feat: Allow float in interpolate_by by column

[agossard]

Ok, splitting this out in to a simpler PR. Fixes #16794

[codecov]

Codecov Report

Attention: Patch coverage is 87.50000% with 1 line in your changes missing coverage. Please review.

Project coverage is 80.34%. Comparing base (d5265d3) to head (c9ff1a8).
Report is 84 commits behind head on main.

Files	Patch %	Lines
...ops/src/series/ops/interpolation/interpolate_by.rs	87.50%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #18015      +/-   ##
==========================================
- Coverage   80.49%   80.34%   -0.16%     
==========================================
  Files        1496     1496              
  Lines      196786   197684     +898     
  Branches     2817     2821       +4     
==========================================
+ Hits       158407   158820     +413     
- Misses      37858    38343     +485     
  Partials      521      521              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[MarcoGorelli]

thanks @agossard 🙏 !

implementation looks good! you just need to run make pre-commit

thanks for amending the hypothesis test - could we also have one of the unit tests use a float by column?

[agossard]

Thanks Marco. I’ll try to fix the remaining style problem. In the unit test, just to confirm… as is, I’ve made it so the tests cover a float by column… by casting the existing example, like the other data types. Are you looking for a different piece of input data that actually has true non integer data in it?

[MarcoGorelli]

thanks @agossard for updating!

sorry I just meant - on top of the hypothesis test, to modify one of the unit tests (say, test_interpolate_by_trailing_nulls) to have a float by column

[MarcoGorelli]

this can be in given, you can use st.sampled_from

[MarcoGorelli]

as in, x_dtype=st.sampled_from([pl.Date, pl.Float64])

[MarcoGorelli]

thanks @agossard , just one pending comment

[MarcoGorelli]

as in, x_dtype=st.sampled_from([pl.Date, pl.Float64])

[MarcoGorelli]

looks like there's a test failure

__________________________ test_interpolate_vs_numpy ___________________________
[gw3] darwin -- Python 3.12.4 /Users/runner/work/polars/polars/.venv/bin/python3
tests/unit/operations/test_interpolate_by.py:168: in test_interpolate_vs_numpy
    def test_interpolate_vs_numpy(data: st.DataObject, x_dtype: pl.DataType) -> None:
tests/unit/operations/test_interpolate_by.py:228: in test_interpolate_vs_numpy
    assert_series_equal(result, expected)
polars/_utils/deprecation.py:91: in wrapper
    return function(*args, **kwargs)
E   AssertionError: Series are different (nan value mismatch)
E   [left]:  [0.0, 0.0, nan, 0.0]
E   [right]: [0.0, 0.0, 0.0, 0.0]
E   Falsifying example: test_interpolate_vs_numpy(
E       data=data(...),
E       x_dtype=Float64,  # or any other generated value
E   )
E   Draw 1: shape: (5, 2)
E   ┌─────────────┬───────┐
E   │ ts          ┆ value │
E   │ ---         ┆ ---   │
E   │ f64         ┆ f64   │
E   ╞═════════════╪═══════╡
E   │ 0.0         ┆ null  │
E   │ 0.0         ┆ 0.0   │
E   │ 9.9792e291  ┆ null  │
E   │ 9.9792e291  ┆ 0.0   │
E   │ -1.7977e308 ┆ 0.0   │
E   └─────────────┴───────┘
E   Explanation:
E       These lines were always and only run by failing examples:
E           /Users/runner/work/polars/polars/py-polars/polars/series/series.py:4039
E           /Users/runner/work/polars/polars/py-polars/polars/series/series.py:696
E           /Users/runner/work/polars/polars/py-polars/polars/series/series.py:739
E           /Users/runner/work/polars/polars/py-polars/polars/series/series.py:752
E           /Users/runner/work/polars/polars/py-polars/polars/series/series.py:782
E           (and 4 more with settings.verbosity >= verbose)
E   
E   You can reproduce this example by temporarily adding @reproduce_failure('6.108.9', b'AXicY2RhwAn2LEBnMgLR/v8QAOIgACMONgMA/UgKew==') as a decorator on your test case

---------- coverage: platform darwin, python 3.12.4-final-0 ----------
Coverage XML written to file main.xml

Required test coverage of 85.0% reached. Total coverage: 90.40%
=========================== short test summary info ============================
FAILED tests/unit/operations/test_interpolate_by.py::test_interpolate_vs_numpy - AssertionError: Series are different (nan value mismatch)
[left]:  [0.0, 0.0, nan, 0.0]
[right]: [0.0, 0.0, 0.0, 0.0]
Falsifying example: test_interpolate_vs_numpy(
    data=data(...),
    x_dtype=Float64,  # or any other generated value
)
Draw 1: shape: (5, 2)
┌─────────────┬───────┐
│ ts          ┆ value │
│ ---         ┆ ---   │
│ f64         ┆ f64   │
╞═════════════╪═══════╡
│ 0.0         ┆ null  │
│ 0.0         ┆ 0.0   │
│ 9.9792e291  ┆ null  │
│ 9.9792e291  ┆ 0.0   │
│ -1.7977e308 ┆ 0.0   │
└─────────────┴───────┘
Explanation:
    These lines were always and only run by failing examples:
        /Users/runner/work/polars/polars/py-polars/polars/series/series.py:4039
        /Users/runner/work/polars/polars/py-polars/polars/series/series.py:696
        /Users/runner/work/polars/polars/py-polars/polars/series/series.py:739
        /Users/runner/work/polars/polars/py-polars/polars/series/series.py:752
        /Users/runner/work/polars/polars/py-polars/polars/series/series.py:782
        (and 4 more with settings.verbosity >= verbose)

You can reproduce this example by temporarily adding @reproduce_failure('6.108.9', b'AXicY2RhwAn2LEBnMgLR/v8QAOIgACMONgMA/UgKew==') as a decorator on your test case

It's probably OK to keep the float case out of the hypothesis test, the unit tests you've added should be enough

[agossard]

OK, hopefully working now. Thanks, Marco!

[MarcoGorelli]

nice one, thanks @agossard !