Fix CI and other issues

pyproject.toml

@@ -22,15 +22,16 @@ classifiers = [
 urls = { Homepage = "https://github.com/sgkit-dev/sgkit" }
 requires-python = ">=3.10"
 dependencies = [
-  "numpy < 2.2",
+  "numpy < 2.3; python_version < \"3.11\"",
+  "numpy >= 2.3.5, < 2.4; python_version >= \"3.11\"", # M4 chip error
   "xarray < 2025.03.1",
   "dask[array,dataframe] >= 2022.01.0, <= 2024.8.0",
   "scipy < 1.16",
   "zarr >= 2.10.0, != 2.11.0, != 2.11.1, != 2.11.2, < 3",
   "numba",
   "fsspec != 2021.6.*",
   "scikit-learn",
-  "pandas",
+  "pandas < 3.0",
   "setuptools >= 41.2",                                   # For pkg_resources
 ]
 dynamic = ["version"]
@@ -69,7 +70,7 @@ fail_under = 100
 
 [tool.pytest.ini_options]
 addopts = "--doctest-modules --ignore=validation --cov-fail-under=100"
-norecursedirs = [".eggs", "build", "docs"]
+norecursedirs = [".eggs", "build", "docs", ".hypothesis"]
 filterwarnings = ["error", "ignore::DeprecationWarning"]
 
 

requirements-numpy1.txt

@@ -6,4 +6,4 @@ numba
 zarr >= 2.10.0, != 2.11.0, != 2.11.1, != 2.11.2, < 3
 fsspec != 2021.6.*
 scikit-learn
-pandas
+pandas < 3.0

requirements.txt

@@ -1,9 +1,10 @@
-numpy < 2.2
+numpy < 2.3; python_version < "3.11"
+numpy >= 2.3.5, < 2.4; python_version >= "3.11"
 xarray < 2025.03.1
 dask[array,dataframe] >= 2023.01.0, <= 2024.8.0
 scipy < 1.16
 numba
 zarr >= 2.10.0, != 2.11.0, != 2.11.1, != 2.11.2, < 3
 fsspec != 2021.6.*
 scikit-learn
-pandas
+pandas < 3.0

sgkit/tests/io/test_dataset.py

@@ -58,7 +58,7 @@ def test_save_unequal_chunks_error():
     # Normal zarr errors shouldn't be caught
     with pytest.raises(
         (FileExistsError, ValueError),
-        match="(path '' contains an array|is not empty)",
+        match="(path '' contains an array|is not empty|contains data)",
     ):
         save_dataset(ds, {".zarray": ""})
 

sgkit/tests/test_association.py

@@ -254,7 +254,11 @@ def run(traits: Sequence[str]) -> Dataset:
 
     traits = [f"trait_{i}" for i in range(ds.attrs["n_trait"])]
     # Run regressions on individual traits and concatenate resulting statistics
-    dfr_single = xr.concat([run([t]) for t in traits], dim="traits").to_dataframe()
+    dfr_single = xr.concat(
+        [run([t]) for t in traits],
+        dim="traits",
+        data_vars="all",
+    ).to_dataframe()
     # Run regressions on all traits simulatenously
     dfr_multi: DataFrame = run(traits).to_dataframe()
     pd.testing.assert_frame_equal(dfr_single, dfr_multi)