Rolling mean / sum incorrect behavior when min_samples < window #22868
Description
Checks
- I have checked that this issue has not already been reported.
- I have confirmed this bug exists on the latest version of Polars.
Reproducible example
pl.DataFrame({"a": [1, 2, 3, 4, 5]}).with_columns(
pl.col("a").rolling_mean(2, weights=[1, 2], min_samples=1)
)Log output
shape: (5, 1)
┌──────────┐
│ a │
│ --- │
│ f64 │
╞══════════╡
│ 0.333333 │
│ 1.666667 │
│ 2.666667 │
│ 3.666667 │
│ 4.666667 │
└──────────┘Issue description
For rolling functions with weights supplied, when min_samples is set smaller than the window size, then the left most values of weights are applied in calculations to the sub-windows, when in fact it should be the right-most.
Expected behavior
In the example above, the first value should either be 1 (as the mean over on sample point is the sample point itsefl), or 0.6666... if we normalize by the total sum of weights. Having the option for both outputs is useful fwiw.
Installed versions
adbc_driver_manager <not installed>
altair <not installed>
azure.identity <not installed>
boto3 <not installed>
cloudpickle <not installed>
connectorx <not installed>
deltalake <not installed>
fastexcel <not installed>
fsspec <not installed>
gevent <not installed>
google.auth 2.23.3
great_tables <not installed>
matplotlib 3.8.0
numpy 1.26.0
openpyxl 3.1.4
pandas 2.2.3
polars_cloud <not installed>
pyarrow 19.0.1
pydantic <not installed>
pyiceberg <not installed>
sqlalchemy 2.0.0
torch <not installed>
xlsx2csv <not installed>
xlsxwriter <not installed>