Error when using `exact_extract` with any extra arguments to statistics

[theroggy]

When I run exact_extract with any statistic having extra arguments (e.g. "mean(coverage_weight=area_spherical_m2"), I get an error. With a "plain" statistic (e.g. "mean") there is no problem. This is an example of such an error:

File "C:\Tools\mambaforge\envs\pysnippets\Lib\site-packages\exactextract\exact_extract.py", line 169, in prep_ops
    ops += prepare_operations([stat], values, weights)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: Unexpected argument(s) to stat: mean

The problem must be related to my system/installation, as I can reproduce it by running an extract from the exactextract test suite, which runs fine on CI.

I'm working on windows with conda. To try to avoid possible side effects from dependencies, I created a new conda environment and just installed pip and the exactextract pip package:

# Name                    Version                   Build  Channel
bzip2                     1.0.8                hcfcfb64_5    conda-forge
ca-certificates           2024.2.2             h56e8100_0    conda-forge
exactextract              0.2.0.dev0               pypi_0    pypi
libexpat                  2.6.2                h63175ca_0    conda-forge
libffi                    3.4.2                h8ffe710_5    conda-forge
libsqlite                 3.45.3               hcfcfb64_0    conda-forge
libzlib                   1.2.13               hcfcfb64_5    conda-forge
openssl                   3.3.0                hcfcfb64_0    conda-forge
pip                       24.0               pyhd8ed1ab_0    conda-forge
python                    3.12.3          h2628c8c_0_cpython    conda-forge
setuptools                69.5.1             pyhd8ed1ab_0    conda-forge
tk                        8.6.13               h5226925_1    conda-forge
tzdata                    2024a                h0c530f3_0    conda-forge
ucrt                      10.0.22621.0         h57928b3_0    conda-forge
vc                        14.3                hcf57466_18    conda-forge
vc14_runtime              14.38.33130         h82b7239_18    conda-forge
vs2015_runtime            14.38.33130         hcb4865c_18    conda-forge
wheel                     0.43.0             pyhd8ed1ab_1    conda-forge
xz                        5.2.6                h8d14728_0    conda-forge

An easy way to reproduce in my case is to run the following extract from one of the tests in the exactextract test suite:

import numpy as np

from exactextract import exact_extract
from exactextract.feature import JSONFeatureSource
from exactextract.raster import NumPyRasterSource


def make_rect(xmin, ymin, xmax, ymax, id=None, properties=None):
    f = {
        "type": "Feature",
        "geometry": {
            "type": "Polygon",
            "coordinates": [
                [[xmin, ymin], [xmax, ymin], [xmax, ymax], [xmin, ymax], [xmin, ymin]]
            ],
        },
    }

    if id is not None:
        f["id"] = id
    if properties is not None:
        f["properties"] = properties

    return f


rast = NumPyRasterSource(np.arange(1, 10, dtype=np.int32).reshape(3, 3))
square = JSONFeatureSource(make_rect(0.5, 0.5, 2.5, 2.5))
result = exact_extract(
    rast,
    square,
    [
        "m1=mean",
        "m2=mean(coverage_weight=area_spherical_m2)",
        "c1=coverage",
        "c2=coverage(coverage_weight=area_spherical_m2)",
        "c3=coverage(coverage_weight=area_spherical_km2)",
        "c4=coverage(coverage_weight=area_cartesian)",
    ],
    strategy="feature-sequential",
)[0]["properties"]

[dbaston]

The package on PyPI isn't up to date with GitHub. I'll push an updated snapshot to Test PyPI at https://test.pypi.org/project/exactextract/. The dependencies are fairly minimal if you want to build the library from source.

[theroggy]

That version indeed works fine. Thanks!

[theroggy]

Maybe better to push the version as well to regular/production pypi, e.g. as an alpha version, as the version available there now is ~broken?

[dbaston]

I don't know about "broken", but it's nearly 2 months old so it doesn't have this feature.

I'm trying to figure out the right frequency to push development releases to regular PyPI. Any info on best practices would be welcome.

[theroggy]

I don't know about "broken", but it's nearly 2 months old so it doesn't have this feature.

Ah, OK, sorry. I misinterpreted the error message as well as the timeline of the pip package vs. when the new feature was introduced. I should have looked better :-(...

I'm trying to figure out the right frequency to push development releases to regular PyPI. Any info on best practices would be welcome.

I wouldn't call it best practices, but based on the projects I'm involved in (not that many), development releases (or rather alpha releases, never encountered development releases) are quite rare and there is no specific frequency. They are only used if there is a need for some broader testing exposure than unit tests can offer. Or, if there is a specific wish/need for someone to get an earlier version (either for broader testing or just to have it earlier), but that's just as rare.
Sometimes there are beta releases or release candidates e.g. a few weeks before a feature release, but even those aren't very common.

I think exactextract is now just a bit in a special situation as there isn't a "stable" python release yet?

[dbaston]

I think exactextract is now just a bit in a special situation as there isn't a "stable" python release yet?

Yes. Hopefully not much longer, I just need to convince myself that the current API is suitable.