Fix the CI (#529)

.github/workflows/docs.yaml

@@ -31,7 +31,7 @@ jobs:
     env:
       DESC: "Documentation build"
     steps:
-      - uses: holoviz-dev/holoviz_tasks/install@v0.1a15
+      - uses: holoviz-dev/holoviz_tasks/install@v0
         with:
           name: Documentation
           python-version: "3.10"

.github/workflows/test.yaml

@@ -5,10 +5,23 @@ on:
       - main
   pull_request:
     branches:
-    - '*'
+      - "*"
   workflow_dispatch:
+    inputs:
+      target:
+        description: "How much of the test suite to run"
+        type: choice
+        default: default
+        options:
+          - default
+          - full
+          - downstream
+      cache:
+        description: "Use cache"
+        type: boolean
+        default: true
   schedule:
-    - cron: '0 17 * * SUN'
+    - cron: "0 17 * * SUN"
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
@@ -17,33 +30,75 @@ concurrency:
 jobs:
   pre_commit:
     name: Run pre-commit
-    runs-on: 'ubuntu-latest'
+    runs-on: "ubuntu-latest"
     steps:
-      - uses: holoviz-dev/holoviz_tasks/[email protected]
+      - uses: holoviz-dev/holoviz_tasks/pre-commit@v0
+
+  setup:
+    name: Setup workflow
+    runs-on: ubuntu-latest
+    outputs:
+      matrix: ${{ env.MATRIX }}
+    steps:
+      - name: Set matrix option
+        run: |
+          if [[ '${{ github.event_name }}' == 'workflow_dispatch' ]]; then
+            OPTION=${{ github.event.inputs.target }}
+          elif [[ '${{ github.event_name }}' == 'schedule' ]]; then
+            OPTION="full"
+          elif [[ '${{ github.event_name }}' == 'push' && '${{ github.ref_type }}' == 'tag' ]]; then
+            OPTION="full"
+          else
+            OPTION="default"
+          fi
+          echo "MATRIX_OPTION=$OPTION" >> $GITHUB_ENV
+      - name: Set test matrix with 'default' option
+        if: env.MATRIX_OPTION == 'default'
+        run: |
+          MATRIX=$(jq -nsc '{
+              "os": ["ubuntu-latest", "macos-latest", "windows-latest"],
+              "python-version": ["3.9", "3.12"]
+          }')
+          echo "MATRIX=$MATRIX" >> $GITHUB_ENV
+      - name: Set test matrix with 'full' option
+        if: env.MATRIX_OPTION == 'full'
+        run: |
+          MATRIX=$(jq -nsc '{
+              "os": ["ubuntu-latest", "macos-latest", "windows-latest"],
+              "python-version": ["3.9", "3.10", "3.11", "3.12"]
+          }')
+          echo "MATRIX=$MATRIX" >> $GITHUB_ENV
+      - name: Set test matrix with 'downstream' option
+        if: env.MATRIX_OPTION == 'downstream'
+        run: |
+          MATRIX=$(jq -nsc '{
+              "os": ["ubuntu-latest"],
+              "python-version": ["3.11"]
+          }')
+          echo "MATRIX=$MATRIX" >> $GITHUB_ENV
+
   unit_test_suite:
-    name: Pytest on ${{ matrix.python-version }}, ${{ matrix.os }}
-    needs: [pre_commit]
+    name: Unit tests on Python ${{ matrix.python-version }}, ${{ matrix.os }}
+    needs: [pre_commit, setup]
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
-      matrix:
-        os: ['ubuntu-latest', 'macos-latest', 'windows-latest']
-        python-version: ${{ ( github.event_name == 'schedule' || github.event_name == 'workflow_dispatch' || ( github.event_name == 'push' && github.ref_type == 'tag' ) ) && fromJSON('["3.9", "3.10", "3.11"]') || fromJSON('["3.9", "3.10", "3.11"]') }}
+      matrix: ${{ fromJson(needs.setup.outputs.matrix) }}
     timeout-minutes: 60
     defaults:
       run:
         shell: bash -l {0}
     env:
       GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
     steps:
-      - uses: holoviz-dev/holoviz_tasks/install@v0.1a18
+      - uses: holoviz-dev/holoviz_tasks/install@v0
         with:
           name: unit_test_suite
           python-version: ${{ matrix.python-version }}
           channel-priority: flexible
           channels: pyviz/label/dev,bokeh,conda-forge,nodefaults
           envs: -o tests -o sql -o tests_ci
-          cache: true
+          cache: ${{ github.event.inputs.cache || github.event.inputs.cache == '' }}
         id: install
       - name: doit test_unit
         run: |
@@ -53,15 +108,16 @@ jobs:
         run: |
           conda activate test-environment
           codecov
+
   core_test_suite:
     name: Core tests on Python ${{ matrix.python-version }}, ${{ matrix.os }}
     needs: [pre_commit]
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
-        os: ['ubuntu-latest']
-        python-version: ['3.12']
+        os: ["ubuntu-latest"]
+        python-version: ["3.12"]
     timeout-minutes: 120
     defaults:
       run:
@@ -70,28 +126,17 @@ jobs:
       DESC: "Python ${{ matrix.python-version }}, ${{ matrix.os }} core tests"
       PYTHON_VERSION: ${{ matrix.python-version }}
     steps:
-      # Add back when this works on Python 3.12
-      # - uses: holoviz-dev/holoviz_tasks/[email protected]
-      #   with:
-      #     name: core_test_suite
-      #     python-version: ${{ matrix.python-version }}
-      #     # channel-priority: strict
-      #     channels: pyviz/label/dev,conda-forge,nodefaults
-      #     envs: "-o tests_core -o tests_ci"
-      #     cache: true
-      #     conda-update: true
-      #     id: install
-      - uses: actions/checkout@v3
-        with:
-          fetch-depth: "100"
-      - name: Fetch unshallow
-        run: git fetch --prune --tags --unshallow -f
-      - uses: actions/setup-python@v4
+      - uses: holoviz-dev/holoviz_tasks/install@v0
         with:
-          python-version: 3.12
-      - run: |
-          python -m pip install -ve '.[tests, tests_ci]'
+          name: core_test_suite
+          python-version: ${{ matrix.python-version }}
+          # channel-priority: strict
+          channels: pyviz/label/dev,conda-forge,nodefaults
+          envs: "-o tests -o tests_ci"
+          cache: ${{ github.event.inputs.cache || github.event.inputs.cache == '' }}
+          conda-update: true
+        id: install
       - name: doit test_unit
         run: |
-          # conda activate test-environment
+          conda activate test-environment
           pytest lumen

