Repository navigation
Conversation
Updates the requirements on [pydata-sphinx-theme](https://github.com/pydata/pydata-sphinx-theme) to permit the latest version. - [Release notes](https://github.com/pydata/pydata-sphinx-theme/releases) - [Changelog](https://github.com/pydata/pydata-sphinx-theme/blob/main/RELEASE.md) - [Commits](pydata/pydata-sphinx-theme@v0.16.1...v0.23.0) --- updated-dependencies: - dependency-name: pydata-sphinx-theme dependency-version: 0.23.0 dependency-type: direct:development ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
…NTRIBUTE NEW DATA template Data contributions become issue-first: one issue carries one or more validated archives (one graph each) plus motivation, and maintainers do all repo work. The standalone grammar template is removed because a grammar is a representation of a query and must ship inside an archive with reference results. The new template makes the four contribution cases explicit, requires per-archive validation output and SHA256, mandates motivation, and states the consistency invariants (no graph without a query, no query without results.mtx) and the declined cases. README and skill pointers are updated to the new template.
Data contributions are issue-first, so the new_graph and new_grammar PR templates — duplicates of the issue template's fields — are removed. A generic PR template covers code contributions and points data contributors to the CONTRIBUTE NEW DATA issue; docs/dataset.rst now says grammars arrive together with the data they apply to.
…contribution guide The old 'How to add a new graph?' section told contributors to open a PR, which contradicts the issue-first flow. The new 'Contributing data' section documents the flow (archives -> validation -> one issue -> maintainers), the four contribution cases, the per-archive validation command, the consistency invariants (no graph without a query, no query without reference results, no item without description and motivation), and the declined cases. Subsection underlines use the file's existing '^' style.
…m an archive Accepting a contributed archive requires a reachable_pairs.csv row per new query (generate_registry reads the graph list and categories from the CSV, and merge_archive refuses to merge without the rows). Typing those rows by hand is error-prone, so the tool derives them from a validated archive: one row per queries/<class>/<query>/ directory, with the query directory name as the grammar stem (representation picked cnf > re > mcfg > rsm, the alphabetically first file per extension) and the results.mtx declared entry count as num_reachable_pairs. The archive must pass the structure checker first, a tarball must be named after the graph, and the CSV's (graph, grammar) sort order is preserved; duplicate rows and category disagreements are reported without writing. --dry-run prints instead of writing.
The add-graph skill gains an 'Accepting an external contribution' section: per archive — download from the Drive link, verify the SHA256 against the issue, re-run the structure checker, check name uniqueness against the registry and the reachable-pairs CSV, derive the CSV rows with utils/add_reachable_pairs.py, then upload (full) or merge (partial); once per issue — regenerate the registry, do the docs wiring, and refresh the reachable-pairs tables and archive sizes. Stale 'PR template' references in both skills point at the issue template now, and add-grammar records that grammars arrive via cases 2/3/4b of a CONTRIBUTE NEW DATA issue.
Issue-first data contributions (unified CONTRIBUTE NEW DATA template, multi-archive issues, motivation and reference-results requirements), the new utils/add_reachable_pairs.py, and the removal of the specialized graph and grammar templates are recorded under Unreleased. Closes #164
…and declined cases Code review found the four-case list and the declined-cases list duplicated near-verbatim between the issue template and the docs guide. Per the single-source-of-truth convention (templates own field and case content), the guide now summarizes both in one sentence each and points at the template's 'Contribution type' and 'Not accepted via this flow' sections; the guide keeps what is uniquely its own — the flow, the validation command, and the consistency invariants.
Code review found the grammar-instantiation rule stated twice inside the template (Reference results and Queries and grammars case 2). The Reference results section now keeps only the graph-side rule, noting that the structure checker enforces it; the grammar-side rule stays with case 2.
The CONTRIBUTE NEW DATA template does not exist on master until a release lands, so blob/master links to it would fail the CI link check on dev. The links now point at blob/dev, where the file exists and which is also the version contributors should follow; the links keep resolving after release.
