Fix test suite #1314
Fix test suite #1314
Conversation
|
Still seems to be problem solving on Windows: Install runtime dependencies with: conda install -y "colorcet" "dask-core" "multipledispatch" "numba" "numpy" "pandas" "param" "pillow" "pyct" "requests" "scipy" "toolz" "xarray" "dask-geopandas" "geopandas" "shapely >=2.0.0" "codecov" "geodatasets" "fastparquet" "flake8" "nbconvert" "nbformat" "nbsmoke >0.5" "netcdf4" "pyarrow" "pytest <8" "pytest-benchmark" "pytest-cov" "rasterio" "rioxarray" "scikit-image" "spatialpandas" "dask-geopandas" "geopandas" "shapely >=2.0.0" "bokeh >3.1" "geodatasets" "holoviews" "matplotlib-base >=3.3" "panel >1.1" "scikit-image" "spatialpandas"
Install runtime dependencies with: conda install -y "colorcet" "dask-core" "multipledispatch" "numba" "numpy" "pandas" "param" "pillow" "pyct" "requests" "scipy" "toolz" "xarray" "dask-geopandas" "geopandas" "shapely >=2.0.0" "codecov" "geodatasets" "fastparquet" "flake8" "nbconvert" "nbformat" "nbsmoke >0.5" "netcdf4" "pyarrow" "pytest <8" "pytest-benchmark" "pytest-cov" "rasterio" "rioxarray" "scikit-image" "spatialpandas" "dask-geopandas" "geopandas" "shapely >=2.0.0" "bokeh >3.1" "geodatasets" "holoviews" "matplotlib-base >=3.3" "panel >1.1" "scikit-image" "spatialpandas"
Channels:
- pyviz/label/dev
- numba
- conda-forge
Platform: win-64
Collecting package metadata (repodata.json): ...working... done
warning libmamba Added empty dependency for problem type SOLVER_RULE_UPDATETook 1.5 hours to solve for one of them.... |
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## main #1314 +/- ##
==========================================
- Coverage 85.65% 85.64% -0.02%
==========================================
Files 52 52
Lines 11288 11291 +3
==========================================
+ Hits 9669 9670 +1
- Misses 1619 1621 +2 |
|
Failing tests on Linux are the same reason as holoviz/geoviews#703, will solve it self (likely by the end of the day) |
|
The slow solve Windows seems to be fixed by removing fastparquet. I don't think it is a culprit by itself, but it is likely caused by some combination with other packages. |
| @@ -835,6 +835,12 @@ def test_sort_values_frame(self): | |||
| def test_where_series(self): | |||
| pass | |||
|
|
|||
| @pytest.mark.xfail(reason="not currently supported") | |||
There was a problem hiding this comment.
Should we instead pin to pandas !=2.2, expecting them to fix it in the next release?
There was a problem hiding this comment.
The line you have commented on is a new test for ExtensionArray, which was added in Pandas 2.2.
I have updated the skip for the upstream regression.
There was a problem hiding this comment.
I'm confused, since it says @pytest.mark.xfail(reason="not currently supported") ("currently"), but in any case, sounds like you're on it!
There was a problem hiding this comment.
duplicated has likely never worked for RaggedArray; I have added a NotImplementedError for it in the class. Before that, this would give a wrong result.
import pandas as pd
from datashader.datatypes import RaggedArray
data = RaggedArray([[0, 1], [1, 2, 3, 4], [], [-1, -2], []], dtype='float64')
arr = data.take([0, 1, 0, 1])
arr.duplicated('first')
arr = pd.array([0, 1, 0, 1])
arr.duplicated('first')Pandas use a hashtable to check if something is duplicated, but it is not designed to work with RaggedArray (which makes sense). I don't think it is impossible to support this, which is why I have added "currently". I hope that clarifies things.
There was a problem hiding this comment.
Ok, thanks! Maybe the way to clarify it would be to state what doesn't currently support it, e.g. "not currently supported by Pandas", "not currently supported by RaggedArray", or "not currently supported by Datshader".
This fixes the failing tests and removes warnings from the test CI.
Also fixes a problem with Windows not solving the conda environment.
Before this PR:
13 failed, 1581 passed, 62 skipped, 10 xfailed, 13569 warningsAfter this PR:
1581 passed, 65 skipped, 11 xfailed