chore(git): add /.worktrees/ to .gitignore - #71
Closed
ywatanabe1989 wants to merge 43 commits into
Closed
Conversation
…0) (#38) Rewrites test_anova test file: - Add AAA marker comments (# Arrange / # Act / # Assert) to every test function, in order, separated. - Split multi-assert tests (TQ007) into single-assert tests with descriptive names that read as one-line specifications. - Result: 22 STX-TQ002+TQ007 violations -> 0. - All 22 tests pass.
#39) * refactor(deps): migrate optional imports to try_import_optional (src/) Per the three-tier dependency policy (general/05_development/11_dependency-tiers.md), raw `try/except ImportError` for optional [all]-tier deps is forbidden ecosystem-wide; the sanctioned helper is `scitex_dev.try_import_optional` (centralizes install hints in `_HINTS` and gives readers a grep target for "this dep is optional"). Migrate every optional-dep import gate in src/ (src-only — no demos/tests/workflows): - torch (descriptive/_describe,_nan,_real,_circular): `HAS_TORCH = torch is not None` - figrecipe (_figrecipe_integration._AVAILABLE; _plot_helpers.get_palette) - scitex umbrella (_plot_helpers.compose_panels/ensure_figure) — no `extra=` because the scitex umbrella can't be a scitex-stats dependency (circular); keeps the hard-require ImportError raise Behavior preserved. Hard-dep gates (pandas/statsmodels/matplotlib/scitex-dev/ scitex-logging/mcp), the user-supplied dynamic import in introspect.py, and the fastmcp availability check in mcp.py are intentionally left as-is — they are not optional-dep violations. Re-does the valid kernel of closed #26 within the narrow scope the maintainer specified (src/ only); #26 itself was closed for overreaching into demos+workflows (broke 22+ demo tests). Demos already moved to examples/ via #35. Verified: 62 passed, 5 skipped (descriptive + plot_helpers + figrecipe + integration tests); all edited modules import cleanly; ruff clean. * test(integration): regenerate cross-package gate for try_import_optional migration The optional-imports migration changed src's static cross-package import set, so the PS-140 §2 gate in test_cross_package_imports.py went stale: - add `scitex_dev` — now imported (`from scitex_dev import try_import_optional`) - drop `scitex` — the umbrella import is now dynamic (`try_import_optional("scitex")`), no longer a static import Matches the delta scitex-dev 0.12.3 reports in CI (audit-project PS-140).
Add AAA markers (# Arrange/# Act/# Assert) to every test and split multi-assertion tests into single-assertion tests (faithful split, no coverage loss), per the PA-307 §3 STX-TQ002/TQ007 rules and the merged anova-test template. 69 tests -> 145 single-assert tests; all pass.
Add AAA markers, split multi-assertion tests into single-assertion tests (faithful split), and rename short test names to >=3 descriptive tokens (STX-TQ002/TQ003/TQ007). Class structure preserved; shared construction factored into small helper methods where it reduced duplication. 57 methods -> 97 single-assert tests; all pass; scitex-linter clean.
- sphinx CLI docs: `mcp installation` -> `mcp install` (deprecated command was renamed; old name prints a deprecation error) - skill CLI reference: same `mcp show-installation` -> `mcp install` - skill env-vars: add `SCITEX_STATS_CONFIG` to the table; drop the stale note claiming no SCITEX_STATS_* vars exist since the CLI precedence chain documents this variable
docs: refresh stale command names and env-var references
Each scitex ecosystem package owns its domain dataclasses. The Stats schemas (Stats, EffectSize, StatResult, GUI positioning/styling, etc.) have lived in the umbrella at scitex.io.bundle.kinds._stats._dataclasses since the schema reorg; that location is the umbrella's bundle dispatcher concern, not stats' domain concern. Move them here so scitex_stats is the single source of truth (umbrella, scitex_io optional provider, and any future GUI / editor tool all import from the same place). The umbrella ships a re-export shim at the same kinds/_stats path to keep its bundle loader/saver/Bundle classes working unchanged.
