Repository navigation
feat(api): expose trend_analysis at top level (broken — listed in __all__ but unimportable) #81
Description
Problem
trend_analysis is listed in __all__ at python/eo_processor/__init__.py:151 and documented as a public top-level function in the README, but it is neither imported from ._core nor defined as a wrapper anywhere in __init__.py.
This means:
from eo_processor import trend_analysis→ImportErrorfrom eo_processor import *→ImportError(star-import resolves every name in__all__, andtrend_analysisis unbound)- The README itself works around the bug by telling users to import from the private module:
from eo_processor._core import trend_analysis
This directly violates the project's own invariant from skills/python-api-surface/SKILL.md: "if it's public, it must be importable, typed, tested, and documented."
The Rust kernel works
The function is fully implemented in src/trends.rs:20 and registered into _core at src/lib.rs:100. Only the Python public surface is broken.
Scope
| File | Change |
|---|---|
python/eo_processor/__init__.py |
Add trend_analysis as _trend_analysis and TrendSegment as _TrendSegment to the from ._core import (...) block. Add a thin wrapper def trend_analysis(y, threshold): ... following the existing pattern (e.g. linear_regression at ~line 349). |
python/eo_processor/__init__.pyi |
Add stub for trend_analysis(y: Sequence[float], threshold: float) -> list[TrendSegment]: ... and a class TrendSegment matching the Rust struct fields (start_index, end_index, slope, intercept). |
tests/test_trends.py |
Add tests: no-break case (large threshold → single segment), a break case (linspace(0,10,50) + linspace(10,0,50) → ≥2 segments), invalid inputs (negative threshold, NaN in y). |
README.md:321 |
Replace from eo_processor._core import trend_analysis with from eo_processor import trend_analysis. |
Acceptance criteria
-
from eo_processor import trend_analysisworks -
from eo_processor import *works (no ImportError) - Type stub present in
__init__.pyi - At least 3 tests covering the function
- README import example updated to use the public API
-
ruff check python/eo_processor/passes - CI green
Version impact
Per the project's SemVer section: exposing a previously-unexported-but-documented function is a minor bump (new public surface, backward-compatible).
Note: A bump:minor label will be needed on the PR to trigger the automated version bump and release pipeline.
Activity
BnJam commentedon Aug 1, 2026
Starting work on this issue. Plan: import trend_analysis/TrendSegment from _core into python/eo_processor/init.py, add a thin wrapper following the linear_regression pattern, add the type stub and TrendSegment class to init.pyi, add tests in tests/test_trends.py, and update the README import example to use the public API.
BnJam commentedon Aug 1, 2026
Opened PR: #82
BnJam commentedon Aug 1, 2026
technocore completed pipeline
issue-handler.Now let me look at the Rust implementation, the stub, tests, and README.
Now let me check the Sphinx docs and skills mentioned in the issue, plus how
linear_regressionis exposed.Now let me check the environment for running tests.
No Python/Rust toolchain locally, so I'll rely on careful review + CI. Let me check CI and tox for what gates run.
Now let me implement. First, the import block in
__init__.py:Now add the wrapper near
linear_regression(line 349):Now the type stub
__init__.pyi:Now the tests in
tests/test_trends.py:Now the README fixes:
All private imports fixed. Let me verify the final state of the changed files:
Let me verify
CoreErrormaps toValueErrorin Python, since the tests rely on it:Let me add a star-import regression test, which directly validates acceptance criterion #2:
Now let me review final state of all changed files and check for any other references:
All changes complete. Now create the branch, commit, and push:
PR created. Now add the
bump:minorlabel and post the PR URL on the issue:Done. PR: #82
Fix summary:
python/eo_processor/__init__.py: importedtrend_analysis as _trend_analysisandTrendSegment as _TrendSegmentfrom._core, added the thintrend_analysis(y, threshold)wrapper (mirroringlinear_regressionat__init__.py:349) and aTrendSegment = _TrendSegmentalias. This binds the name in__all__(line 151), fixing bothfrom eo_processor import trend_analysisand star-imports.__init__.pyi: addedclass TrendSegment(start_index, end_index, slope, intercept) and thetrend_analysis(y: Sequence[float], threshold: float) -> list[TrendSegment]stub.tests/test_trends.py: 5 tests — no-break single segment, break detection (≥2 segments), negative threshold rejection, NaN rejection, and a star-import regression test.README.md+docs/source/README.md: replaced theeo_processor._coreworkaround with the public import.No local Python/Rust toolchain exists in this environment, so I couldn't run
ruff/pytestlocally; CI (tox lint + py tests) will verify. Added thebump:minorlabel to the PR to trigger the automated version bump/release.