This is an automated email from the ASF dual-hosted git repository. github-merge-queue[bot] pushed a commit to branch gh-readonly-queue/main/pr-8040-50321e403c82df299a13deb50a7f9849dd93bdba in repository https://gitbox.apache.org/repos/asf/texera.git
commit 5e48e30ae452d7a50889a82dd6847ea16234a577 Author: Xinyuan Lin <[email protected]> AuthorDate: Mon Aug 31 11:54:52 2026 +0000 ci(pyamber): measure branch coverage (#8040) ### What changes were proposed in this PR? Enables branch coverage for pyamber. Two files, no production or test code: - **`amber/pyproject.toml`** — adds `[tool.coverage.run] branch = true`. - **`.github/workflows/build.yml`** — a comment on the pytest step recording that branch coverage is on and where it is configured. The `run:` command itself is unchanged. ### Why the config rather than `--cov-branch` `--cov-branch` on the CI command would work, but it would put CI and a developer's laptop on **different metrics**: a local `pytest --cov` would keep reporting statement-only coverage, so a half-taken `if` would read as covered locally and partial on Codecov. Setting it in `pyproject.toml` gives both the same metric from one switch. The comment in `build.yml` exists so the pipeline is not silent about the change, and it says explicitly not to add the flag back — two switches for one boolean invites someone to remove one and believe it is off. ### What this surfaces Measured on `97e3585a6a`, with `codecov.yml`'s ignore rules applied (`proto/**`, `test_*.py`): | | Before | After | |---|---|---| | Fully covered lines | 4352 | 4290 | | **Partial lines** | **0** | **62** | | Missed lines | 122 | 122 | | Branch arms (taken/total) | not measured | 814/894 | **62 partial lines and 80 untaken branch arms become visible.** No line's actual execution changed — the same 1159 tests run and the same statements execute. This only stops the report from rounding a half-taken branch up to "covered". ### The reported number will go down, and that is the point **pyamber's flag figure drops 97.27% → 95.89%, about 1.4 points.** That is expected: partial lines stop counting as hits. `codecov.yml` sets `coverage.status.project` to `target: auto, threshold: 1%`, so it is worth being precise about whether this trips the check. It does not: - The project status is computed **repo-wide**, and `flag_management` in `codecov.yml` sets only `carryforward` — no per-flag statuses are configured, so pyamber's own 1.4-point drop is not itself gated. - Repo-wide, 62 lines out of 47,713 moves coverage **93.32% → 93.19%, a 0.13-point drop** — comfortably inside the 1% slack. - The `patch` status is unaffected: this PR changes no coverable lines. If a per-flag pyamber status is ever added, it should be added after this lands so its baseline is the branch-aware figure. ### Verification Run locally on Python 3.12.10 with coverage 7.13.5 and pytest 7.4.0 — the same coverage version CI resolves. - **The config alone does the job.** Ran the CI command **verbatim, without any `--cov-branch` flag**, and `coverage.xml` came back with `branch-rate="0.9105"` and 447 lines carrying `condition-coverage` data. Before the change the same command gives `branch-rate="0"` and zero such lines. coverage.py finds `amber/pyproject.toml` because CI does `cd amber` first, which is also how the repo documents running pytest. - **No behavioural change.** 1159 passed on both sides. The pre-existing local failures (5 failed, 7 errors — Iceberg/Windows environment, present on `main`) are **identical in identity**, not merely in count: the two `FAILED`/`ERROR` sets diff clean at 12 entries. - **`build.yml` still parses** as YAML (9 jobs) and the pyamber step's `run:` command is byte-identical to `main`. - **`amber/pyproject.toml` still parses** via `tomllib`, with `[tool.ruff]` and `[tool.pytest.ini_options]` intact. - **Lint clean** using CI's exact scoping — `ruff check src/main/python src/test/python && ruff format --check src/main/python src/test/python` → `All checks passed!` / `211 files already formatted`. One note in case someone reproduces this: running `ruff format --check .` from `amber/` reports two files needing reformatting, both `.py` files living under `src/main/scala/.../aiassistant/`. They are pre-existing (identical output with this change reverted) and outside CI's ruff scope, so they are unrelated to this PR — mentioned only so the broader command's output is not mistaken for a regression. ### Any related issues, documentation, discussions? Closes #8039 ### How was this PR tested? ``` cd amber && pytest -m "not integration" --cov=src/main/python --cov-report=xml -q ``` ``` 1159 passed, 1 deselected, 1 xfailed (plus the pre-existing local Iceberg/Windows failures, unchanged from main) branch-rate = 0.9105 (was 0.0) ``` CI's own lint scope: ``` cd amber && ruff check src/main/python src/test/python && ruff format --check src/main/python src/test/python ``` ### Was this PR authored or co-authored using generative AI tooling? Generated-by: Claude Code (Opus 5) --- .github/workflows/build.yml | 6 ++++++ amber/pyproject.toml | 11 ++++++++++- 2 files changed, 16 insertions(+), 1 deletion(-) diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml index 35e921a3d2..c0107997b3 100644 --- a/.github/workflows/build.yml +++ b/.github/workflows/build.yml @@ -1140,6 +1140,12 @@ jobs: # --junit-xml emits a JUnit-XML report alongside the coverage XML # so the Test Analytics upload below can feed Codecov's failing- # test PR comments and flaky-test detection on main. + # + # Branch coverage is ON, but via `[tool.coverage.run] branch = true` + # in amber/pyproject.toml rather than a `--cov-branch` flag here, so + # that a local `pytest --cov` reports the same metric as CI. Don't + # add the flag back: it would be redundant, and two switches for one + # boolean invites someone to remove one and think it is off. env: # Pytest runs with -s, so loguru's default stderr sink prints # straight into the CI log; pin it to WARNING (loguru has no WARN) diff --git a/amber/pyproject.toml b/amber/pyproject.toml index dd9f2f939e..3915e16ebb 100644 --- a/amber/pyproject.toml +++ b/amber/pyproject.toml @@ -42,4 +42,13 @@ testpaths = ["src/test/python"] addopts = "--import-mode=importlib" markers = [ "integration: end-to-end test routed to the amber-integration CI job", -] \ No newline at end of file +] + +# Branch coverage. Without this, coverage.py records only whether each +# statement ran, so a half-taken `if` reads as fully covered and Codecov +# has no partial-line data for any pyamber file. Setting it here rather +# than passing `--cov-branch` in CI keeps a local `pytest --cov` run on +# the same metric as the pipeline — a local-vs-CI split is exactly how a +# line comes to look green on a laptop and partial on Codecov. +[tool.coverage.run] +branch = true
