Add Pandas 2.0 support

.github/workflows/test.yaml

@@ -27,7 +27,7 @@ jobs:
       - name: pre-commit
         uses: pre-commit/action@v3.0.0
   test_suite:
-    name: Pytest on ${{ matrix.python-version }}, Bokeh ${{ matrix.bokeh-version }}, ${{ matrix.os }}
+    name: Pytest on ${{ matrix.python-version }}, Bokeh ${{ matrix.bokeh-version }}, Pandas ${{ matrix.pandas-version }}, ${{ matrix.os }}
     needs: [pre_commit]
     runs-on: ${{ matrix.os }}
     strategy:
@@ -37,16 +37,22 @@ jobs:
         # Run on the full set on schedule, workflow_dispatch and push&tags events, otherwise on a subset.
         python-version: ${{ ( github.event_name == 'schedule' || github.event_name == 'workflow_dispatch' || ( github.event_name == 'push' && github.ref_type == 'tag' ) ) && fromJSON('["3.7", "3.8", "3.9", "3.10", "3.11"]') || fromJSON('["3.7", "3.9", "3.11"]') }}
         bokeh-version: ['2', '3']
+        pandas-version: ['1']
         exclude:
             # Bokeh 3 does not support Python 3.7
           - bokeh-version: '3'
             python-version: '3.7'
+        include:
+          - pandas-version: '2'
+            bokeh-version: '3'
+            python-version: '3.10'
+            os: 'ubuntu-latest'
     timeout-minutes: 120  # Because slow conda solve on Python 3.7
     defaults:
       run:
         shell: bash -el {0}
     env:
-      DESC: "Python ${{ matrix.python-version }} - Bokeh ${{ matrix.bokeh-version }} tests"
+      DESC: "Python ${{ matrix.python-version }} - Bokeh ${{ matrix.bokeh-version }} - Pandas ${{ matrix.pandas-version }} tests"
       PYTHON_VERSION: ${{ matrix.python-version }}
       SETUPTOOLS_ENABLE_FEATURES: "legacy-editable"
       DISPLAY: ":99.0"
@@ -56,7 +62,7 @@ jobs:
     steps:
       - uses: pyviz-dev/holoviz_tasks/install@v0.1a9
         with:
-          name: unit_test_suite_bokeh${{ matrix.bokeh-version }}
+          name: unit_test_suite_bokeh${{ matrix.bokeh-version }}_pandas${{ matrix.pandas-version }}
           python-version: ${{ matrix.python-version }}
           channel-priority: strict
           channels: pyviz/label/dev,bokeh/label/dev,conda-forge,nodefaults
@@ -65,6 +71,13 @@ jobs:
           conda-update: true
           conda-mamba: mamba
         id: install
+      - name: Pandas 2.0 install
+        if: steps.install.outputs.cache-hit != 'true' && matrix.pandas-version == '2'
+        run: |
+          conda activate test-environment
+          conda uninstall pandas dask dask-core --force --offline -y
+          mamba install pandas -c conda-forge/label/pandas_rc -y
+          mamba install dask-core -y
       - name: bokeh sampledata
         run: |
           conda activate test-environment

examples/gallery/demos/bokeh/bars_economic.ipynb

@@ -34,7 +34,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', '\\t')\n",
+    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', delimiter='\\t')\n",
     "key_dimensions   = [('year', 'Year'), ('country', 'Country')]\n",
     "value_dimensions = [('unem', 'Unemployment'), ('capmob', 'Capital Mobility'),\n",
     "                    ('gdp', 'GDP Growth'), ('trade', 'Trade')]\n",

examples/gallery/demos/bokeh/dropdown_economic.ipynb

@@ -34,7 +34,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', '\\t')\n",
+    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', delimiter='\\t')\n",
     "key_dimensions   = [('year', 'Year'), ('country', 'Country')]\n",
     "value_dimensions = [('unem', 'Unemployment'), ('capmob', 'Capital Mobility'),\n",
     "                    ('gdp', 'GDP Growth'), ('trade', 'Trade')]\n",

