numpy-groupies
numpy-groupies copied to clipboard
ml31415
Optimised tools for group-indexing operations: aggregated sum and more
Hi Michael, What do you think of adding a pypi release workflow like https://github.com/pydata/xarray/blob/main/.github/workflows/pypi-release.yaml It makes it really easy to issue a release, all we have to do is tag...
Out of curiosity, some benchmarking with the pandas-replacement [polars](https://github.com/pola-rs/polars). ``` function ufunc numpy numba pandas polars ------------------------------------------------------------------------------- sum 37.351 1.575 0.705 11.804 5.760 prod 37.858 38.116 0.696 11.855 ---- amin...
This makes anything that only uses bincount work. max., min etc are more complicated.
TODO: - [ ] `aggregate_numpy.py` implementation: `sum`, `mean`, `var`, `sort`. Others are already ok I think. [Done: `sum` and `mean`] - [ ] write tests - [ ] implement in...
For convenience, it'd be nice to support medians even if it is relatively slow.
Thanks for this great package! Would it be possible to support `np.trapz` in the allowed callable functions? I'd be happy to contribute if it's possible, and any tips for how...
Firstly thanks for the impressive package. We're considering using it in https://github.com/pydata/xarray to provide faster groupby operations. It looks like some forms of stack / unstacking are supported too, if...
Hi, I'm working on a numba implementation of `np.trapz` for numpy-groupies. This seems to work for my purposes – might this be useful to anyone else? Thanks for the great...
``` import numpy as np import numpy_groupies as npg idx = np.arange(100).repeat(100) d = np.random.random(10000) npg.aggregate(idx, d, np.median) ``` And we get: ``` TypingError: Failed in nopython mode pipeline (step:...
Unable to run the example "Use the axis arg in order to do a sum-aggregation on three rows simultaneously" in [README.md](https://github.com/ml31415/numpy-groupies/blob/master/README.md) ``` import numpy as np import numpy_groupies as npg...