Mirrors figrecipe's .plt.zip / .fig.zip pattern: a single-file zipped form is the portable target, while the legacy directory bundle stays supported for iteration. save_stats_bundle / load_stats_bundle now branch on the suffix: *.stats.zip → zipped bundle (recommended) anything else → directory bundle (legacy) Both write the same canonical layout: <bundle>/stats.json # spec <bundle>/data.csv # supplementary data (optional) <bundle>/report.md # human summary (optional) This is what scitex_io's optional provider will register so that 'scitex_io.save(stats_dict, "results.stats.zip")' round-trips through the same .stats.zip API.
…202) PS-202 was firing because src/scitex_stats/_dataclasses/ (the new location of the Stats schema after the umbrella move) had no matching tests/scitex_stats/_dataclasses/ mirror. Adds test__Stats.py with: - Parametrized presence test covering every name in __all__ (STATS_VERSION, type aliases, GUI classes, core dataclasses). - STATS_VERSION shape check (semver-ish 3-part dotted string). - Stats is exposed as a real dataclass (not a stub/alias). - DataRef dict round-trip preserves path and columns. Also fixes PS-140: scitex umbrella was missing from the cross-package import gate. The umbrella imports scitex_stats.io via its scitex.io.bundle._core dispatcher, so it's a legitimate cross-package dependency. Plus pre-existing STX-TQ001/TQ002 lint debt on the parametrized test_cross_package_import function — added the required AAA markers and a real assertion on the importorskip return value.
CI (scitex-dev v0.14.0) reports 'scitex' as stale in the gate — the gate must reflect actual src/ imports, and scitex_stats does not import the scitex umbrella anywhere (correct — scitex is downstream). My earlier addition was based on local v0.12.4 reporting 'scitex' as missing; v0.14 fixed that auto-gen bug. CI is authoritative.
…SOC R4)
scitex_stats was hard-importing private figrecipe modules
(figrecipe._wrappers, figrecipe.styles._internal,
figrecipe._diagram), which made figrecipe a de-facto required
dep instead of optional. Per SOC.md R4 ("scitex-stats knows
figrecipe but as optional"), route these through
scitex_dev.try_import_optional with extra='figrecipe' so
scitex-stats installs and runs cleanly without figrecipe.
- _plot_helpers.violin_swarm: gate 3 figrecipe imports
- auto/_decision_tree: hoist GraphDiagram resolution into
_require_graph_diagram() helper, used by both
render_flowchart_mermaid and render_flowchart_svg
If figrecipe is absent, the functions raise a clear
ImportError pointing at `pip install scitex-stats[figrecipe]`
instead of an opaque ModuleNotFoundError at import time.
Add per-edge tests for the OPTIONAL figrecipe edge in scitex_stats._figrecipe_integration, mirroring the canonical scitex-io <-> figrecipe template. - INTEGRATION (figrecipe PRESENT, importorskip-guarded): to_figrecipe / annotate work and AxisWrapper unwrapping resolves. - DEGRADATION (figrecipe ABSENT, reversible sys.modules stub-shim): _AVAILABLE flips False and to_figrecipe/annotate/load_and_annotate each raise the documented ImportError carrying the pip-install hint, instead of leaking an opaque AttributeError. One-assert-per-test + explicit AAA markers; no monkeypatch/mocker.
test: figrecipe integration + degradation edge tests
refactor(stats): gate figrecipe imports through try_import_optional (SOC R4)
The _dataclasses.Stats schema has lived on develop unreleased (PyPI 0.2.21 lacks it), forcing scitex-io to git-pin scitex-stats. Publishing it unblocks scitex-io to depend on scitex-stats>=0.2.22 and exercises the .stats.zip bundle integration on PyPI.
release: v0.2.22 (publish _dataclasses Stats schema)
…load_stats)
Port the scitex-stats↔bundle integration from the scitex-python umbrella
into the standalone so the umbrella can drop its stats/ dir and pure-alias
to scitex_stats.
New module src/scitex_stats/_bundle_io.py exposes three public functions:
- test_result_to_stats(result) -> Stats: convert a run_test-shaped result
dict (flat or nested) to the canonical scitex_stats._dataclasses.Stats
schema.
