compat: Pandas 3.0.0

holoviews/core/data/narwhals.py

@@ -192,17 +192,12 @@ def range(cls, dataset, dimension):
                 return np.nan, np.nan
             return cmin.item(), cmax.item()
         else:
+            col = nw.col(name)
             if dimension.nodata is not None:
-                expr = (
-                    nw.when(nw.col(name) == dimension.nodata)
-                    .then(None)
-                    .otherwise(nw.col(name))
-                    .alias(name)
+                df_column = df_column.select(
+                    nw.when(col != dimension.nodata).then(col)
                 )
-                df_column = df_column.select(expr)
-            if dataset.data.implementation in _NO_DROP_NULL:
-                col = nw.col(name)
-            else:
+            if dataset.data.implementation not in _NO_DROP_NULL:
                 col = nw.col(name).drop_nulls()
             # NOTE: Some narwhals backends (duckdb) will return nan as
             # the max value

holoviews/core/data/pandas.py

@@ -460,7 +460,12 @@ def values(
             data = (data if isindex else data.dt).tz_localize(None)
         if not expanded:
             return pd.unique(data)
-        return data.values if hasattr(data, 'values') else data
+        if hasattr(data, 'values'):
+            data = data.values
+            # https://pandas.pydata.org/docs/dev/user_guide/copy_on_write.html#read-only-numpy-arrays
+            if hasattr(data, "flags") and not data.flags.writeable:
+                data = data.copy()
+        return data
 
 
     @classmethod

holoviews/core/data/spatialpandas_dask.py

@@ -34,12 +34,17 @@ def frame_type(cls):
 
     @classmethod
     def init(cls, eltype, data, kdims, vdims):
+        import dask
         import dask.dataframe as dd
         data, dims, params = super().init(
             eltype, data, kdims, vdims
         )
         if not isinstance(data, cls.frame_type()):
-            data = dd.from_pandas(data, npartitions=1)
+            # convert-string will convert the object dtype to string
+            # even if the object is a list
+            # https://github.com/dask/dask/issues/10631
+            with dask.config.set({"dataframe.convert-string": False}):
+                data = dd.from_pandas(data, npartitions=1)
         return data, dims, params
 
     @classmethod

holoviews/core/util/__init__.py

@@ -860,9 +860,7 @@ def asarray(arraylike, strict=True):
         return arraylike
     elif isinstance(arraylike, list):
         return np.asarray(arraylike, dtype=object)
-    elif not isinstance(arraylike, np.ndarray) and isinstance(arraylike, arraylike_types):
-        return arraylike.values
-    elif hasattr(arraylike, '__array__'):
+    elif hasattr(arraylike, '__array__') or isinstance(arraylike, arraylike_types):
         return np.asarray(arraylike)
     elif strict:
         raise ValueError(f'Could not convert {type(arraylike)} type to array')

holoviews/core/util/dependencies.py

@@ -112,5 +112,15 @@ def __version__(self):
 NUMPY_GE_2_0_0 = NUMPY_VERSION >= (2, 0, 0)
 PANDAS_GE_2_1_0 = PANDAS_VERSION >= (2, 1, 0)
 PANDAS_GE_2_2_0 = PANDAS_VERSION >= (2, 2, 0)
-
-__all__ = ["NUMPY_GE_2_0_0", "NUMPY_VERSION", "PANDAS_GE_2_1_0", "PANDAS_GE_2_2_0", "PANDAS_VERSION", "PARAM_VERSION", "_LazyModule"]
+PANDAS_GE_3_0_0 = PANDAS_VERSION >= (3, 0, 0)
+
+__all__ = [
+    "NUMPY_GE_2_0_0",
+    "NUMPY_VERSION",
+    "PANDAS_GE_2_1_0",
+    "PANDAS_GE_2_2_0",
+    "PANDAS_GE_3_0_0",
+    "PANDAS_VERSION",
+    "PARAM_VERSION",
+    "_LazyModule",
+]

holoviews/operation/datashader.py

@@ -489,15 +489,21 @@ def _apply_datashader(self, dfdata, cvs_fn, agg_fn, agg_kwargs, x, y, agg_state:
             for col in dfdata.columns:
                 if col in agg.coords:
                     continue
-                val = dfdata[col].values[index]
-                if dtype_kind(val) == 'f':
+                ser = dfdata[col]
+                kind = dtype_kind(ser)
+                if kind == "O" and getattr(ser.dtype, "storage", None) == "pyarrow":
+                    # pyarrow can only handle 1-dimensional
+                    val = np.asarray(ser)[index]
+                else:
+                    val = ser.values[index]
+                if kind == 'f':
                     val[neg1] = np.nan
                 elif isinstance(val.dtype, pd.CategoricalDtype):
                     val = val.to_numpy()
                     val[neg1] = "-"
-                elif dtype_kind(val) == "O":
+                elif kind == "O":
                     val[neg1] = "-"
-                elif dtype_kind(val) == "M":
+                elif kind == "M":
                     val[neg1] = np.datetime64("NaT")
                 else:
                     val = val.astype(np.float64)

holoviews/operation/element.py

@@ -32,6 +32,7 @@
 from ..core.util import (
     datetime_types,
     dt_to_int,
+    dtype_kind,
     group_sanitizer,
     is_cupy_array,
     is_dask_array,
@@ -1065,16 +1066,23 @@ class interpolate_curve(Operation):
 
