Add Pandas 2.0 support

.github/workflows/test.yaml

@@ -45,7 +45,7 @@ jobs:
             # Bokeh 3 does not support Python 3.7
           - bokeh-version: '3'
             python-version: '3.7'
-    timeout-minutes: 120  # Because slow conda solve on Python 3.7
+    timeout-minutes: 120
     defaults:
       run:
         shell: bash -el {0}
@@ -63,7 +63,7 @@ jobs:
           name: unit_test_suite_bokeh${{ matrix.bokeh-version }}
           python-version: ${{ matrix.python-version }}
           channel-priority: strict
-          channels: pyviz/label/dev,bokeh/label/dev,conda-forge,nodefaults
+          channels: pyviz/label/dev,conda-forge,nodefaults
           envs: "-o flakes -o tests -o examples_tests -o bokeh${{ matrix.bokeh-version }}"
           cache: true
           conda-update: true
@@ -78,20 +78,11 @@ jobs:
           conda activate test-environment
           doit test_unit
       - name: test examples
-        if: matrix.python-version != '3.11'
         run: |
           conda activate test-environment
           mkdir -p ~/.jupyter/
           echo "c.ExecutePreprocessor.startup_timeout=600" >> ~/.jupyter/jupyter_nbconvert_config.py
           doit test_examples
-      - name: test examples - python 3.11
-        # Should be removed when numba support python 3.11
-        if: matrix.python-version == '3.11'
-        run: |
-          conda activate test-environment
-          mkdir -p ~/.jupyter/
-          echo "c.ExecutePreprocessor.startup_timeout=600" >> ~/.jupyter/jupyter_nbconvert_config.py
-          pytest -n auto --dist loadscope --nbval-lax examples/reference/elements
       - name: codecov
         run: |
           conda activate test-environment

examples/conftest.py

@@ -0,0 +1,32 @@
+import sys
+
+import pandas as pd
+from packaging.version import Version
+
+PD2 = Version(pd.__version__) >= Version("2.0")
+
+collect_ignore_glob = [
+    # Needs selenium, phantomjs, firefox, and geckodriver to save a png picture
+    "user_guide/Plotting_with_Bokeh.ipynb",
+    # Possible timeout error
+    "user_guide/17-Dashboards.ipynb",
+    # Give file not found
+    "user_guide/Plots_and_Renderers.ipynb",
+]
+
+# Numba incompatibility
+if sys.version_info >= (3, 11):
+    collect_ignore_glob += [
+        "user_guide/15-Large_Data.ipynb",
+        "user_guide/16-Streaming_Data.ipynb",
+        "user_guide/Linked_Brushing.ipynb",
+        "user_guide/Network_Graphs.ipynb",
+    ]
+
+# Pandas bug: https://github.com/pandas-dev/pandas/issues/52451
+if PD2 and sys.platform == "win32":
+    collect_ignore_glob += [
+        "gallery/demos/bokeh/point_draw_triangulate.ipynb",
+        "reference/elements/*/TriMesh.ipynb",
+        "user_guide/15-Large_Data.ipynb",
+    ]

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/core/util.py

@@ -2108,7 +2108,7 @@ def dt_to_int(value, time_unit='us'):
         try:
             value = np.datetime64(value, 'ns')
             tscale = (np.timedelta64(1, time_unit)/np.timedelta64(1, 'ns'))
-            return value.tolist()/tscale
+            return int(value.tolist() / tscale)
         except Exception:
             # If it can't handle ns precision fall back to datetime
             value = value.tolist()

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

@@ -222,11 +222,11 @@ def _get_sampling(self, element, x, y, ndim=2, default=None):
     def _dt_transform(self, x_range, y_range, xs, ys, xtype, ytype):
         (xstart, xend), (ystart, yend) = x_range, y_range
         if xtype == 'datetime':
-            xstart, xend = (np.array([xstart, xend])/1e3).astype('datetime64[us]')
-            xs = (xs/1e3).astype('datetime64[us]')
+            xstart, xend = np.array([xstart, xend]).astype('datetime64[ns]')
+            xs = xs.astype('datetime64[ns]')
         if ytype == 'datetime':
-            ystart, yend = (np.array([ystart, yend])/1e3).astype('datetime64[us]')
-            ys = (ys/1e3).astype('datetime64[us]')
+            ystart, yend = np.array([ystart, yend]).astype('datetime64[ns]')
+            ys = ys.astype('datetime64[ns]')
         return ((xstart, xend), (ystart, yend)), (xs, ys)
 
