lookup
participant toolkit_extractor
participant XamlBuilder
extract_xaml->>code_behind_resolver: resolve x:Class and event handlers
extract_xaml->>viewmodel_resolver: detect DataContext / Prism / naming
viewmodel_resolver->>cs_class_lookup: scan *ViewModel classes
extract_xaml->>toolkit_extractor: emit ObservableProperty / RelayCommand members
extract_xaml->>XamlBuilder: deduplicate nodes and edges
Estimated code review effort🎯 5 (Critical) | ⏱️ ~150 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
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 |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
Resolve the findings from the second CodeRabbit pass. Genuine bugs and determinism gaps are fixed (several are bugs `graphify-py` shares, fixed per AGENTS.md); findings that faithfully mirror `graphify-py` are kept, with the parity reasoning recorded in code. Fixes: - `with_xaml_extract_root` restores the thread-local extract root through an RAII guard, so a panic can't leak it to a reused rayon worker (parity with graphify-py's `try`/`finally`). - `.xaml` joins the AST disk-cache bypass list: XAML resolution scans sibling `.cs` code-behind, so a content-keyed cache served stale `ViewModel` members across runs. - `origin_file` is relativized alongside `source_file` (extraction and build) so the #1462 stub field no longer leaks absolute machine paths into graph JSON. - objc and PowerShell `ensure_named_node` emit sourceless stubs (empty `source_file`, `origin_file` set) so cross-file references rewire onto real definitions instead of becoming phantom duplicates (#1402/#1462). - Obsidian `owned_write` refuses to follow a symlink at the target or a parent component, so a symlinked note can't escape the vault (#1506). - `find_node` matches a source path by its normalized full path rather than tokenized form, so distinct paths no longer collide. - `cluster-only --missing-only` preserves a malformed labels file instead of relabeling over it, matching the non-`--missing-only` path. - `#[must_use]` on the pure ID/stem helpers; corrected `extract` and `label` doc-comments. Kept as parity (documented in code): the objc compound-selector node-ID form, C# nested-namespace registration deferral, and the wiki/`index_md` god-node list all mirror `graphify-py` exactly. Test strengthening: `lessons_fresh` now covers both graph sidecars (matching the Python parametrization), the trailing-slash seed test exercises the re-keyed file node, and the affected test suites return `Result` with `?` per the repo convention. By the will of the Machine God
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
A `src\foo.rs` query reached the `source_exact` tier (its `query_path` normalizes `\` to `/`), but the L1 file-node preference never fired. `query_basename` came from `Path::file_name` on the raw label, and on Unix `\` is not a path separator, so the basename kept the whole backslash string instead of `foo.rs` — so the file node never floated ahead of its symbols. Derive both `query_basename` and `query_path` from one slash-normalized string so Windows-style path queries match the forward-slash behavior (#1503). The `#[must_use]` nitpick on `find_node` was already satisfied; no change there. Ave Deus Mechanicus
Three CodeRabbit findings (C# nested namespaces, ObjC selector colon collapse, wiki `index_md` god-node list) were re-raised. Verified each against `graphify-py @ 92e682f`: all faithfully mirror the reference, none is a still-valid bug, and "fixing" any would break byte-identical parity. No behavioral change — only inline parity notes: - ObjC (`objc.rs`): graphify-py joins selector identifiers colon-less (`extract.py:9623`/`:9657`), pinned by `objc_compound_selector_call_ resolves`. Firmed the comment to name the deliberate collision-risk trade-off (`setFoo:bar:`/`setFoobar:` collapse); preserving `:` would re-key every keyword selector and break parity. - Wiki (`generate.rs`): `index_md` intentionally gets the full `god_nodes_data` (`wiki.py:333`); a node absent from the graph is rendered as plain text by `md_link`, not a broken link, so filtering would diverge from byte-identical `index.md`. Documented at the call site. The C# site was already documented (`csharp.py:106-111`, flat `own_ns` with `# len > 1: skip`); left unchanged. By the will of the Machine God
`cluster-only --no-label --missing-only` with an existing `.graphify_labels.json` wiped every community to a `Community N` placeholder, clobbering hand-curated names. The `--no-label` branch fired before the `--missing-only` handling and never loaded the existing file. graphify-py preserves here: `__main__.py:3489` loads existing labels for any `exists && !force_relabel` case, with no `missing_only` exclusion. The Rust port had added `&& !opts.missing_only` to that branch, dropping `--no-label --missing-only` into the placeholder path. Fire the preserve branch whenever no LLM naming will happen (`!missing_only || no_label`), since `--no-label` forbids any LLM call. Curated labels survive; only true gaps fall back to placeholders. Adds a `cluster-only` regression test. By the will of the Machine God
Address the latest CodeRabbit findings (verified against `graphify-py @ 92e682f`): - `cluster-only` marked the `export` stage before the HTML render, so `--timing` excluded `to_html` time. graphify-py marks it after (`__main__.py:3555`); moved the mark past the render to match. - Strengthened the label-preservation tests to assert community 0 still maps to `Authentication` by key (parse + `["0"]`), not a bare substring; added a concrete per-stage line to the timing test. Adds `serde_json` as a dev-dependency for the structural assertion. Skipped (parity): the `--no-label` branch keeps `&& !force_relabel` — `label` does not accept `--no-label`, so the combination is unreachable, and the guard matches graphify-py `__main__.py:3494` (`elif no_label and not force_relabel`). By the will of the Machine God
rblaine95
force-pushed
the
resync/graphify-py/92e682f
branch
from
7985376 to
6bf2f0d
Compare
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.
Ports the applicable changes from
graphify-pyv0.8.49 → v0.9.0 (38 upstream commits) and bumps the workspace to0.9.0. Submodule pointer advanced to92e682f.Breaking change
_), e.g.src/auth/session.py→src_auth_session,validate()inside →src_auth_session_validate(#1504/#1509). Same-named files in different directories no longer collide. Abuild_from_jsonre-key migrates pre-#1504 graphs; read-only consumers (query/serve) print a one-line rebuild nudge instead of silently re-keying.Ported
origin_fileso same-label stubs from different files stay distinct (#1462/#1500).enum/struct/recordnode types (#1466).@import(#1475).x:Class, named controls,{Binding}paths, ViewModel / code-behind resolution (#1460/#1473)..metalrouted through the C++ extractor (#1468/#1480).ceil(sqrt(n))grid, case-fold filename dedup (#1506/#1452/#1453); wiki portable relative-Markdown links (#1444).stream: false;kimi/gemini/deepseekhonour a bare*_BASE_URL(#1223/#1458).--timingonextract/cluster-only/label;--missing-onlyoncluster-only/label(#1490/#1481).--if-staleweighs graph sidecars; the Read|Glob hook matches the real trailing extension;affected/explainresolve a source-file path to its file-level node (#1470/#1463/#1503).Docs (
README.md,USAGE.md) reconciled: new flags, formats, env vars, node-ID scheme, and export behaviours. Parity tests ported for every new upstream test case.Verification
cargo clippy --all-targets --all-features --workspace— cleancargo nextest run --workspace— 2283 passed, 2 skippedhk check— pass--type committed—findings: 0(24 first-pass findings resolved: genuine divergences fixed, graphify-py-matching ones kept with in-code// Parity dispute (CodeRabbit): …comments)Summary by CodeRabbit
--timingand--missing-only; expanded backend base-URL env overrides (KIMI/GEMINI/DEEPSEEK).