     _per_element = True
 
+    @staticmethod
+    def _get_dtype(obj):
+        if dtype_kind(obj) == "O":
+            # Handle non-numpy objects like Pandas 3.0 StringArray
+            return "O"
+        return obj.dtype
+
     @classmethod
     def pts_to_prestep(cls, x, values):
         steps = np.zeros(2 * len(x) - 1)
-        value_steps = tuple(np.empty(2 * len(x) - 1, dtype=v.dtype) for v in values)
+        value_steps = tuple(np.empty(2 * len(x) - 1, dtype=cls._get_dtype(v)) for v in values)
 
         steps[0::2] = x
         steps[1::2] = steps[0:-2:2]
 
         val_arrays = []
-        for v, s in zip(values, value_steps, strict=None):
+        for v, s in zip(values, value_steps, strict=True):
             s[0::2] = v
             s[1::2] = s[2::2]
             val_arrays.append(s)
@@ -1084,13 +1092,13 @@ def pts_to_prestep(cls, x, values):
     @classmethod
     def pts_to_midstep(cls, x, values):
         steps = np.zeros(2 * len(x))
-        value_steps = tuple(np.empty(2 * len(x), dtype=v.dtype) for v in values)
+        value_steps = tuple(np.empty(2 * len(x), dtype=cls._get_dtype(v)) for v in values)
 
         steps[1:-1:2] = steps[2::2] = x[:-1] + (x[1:] - x[:-1])/2
         steps[0], steps[-1] = x[0], x[-1]
 
         val_arrays = []
-        for v, s in zip(values, value_steps, strict=None):
+        for v, s in zip(values, value_steps, strict=True):
             s[0::2] = v
             s[1::2] = s[0::2]
             val_arrays.append(s)
@@ -1100,13 +1108,13 @@ def pts_to_midstep(cls, x, values):
     @classmethod
     def pts_to_poststep(cls, x, values):
         steps = np.zeros(2 * len(x) - 1)
-        value_steps = tuple(np.empty(2 * len(x) - 1, dtype=v.dtype) for v in values)
+        value_steps = tuple(np.empty(2 * len(x) - 1, dtype=cls._get_dtype(v)) for v in values)
 
         steps[0::2] = x
         steps[1::2] = steps[2::2]
 
         val_arrays = []
-        for v, s in zip(values, value_steps, strict=None):
+        for v, s in zip(values, value_steps, strict=True):
             s[0::2] = v
             s[1::2] = s[0:-2:2]
             val_arrays.append(s)

holoviews/plotting/bokeh/path.py

@@ -6,7 +6,7 @@
 
 from ...core import util
 from ...core.dimension import Dimension
-from ...core.util import dtype_kind
+from ...core.util import arraylike_types, dtype_kind
 from ...element import Contours, Polygons
 from ...util.transform import dim
 from .callbacks import PolyDrawCallback, PolyEditCallback
@@ -50,7 +50,7 @@ def _element_transform(self, transform, element, ranges):
             data = super()._element_transform(transform, element, ranges)
             new_data = []
             for d in data:
-                if isinstance(d, np.ndarray) and len(d) == 1:
+                if isinstance(d, arraylike_types) and len(d) == 1:
                     new_data.append(d[0])
                 else:
                     new_data.append(d)

holoviews/plotting/mpl/chart.py

@@ -966,7 +966,7 @@ def _create_bars(self, axis, element, ranges, style):
                 xdiff = 1
             else:
                 xdiff = np.min(xdiff)
-            width = (1 - self.bar_padding) * (date2num(xdiff) / 1000 if is_dt else xdiff)
+            width = (1 - self.bar_padding) * (date2num(xdiff.astype("timedelta64[us]")) if is_dt else xdiff)
             data_width = (1 - self.bar_padding) * xdiff
         else:
             xdiff = len(values.get('category', [None]))

holoviews/plotting/mpl/element.py

@@ -715,6 +715,8 @@ def _apply_transforms(self, element, ranges, style):
             if (not np.isscalar(val) and len(util.unique_array(val)) == 1 and
                 ("color" not in k or validate('color', val))):
                 val = val[0]
+            if isinstance(val, util.arraylike_types):
+                val = np.asarray(val)
 
             if not np.isscalar(val) and k in self._nonvectorized_styles:
                 element = type(element).__name__

holoviews/tests/conftest.py

@@ -54,13 +54,6 @@ def pytest_collection_modifyitems(config, items):
     mpl.use("agg")
 