 
@@ -528,9 +528,9 @@ def _process(self, element, key=None):
         if 'x_axis' in agg.coords and 'y_axis' in agg.coords:
             agg = agg.rename({'x_axis': x, 'y_axis': y})
         if xtype == 'datetime':
-            agg[x.name] = (agg[x.name]/1e3).astype('datetime64[us]')
+            agg[x.name] = agg[x.name].astype('datetime64[ns]')
         if ytype == 'datetime':
-            agg[y.name] = (agg[y.name]/1e3).astype('datetime64[us]')
+            agg[y.name] = agg[y.name].astype('datetime64[ns]')
 
         if agg.ndim == 2:
             # Replacing x and y coordinates to avoid numerical precision issues
@@ -700,7 +700,7 @@ def _process(self, element, key=None):
 
         agg = cvs.area(df, x.name, y.name, agg_fn, axis=0, y_stack=ystack)
         if xtype == "datetime":
-            agg[x.name] = (agg[x.name]/1e3).astype('datetime64[us]')
+            agg[x.name] = agg[x.name].astype('datetime64[ns]')
 
         return self.p.element_type(agg, **params)
 
@@ -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))
 
@@ -797,7 +797,7 @@ def _process(self, element, key=None):
 
         agg = cvs.line(df, x.name, yagg, agg_fn, axis=1, **agg_kwargs).rename(rename_dict)
         if xtype == "datetime":
-            agg[x.name] = (agg[x.name]/1e3).astype('datetime64[us]')
+            agg[x.name] = agg[x.name].astype('datetime64[ns]')
 
         return self.p.element_type(agg, **params)
 
@@ -821,12 +821,14 @@ def _process(self, element, key=None):
         ((x0, x1), (y0, y1)), (xs, ys) = self._dt_transform(x_range, y_range, xs, ys, xtype, ytype)
 
         df = element.interface.as_dframe(element)
+        if xtype == 'datetime' or ytype == 'datetime':
+            df = df.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')
@@ -843,9 +845,9 @@ def _process(self, element, key=None):
 
         xdim, ydim = list(agg.dims)[:2][::-1]
         if xtype == "datetime":
-            agg[xdim] = (agg[xdim]/1e3).astype('datetime64[us]')
+            agg[xdim] = agg[xdim].astype('datetime64[ns]')
         if ytype == "datetime":
-            agg[ydim] = (agg[ydim]/1e3).astype('datetime64[us]')
+            agg[ydim] = agg[ydim].astype('datetime64[ns]')
 
         params['kdims'] = [xdim, ydim]
 
@@ -1017,9 +1019,9 @@ def _process(self, element, key=None):
 
             # Convert datetime coordinates
             if xtype == "datetime":
-                rarray[x.name] = (rarray[x.name]/1e3).astype('datetime64[us]')
+                rarray[x.name] = rarray[x.name].astype('datetime64[ns]')
             if ytype == "datetime":
-                rarray[y.name] = (rarray[y.name]/1e3).astype('datetime64[us]')
+                rarray[y.name] = rarray[y.name].astype('datetime64[ns]')
             regridded[vd] = rarray
         regridded = xr.Dataset(regridded)
 
@@ -1197,9 +1199,9 @@ def _process(self, element, key=None):
         info = self._get_sampling(element, x, y)
         (x_range, y_range), (xs, ys), (width, height), (xtype, ytype) = info
         if xtype == 'datetime':
-            data[x.name] = data[x.name].astype('datetime64[us]').astype('int64')
+            data[x.name] = data[x.name].astype('datetime64[ns]').astype('int64')
         if ytype == 'datetime':
-            data[y.name] = data[y.name].astype('datetime64[us]').astype('int64')
+            data[y.name] = data[y.name].astype('datetime64[ns]').astype('int64')
 
         # Compute bounds (converting datetimes)
         ((x0, x1), (y0, y1)), (xs, ys) = self._dt_transform(
@@ -1218,9 +1220,9 @@ def _process(self, element, key=None):
         agg = cvs.quadmesh(data[vdim], x.name, y.name, agg_fn)
         xdim, ydim = list(agg.dims)[:2][::-1]
         if xtype == "datetime":
-            agg[xdim] = (agg[xdim]/1e3).astype('datetime64[us]')
+            agg[xdim] = agg[xdim].astype('datetime64[ns]')
         if ytype == "datetime":
-            agg[ydim] = (agg[ydim]/1e3).astype('datetime64[us]')
+            agg[ydim] = agg[ydim].astype('datetime64[ns]')
 
         return Image(agg, **params)
 

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("2.0"),
+        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("2.0"),
+        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

tox.ini

@@ -28,10 +28,7 @@ commands = pytest holoviews --cov=./holoviews -vv
 [_examples]
 description = Test that default examples run
 deps = .[examples_tests]
-commands = pytest -n auto --dist loadscope --nbval-lax examples --force-flaky --max-runs=5 -k "not Plotting_with_Bokeh and not 17-Dashboards and not Plots_and_Renderers"
-# 'Plotting_with_Bokeh' needs selenium, phantomjs, firefox and geckodriver to save a png picture.
-# '17-Dashboards' can give a timeout error.
-# 'Plots_and_Renderers' can give file not found here.
+commands = pytest -n auto --dist loadscope --nbval-lax examples --force-flaky --max-runs=5
 
 [_all_recommended]
 description = Run all recommended tests