ENH: per-directory files for the directory index-location map - #772
Merged
Conversation
Replace the single shared JSON index-location map with one small JSON file per data directory under a cache directory. This is the simplified successor to the SQLite approach in #764. Different directories use different files, so concurrent writers to distinct directories cannot lose each other's entries, and a corrupt entry only affects its own directory and self-heals on the next write (see #508) -- with no locks, corruption-recovery protocol, or at-fork hooks. Writes swap a sibling temp file into place, so readers never see a half-written entry. The read-only-directory fallback derives a deterministic index name and needs no map entry at all; only a user-specified custom index path is recorded. Config option `directory_index_map_path` becomes `directory_index_map_dir` (a directory). The map is a disposable cache; any old shared `cache_paths.json` is ignored and left in place.
…ling - _get_mapped_index_path no longer unlinks corrupt entries on read (that could delete a concurrent writer's repair); a miss self-heals on the next atomic write instead. - Treat non-UTF-8 / malformed JSON and non-string/empty index_path payloads as misses (catch ValueError; validate shape). - Use the full sha256 via os.fsencode in _path_digest: avoids a read-only index-file-name collision and handles non-UTF-8 path bytes.
4 tasks
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughChangesThe index-location map now uses per-directory JSON entries under Index map refactor
Possibly related PRs
Suggested labels: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
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 |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## dev #772 +/- ##
=========================================
Coverage 100.00% 100.00%
=========================================
Files 164 164
Lines 17357 17356 -1
=========================================
- Hits 17357 17356 -1
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
d-chambers
commented
Jul 24, 2026
| file) and non-UTF-8 filename bytes digest without error. | ||
| """ | ||
| Return a fresh dict of index locations read from disk. | ||
| return hashlib.sha256(os.fsencode(path)).hexdigest() |
Contributor
Author
There was a problem hiding this comment.
We may also need to support URLs in the future. Please just use fsencode if possible, fallback to string only hash if not.
Contributor
Author
There was a problem hiding this comment.
Done in 34c4bb1: _path_digest now tries os.fsencode first (exact local-path bytes) and falls back to a plain string encoding for inputs it rejects — e.g. a remote URL/UPath directory whose __fspath__ isn't available — instead of raising. Added a regression covering the fallback branch.
…nput Prefer os.fsencode for exact local-path digests, but fall back to a plain string encoding for inputs it rejects (e.g. remote URL/UPath directories we may support later) instead of raising.
This was referenced Jul 24, 2026
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.
Description
Simplified successor to #764: it replaces the shared JSON directory index-location map with one small JSON file per data directory, instead of the SQLite database (with corruption-recovery locks and
os.register_at_forkhooks) that #764 proposed.The index-location map records where a directory's index lives when the data directory itself is read-only, or when a user supplies a custom index path. Its goals are (a) concurrent writers to distinct data directories must not lose each other's entries, and (b) a corrupt entry (see #508) must not break indexing.
What changed
<sha256(dir)>.json) under a cache directory, replacing the single sharedcache_paths.json. Different directories use different files, so concurrent writers to distinct directories cannot clobber one another.directory_index_map_pathbecomesdirectory_index_map_dir(now a directory). The map is a disposable cache; any old sharedcache_paths.jsonis ignored and left in place.Why not the SQLite approach from #764
For a disposable convenience cache whose only production reads/writes are single-key, per-entry files give the same lost-update and corruption guarantees by construction, with ~10% of the code and none of the process-wide side effects (the
os.register_at_forkhooks in #764 serialize every fork in the host process through a library lock). It is also friendlier to NFS home directories, where SQLite locking is unreliable.Supersedes #764. Part of the free-threading work extracted from #763.
Validation
tests/test_io/test_indexer.py,tests/test_io/test_index/,tests/test_utils/test_config.py: pass.tests/test_io/: 2053 passed, 63 skipped, 1 xfailed.prek/pre-commit hooks pass on changed files.Changelog
directory_index_map_dirinstead of a sharedcache_paths.json, so a corrupt entry only affects its own directory. Any oldcache_paths.jsonis ignored and left in place.Checklist
I have (if applicable):
Summary by CodeRabbit
New Features
API Changes
directory_index_map_pathsetting withdirectory_index_map_dir.Bug Fixes