 
-with contextlib.suppress(Exception):
-    # From Dask 2023.7.1 they now automatically convert strings
-    # https://docs.dask.org/en/stable/changelog.html#v2023-7-1
-    import dask
-
-    dask.config.set({"dataframe.convert-string": False})
-
 
 @pytest.fixture
 def ibis_sqlite_backend():

holoviews/tests/core/data/test_daskinterface.py

@@ -5,16 +5,26 @@
 import pandas as pd
 
 try:
+    import dask
     import dask.dataframe as dd
 except ImportError:
     raise SkipTest("Could not import dask, skipping DaskInterface tests.")
 
 from holoviews.core.data import Dataset
-from holoviews.core.util import PANDAS_VERSION
+from holoviews.core.util.dependencies import (
+    PANDAS_GE_3_0_0,
+    PANDAS_VERSION,
+    _no_import_version,
+)
 from holoviews.util.transform import dim
 
 from .test_pandasinterface import BasePandasInterfaceTests
 
+_DASK_CONVERT_STRING = (
+    _no_import_version("dask") >= (2023, 7, 1)
+    and dask.config.get("dataframe.convert-string") in (True, None)
+)
+
 
 class DaskDatasetTest(BasePandasInterfaceTests):
     """
@@ -137,3 +147,16 @@ def test_dataset_get_dframe_by_dimension(self):
         df = self.dataset_hm.dframe(['x'])
         expected = self.frame({'x': self.xs}, dtype=df.dtypes.iloc[0]).compute()
         self.assertEqual(df, expected)
+
+    def test_dataset_dataset_ht_dtypes(self):
+        ds = self.table
+        if _DASK_CONVERT_STRING:
+            string_dtype = pd.StringDtype(na_value=pd.NA, storage="pyarrow")
+        elif PANDAS_GE_3_0_0:
+            string_dtype = pd.StringDtype(na_value=np.nan)
+        else:
+            string_dtype = np.dtype('object')
+        assert ds.interface.dtype(ds, 'Gender') == string_dtype
+        assert ds.interface.dtype(ds, 'Age') == np.dtype(int)
+        assert ds.interface.dtype(ds, 'Weight') == np.dtype(int)
+        assert ds.interface.dtype(ds, 'Height') == np.dtype('float64')

holoviews/tests/core/data/test_pandasinterface.py

@@ -7,6 +7,7 @@
 from holoviews.core.data.pandas import PandasInterface
 from holoviews.core.dimension import Dimension
 from holoviews.core.spaces import HoloMap
+from holoviews.core.util.dependencies import PANDAS_GE_3_0_0
 from holoviews.element import Distribution, Points, Scatter
 
 from .base import HeterogeneousColumnTests, InterfaceTests
@@ -172,6 +173,14 @@ def test_dataset_range_with_object_index(self):
         ds = Dataset(df, kdims='index')
         assert ds.range('index') == ('A', 'D')
 
+    def test_dataset_dataset_ht_dtypes(self):
+        ds = self.table
+        string_dtype = pd.StringDtype(na_value=np.nan) if PANDAS_GE_3_0_0 else np.dtype('object')
+        assert ds.interface.dtype(ds, 'Gender') == string_dtype
+        assert ds.interface.dtype(ds, 'Age') == np.dtype(int)
+        assert ds.interface.dtype(ds, 'Weight') == np.dtype(int)
+        assert ds.interface.dtype(ds, 'Height') == np.dtype('float64')
+
 
 class PandasInterfaceTests(BasePandasInterfaceTests):
 
@@ -412,6 +421,14 @@ def test_groupby_one_index_one_column(self):
         for k, v in grouped.items():
             pd.testing.assert_frame_equal(v.data, ds.select(values=k).data)
 
+    def test_dataset_dataset_ht_dtypes(self):
+        ds = self.table
+        string_dtype = pd.StringDtype(na_value=np.nan) if PANDAS_GE_3_0_0 else np.dtype('object')
+        assert ds.interface.dtype(ds, 'Gender') == string_dtype
+        assert ds.interface.dtype(ds, 'Age') == np.dtype(int)
+        assert ds.interface.dtype(ds, 'Weight') == np.dtype(int)
+        assert ds.interface.dtype(ds, 'Height') == np.dtype('float64')
+
     def test_regression_no_auto_index(self):
         # https://github.com/holoviz/holoviews/issues/6298
 

holoviews/tests/core/test_utils.py

@@ -7,6 +7,7 @@
 import unittest
 from itertools import product
 from pathlib import Path
+from zoneinfo import ZoneInfo
 
 import narwhals.stable.v2 as nw
 import numpy as np
@@ -589,8 +590,7 @@ def test_date_range_1_sec(self):
         self.assertEqual(drange[-1], end-np.timedelta64(50, 'ms'))
 
     def test_timezone_to_int(self):
-        import pytz
-        timezone = pytz.timezone("Europe/Copenhagen")
+        timezone = ZoneInfo("Europe/Copenhagen")
 
         values = [
             datetime.datetime(2021, 4, 8, 12, 0, 0, 0),