examples/gallery/demos/bokeh/scatter_economic.ipynb

@@ -34,7 +34,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', '\\t')\n",
+    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', delimiter='\\t')\n",
     "key_dimensions   = [('year', 'Year'), ('country', 'Country')]\n",
     "value_dimensions = [('unem', 'Unemployment'), ('capmob', 'Capital Mobility'),\n",
     "                    ('gdp', 'GDP Growth'), ('trade', 'Trade')]\n",

examples/gallery/demos/bokeh/us_unemployment.ipynb

@@ -39,7 +39,7 @@
    "source": [
     "from bokeh.sampledata.unemployment1948 import data\n",
     "\n",
-    "data = pd.melt(data.drop('Annual', 1), id_vars='Year', var_name='Month', value_name='Unemployment')\n",
+    "data = pd.melt(data.drop('Annual', axis=1), id_vars='Year', var_name='Month', value_name='Unemployment')\n",
     "heatmap = hv.HeatMap(data, label=\"US Unemployment (1948 - 2013)\")"
    ]
   },

examples/gallery/demos/matplotlib/bars_economic.ipynb

@@ -35,7 +35,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', '\\t')\n",
+    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', delimiter='\\t')\n",
     "key_dimensions   = [('year', 'Year'), ('country', 'Country')]\n",
     "value_dimensions = [('unem', 'Unemployment'), ('capmob', 'Capital Mobility'),\n",
     "                    ('gdp', 'GDP Growth'), ('trade', 'Trade')]\n",

examples/gallery/demos/matplotlib/dropdown_economic.ipynb

@@ -35,7 +35,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', '\\t')\n",
+    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', delimiter='\\t')\n",
     "key_dimensions   = [('year', 'Year'), ('country', 'Country')]\n",
     "value_dimensions = [('unem', 'Unemployment'), ('capmob', 'Capital Mobility'),\n",
     "                    ('gdp', 'GDP Growth'), ('trade', 'Trade')]\n",

examples/gallery/demos/matplotlib/scatter_economic.ipynb

@@ -36,7 +36,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', '\\t')\n",
+    "macro_df = pd.read_csv('http://assets.holoviews.org/macro.csv', delimiter='\\t')\n",
     "key_dimensions   = [('year', 'Year'), ('country', 'Country')]\n",
     "value_dimensions = [('unem', 'Unemployment'), ('capmob', 'Capital Mobility'),\n",
     "                    ('gdp', 'GDP Growth'), ('trade', 'Trade')]\n",

examples/gallery/demos/matplotlib/us_unemployment.ipynb

@@ -39,7 +39,7 @@
     "\n",
     "colors = [\"#75968f\", \"#a5bab7\", \"#c9d9d3\", \"#e2e2e2\", \"#dfccce\", \"#ddb7b1\", \"#cc7878\", \"#933b41\", \"#550b1d\"]\n",
     "\n",
-    "data = pd.melt(data.drop('Annual', 1), id_vars='Year', var_name='Month', value_name='Unemployment')\n",
+    "data = pd.melt(data.drop('Annual', axis=1), id_vars='Year', var_name='Month', value_name='Unemployment')\n",
     "\n",
     "heatmap = hv.HeatMap(data, label=\"US Unemployment (1948 - 2013)\")"
    ]

holoviews/element/util.py

@@ -294,7 +294,7 @@ def connect_tri_edges_pd(trimesh):
     edges = edges.drop("color", errors="ignore", axis=1).reset_index()
     nodes = trimesh.nodes.dframe().copy()
     nodes.index.name = 'node_index'
-    nodes = nodes.drop("color", errors="ignore", axis=1)
+    nodes = nodes.drop(["color", "z"], errors="ignore", axis=1)
     v1, v2, v3 = trimesh.kdims
     x, y, idx = trimesh.nodes.kdims[:3]
 

holoviews/operation/datashader.py

@@ -781,7 +781,7 @@ def _process(self, element, key=None):
             df['y0'] = np.array(0, df.dtypes[y.name])
             yagg = ['y0', y.name]
         if xtype == 'datetime':
-            df[x.name] = cast_array_to_int64(df[x.name].astype('datetime64[us]'))
+            df[x.name] = cast_array_to_int64(df[x.name].astype('datetime64[ns]'))
 
         params = self._get_agg_params(element, x, y, agg_fn, (x0, y0, x1, y1))
 