Move utils/check_archive_structure.py into flpq_data/tools/ and register the flpq-check-archive console entry point, so external contributors validate archives with pip install flpq-data alone instead of downloading the script from the repo. The S3 client factory and bucket constants move to flpq_data/dataset/s3.py (boto3 imported lazily — it is a dev-only dependency), and --audit reports a clean error when boto3 is missing. The utils importers now use the package path; upload_to_s3 keeps re-exporting the moved names so existing imports work unchanged. The module now counts toward the 95% coverage gate, which exposed 28 never-tested branch edges. Sixteen new tests cover the real defensive branches (malformed MTX headers and entry lines, nnz mismatches, blank body lines, unexpected queries/ entries, missing queries README, subdirectories in query dirs, forbidden label components, same-directory terminal-set consistency, multi-top-level tarballs, broken symlinks, the no-path CLI). Two provably dead constructs are removed: the indexed- reversed label check (every stem it matches also matches the reverse regex, so it could never fire) and the __main__ guard (replaced by the console entry point; no other package module has one). Shared archive-tree builders move to tests/archive_helpers.py with a root tests/conftest.py putting tests/ on sys.path; the four utils test files import from there. New tests also pin the missing-boto3 --audit branch and the declared console entry point via importlib.metadata.
The checker is now part of the installed package, so every documented invocation becomes the flpq-check-archive console command: the contributor recipe in the issue template and the Graphs page is now pip install flpq-data (no script download), and the maintainer steps in the add-graph skill use the command too. The "Archive structure" section moves from docs/utils.rst to a new docs/reference/tools/index.rst reference page (automodule + autosummary of the checker's public API, generated stubs committed), carrying the archive_structure anchor so existing :ref: links keep resolving. The Labels check bullet drops the indexed-reversed ordering clause — the checker never enforced it (the code path was dead and is removed in S1). AGENTS.md points at the new module path and gains a flpq_data/tools/ layout line; the changelog records the command and the move. Closes #165
The S1 commit picked up a locally regenerated uv.lock (the pre-existing dependabot #162 specifier drift plus a lockfile revision bump from a newer local uv). That change is out of scope for this task; restore the lock to the dev state.
Bump the version to 6.1.0 and promote the [Unreleased] changelog section to [6.1.0] - 2026-10-14. The release is minor: everything since 6.0.0 is additive (the flpq-check-archive console command, the flpq_data.tools and flpq_data.dataset.s3 modules, utils/add_reachable_pairs.py) with no changes to the existing public API, and the dataset prefix stays at 6.0.0. uv.lock is re-locked in the same commit: it must track the new package version for CI's 'uv sync --frozen', and it also picks up the pydata-sphinx-theme specifier that #162 changed in pyproject.toml without re-locking, which left the lockfile stale.
The 'How to contribute new data?' heading added by #164 has an underline one character shorter than the title, which is invalid RST. twine check (the same validation the Publish workflow runs before the TestPyPI pre-publish) rejects both dists with 'Title underline too short', so no release could be published until the underline matches the title.
The assertion flpq_data.__version__ == "6.0.0" hardcodes the release version and fails on every bump — it broke the 6.1.0 release commit on all CI platforms. Assert against the canonical VERSION from flpq_data.config instead: the wiring of __version__ to config is still checked, while the dataset-prefix literals stay pinned because they move only on a major release.
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.
Merge
devintomasterfor the 6.1.0 release (minor: no changes to the existing public API, dataset prefix stays at 6.0.0).What this carries:
flpq-check-archiveconsole command (flpq_data/tools/, issue Include the archive-checking tooling in flpq-data #165), so contributors validate archives withpip install flpq-dataalone.utils/add_reachable_pairs.py: derives the reachable-pairs CSV rows of one validated archive (issue Design external users friendly way to contribute new data #164).flpq_data.dataset.s3.6.1.0inconfig.py/pyproject.toml, the changelog promoted from[Unreleased]to[6.1.0] - 2026-10-14, anduv.lockre-locked (it also picks up the pydata-sphinx-theme specifier that chore(deps-dev): update pydata-sphinx-theme requirement from <0.22.0,>=0.16.1 to >=0.16.1,<0.24.0 #162 left stale).After this merges,
v6.1.0is tagged onmaster; the Publish workflow then builds the sdist/wheel, publishes them to PyPI (Trusted Publishing), and creates the GitHub Release with the[6.1.0]changelog section. The PR's Publish check pre-publishes to TestPyPI as a packaging validation for exactly this code.Task issues #164 and #165 close via their
Closes #Ncommit references when this lands on master.