- save_stats(comparisons, path, metadata=None, as_zip=False): persist as a
SciTeX bundle (kind="stats") via scitex_io.bundle.Bundle.
- load_stats(path): load a stats bundle into a flat, plot-friendly dict.
scitex-io is an OPTIONAL dependency (new extra [bundle] = scitex-io>=0.2.16):
the module imports cleanly without it and each function raises a clear
ImportError("scitex-io is required for stats bundle I/O; install
scitex-stats[bundle]") when called. The Stats dataclasses live in
scitex_stats (single source of truth; scitex-io imports them back), so the
extra stays optional with no hard/circular dependency and no direct refs.
Functions are exposed lazily on the package root (PEP 562) and listed in
__all__. Tests cover integration round-trips (importorskip scitex_io.bundle)
and graceful degradation via a reversible sys.modules shim (no monkeypatch).
All 28 new tests pass. Pre-existing test__server.py / MCP-CLI failures are
unrelated (FastMCP API drift in the env) and reproduce on clean develop.
feat(bundle): host stats↔bundle I/O (test_result_to_stats/save_stats/load_stats)
…le) into scitex_stats (#48) * feat: move scitex stats integration (Stats/annotate/to_figrecipe/bundle) into scitex_stats Add public _integration.py aggregator re-exporting the eight scitex-specific stats integration names so they import directly from scitex_stats: Stats, BUNDLE_AVAILABLE, test_result_to_stats, save_stats, load_stats, to_figrecipe, annotate, load_and_annotate - Stats schema sourced from scitex_stats._dataclasses (single source of truth). - Bundle I/O (test_result_to_stats/save_stats/load_stats/BUNDLE_AVAILABLE) re-exported from _bundle_io; scitex-io imported OPTIONALLY (extra [bundle]). - figrecipe annotations (to_figrecipe/annotate/load_and_annotate) re-exported from _figrecipe_integration; figrecipe imported OPTIONALLY (extra [figrecipe]). - Wire all eight names into the PEP 562 lazy dispatch table + __all__. - Add tests/scitex_stats/test__integration.py (AAA, importorskip-gated). Also migrate FastMCP 2.12 API: FastMCP.list_tools() was removed. - CLI mcp list-tools/doctor now use mcp.get_tools() (in-process accessor). - test__server.py: enumerate via in-memory fastmcp.Client; invoke decorated tools through FunctionTool.fn; read schemas via Tool.inputSchema. * fix: make FastMCP tool enumeration version-robust (2.x + 3.x) and close audit gaps CI runs FastMCP 3.3.1 where get_tools()/FunctionTool.fn do not exist; local dev had 2.12.4. Make all tool access version-robust: - _cli/mcp.py _list_tool_objects(): prefer in-process get_tools() (2.x, keeps return-type annotations), else fall back to the in-memory fastmcp.Client transport (stable across 2.x/3.x; MCP wire Tool has name/description/inputSchema). - _format_tool_signature(): accept .parameters (2.x) or .inputSchema schema attr. - test__server.py: _tool_fn() resolves the tool coroutine via .fn (2.x FunctionTool) or the plain function (3.x). Close audit gaps flagged by CI's scitex-dev 0.14.1: - PS-210: add scitex-io + figrecipe to [dev] so a fresh install runs the full suite (the [bundle]/[all] integration tests). - PS-140: add scitex_io.bundle to the cross-package import gate.
Contributor
|
Please sign the SciTeX CLA before your contribution can be merged. I have read and agree to the SciTeX CLA. You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot. |
Collaborator
Author
|
Closing as stale — fleet PR/Issue backlog sweep, 2026-08-11. What this PR attempted: chore(git): add /.worktrees/ to .gitignore Why it is being closed: newest generation for this repo but mergeState=DIRTY; a one-line .gitignore change is faster to redo than to rescue. Nothing is deleted. Branch |
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
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.
Housekeeping: live git worktrees under .worktrees/ show up as untracked noise in git status. Adds /.worktrees/ to .gitignore so this cannot recur. No code changes.