@@ -820,13 +820,13 @@ def _process(self, element, key=None):
         (x_range, y_range), (xs, ys), (width, height), (xtype, ytype) = info
         ((x0, x1), (y0, y1)), (xs, ys) = self._dt_transform(x_range, y_range, xs, ys, xtype, ytype)
 
-        df = element.interface.as_dframe(element)
+        df = element.interface.as_dframe(element).copy()
         if xtype == 'datetime':
-            df[x0d.name] = cast_array_to_int64(df[x0d.name].astype('datetime64[us]'))
-            df[x1d.name] = cast_array_to_int64(df[x1d.name].astype('datetime64[us]'))
+            df[x0d.name] = cast_array_to_int64(df[x0d.name].astype('datetime64[ns]'))
+            df[x1d.name] = cast_array_to_int64(df[x1d.name].astype('datetime64[ns]'))
         if ytype == 'datetime':
-            df[y0d.name] = cast_array_to_int64(df[y0d.name].astype('datetime64[us]'))
-            df[y1d.name] = cast_array_to_int64(df[y1d.name].astype('datetime64[us]'))
+            df[y0d.name] = cast_array_to_int64(df[y0d.name].astype('datetime64[ns]'))
+            df[y1d.name] = cast_array_to_int64(df[y1d.name].astype('datetime64[ns]'))
 
         if isinstance(agg_fn, ds.count_cat) and df[agg_fn.column].dtype.name != 'category':
             df[agg_fn.column] = df[agg_fn.column].astype('category')

holoviews/tests/core/data/test_daskinterface.py

@@ -1,7 +1,9 @@
 from unittest import SkipTest
+import unittest
 
 import numpy as np
 import pandas as pd
+from packaging.version import Version
 
 try:
     import dask.dataframe as dd
@@ -10,6 +12,7 @@
 
 from holoviews.core.data import Dataset
 from holoviews.util.transform import dim
+from holoviews.core.util import pandas_version
 
 from .test_pandasinterface import BasePandasInterfaceTests
 
@@ -73,6 +76,20 @@ def test_dataset_2D_partial_reduce_ht(self):
     def test_dataset_aggregate_string_types(self):
         raise SkipTest("Temporarily skipped")
 
+    @unittest.skipIf(
+        pandas_version >= Version("2.0rc0"),
+        reason="Not supported yet, https://github.com/dask/dask/issues/9913"
+    )
+    def test_dataset_aggregate_ht(self):
+        super().test_dataset_aggregate_ht()
+
+    @unittest.skipIf(
+        pandas_version >= Version("2.0rc0"),
+        reason="Not supported yet, https://github.com/dask/dask/issues/9913"
+    )
+    def test_dataset_aggregate_ht_alias(self):
+        super().test_dataset_aggregate_ht_alias()
+
     def test_dataset_from_multi_index(self):
         raise SkipTest("Temporarily skipped")
         df = pd.DataFrame({'x': np.arange(10), 'y': np.arange(10), 'z': np.random.rand(10)})

holoviews/tests/plotting/bokeh/test_callbacks.py

@@ -406,10 +406,8 @@ def test_cds_resolves(self):
         self.assertEqual(resolved, {'id': cds.ref['id'],
                                     'value': points.columns()})
 
-    @pytest.mark.filterwarnings("ignore::FutureWarning")
     def test_rangexy_datetime(self):
-        # Raises a warning because makeTimeDataFrame isn't part of the public API.
-        curve = Curve(pd.util.testing.makeTimeDataFrame(), 'index', 'C')
+        curve = Curve(pd._testing.makeTimeDataFrame(), 'index', 'C')
         stream = RangeXY(source=curve)
         plot = bokeh_server_renderer.get_plot(curve)
         callback = plot.callbacks[0]

setup.cfg

@@ -6,3 +6,4 @@ namespace_map =
     ibis-framework=ibis-sqlite
     ; dask pins to bokeh<3 right now
     dask=dask-core
+    geoviews=geoviews-core