lumen/tests/conftest.py

@@ -1,3 +1,4 @@
+import contextlib
 import os
 import tempfile
 
@@ -8,13 +9,20 @@
 import pytest
 
 from bokeh.document import Document  # type: ignore
+from hvplot.tests.util import makeMixedDataFrame
 
 from lumen.config import config
 from lumen.sources.base import FileSource, Source
 from lumen.state import state
 from lumen.variables.base import Variables
 
-pd.set_option('mode.string_storage', 'pyarrow')
+with contextlib.suppress(Exception):
+    # From Dask 2024.3.0 they now use `dask_expr` by default
+    # https://github.com/dask/dask/issues/10995
+    import dask
+
+    dask.config.set({"dataframe.query-planning": False})
+
 
 @pytest.fixture
 def set_root():
@@ -44,11 +52,11 @@ def make_variable_filesource():
     def create(root, **kwargs):
         config._root = root
         state._variable = Variables.from_spec({'tables': {'type': 'constant', 'default': {'test': 'test.csv'}}})
-        source = Source.from_spec(dict({
+        source = Source.from_spec({
             'type': 'file',
             'tables': '$variables.tables',
             'kwargs': {'parse_dates': ['D']}
-        }))
+        })
         state.sources['original'] = source
         return source
     yield create
@@ -61,14 +69,20 @@ def create(root, **kwargs):
 @pytest.fixture
 def mixed_df():
     pytest.importorskip("pyarrow", "7.0", "Pyarrow is not out on Python 3.12 yet")
-    df = pd._testing.makeMixedDataFrame()
+    string = pd.get_option('mode.string_storage')
+    pd.set_option('mode.string_storage', 'pyarrow')
+    df = makeMixedDataFrame()
     df['C'] = df.C.astype("string")
     yield df
+    pd.set_option('mode.string_storage', string)
 
 @pytest.fixture
 def mixed_df_object_type():
-    df = pd._testing.makeMixedDataFrame()
+    string = pd.get_option('mode.string_storage')
+    pd.set_option('mode.string_storage', 'python')
+    df = makeMixedDataFrame()
     yield df
+    pd.set_option('mode.string_storage', string)
 
 @pytest.fixture
 def yaml_file():

lumen/tests/sources/conftest.py

@@ -1,17 +1,21 @@
 import pandas as pd
 import pytest
 
-pd.set_option('mode.string_storage', 'pyarrow')
+from hvplot.tests.util import makeMixedDataFrame
+
 
 @pytest.fixture
 def source_tables():
-    df_test = pd._testing.makeMixedDataFrame()
-    df_test_sql = pd._testing.makeMixedDataFrame()
-    df_test_sql_none = pd._testing.makeMixedDataFrame()
+    string = pd.get_option('mode.string_storage')
+    pd.set_option('mode.string_storage', 'pyarrow')
+    df_test = makeMixedDataFrame()
+    df_test_sql = makeMixedDataFrame()
+    df_test_sql_none = makeMixedDataFrame()
     df_test_sql_none['C'] = ['foo1', None, 'foo3', None, 'foo5']
     tables = {
         'test': df_test,
         'test_sql': df_test_sql,
         'test_sql_with_none': df_test_sql_none,
     }
-    return tables
+    yield tables
+    pd.set_option('mode.string_storage', string)

lumen/tests/sources/test_base.py

@@ -6,6 +6,8 @@
 import pandas as pd
 import pytest
 
+from hvplot.tests.util import makeMixedDataFrame
+
 from lumen.sources.base import Source
 from lumen.state import state
 from lumen.transforms.sql import SQLLimit
@@ -155,7 +157,7 @@ def test_file_source_get_query_cache_to_file(make_filesource, cachedir):
     df.index.names = [None]
     pd.testing.assert_frame_equal(
         df,
-        pd._testing.makeMixedDataFrame().iloc[1:3]
+        makeMixedDataFrame().iloc[1:3]
     )
 
 

lumen/tests/sources/test_duckdb.py

@@ -1,7 +1,6 @@
 import datetime as dt
 import os
 
-import numpy as np
 import pandas as pd
 import pytest
 
@@ -15,26 +14,6 @@
 pytestmark = pytest.mark.skipif(DuckDBSource is None, reason="Duckdb is not installed")
 
 
-def assert_frame_equal_ignore_null_like(a, b):
-    """
-    From Pandas 2.1 we are getting this FutureWarning:
-        Mismatched null-like values nan and None found. In a future version,
-        pandas equality-testing functions (e.g. assert_frame_equal)
-        will consider these not-matching and raise.
-
-    Here we are converting all null-like values to np.nan.
-
-    Could be that in future release we get a keyword argument to ignore
-    this strict behavior
-
-    Reference: https://github.com/pandas-dev/pandas/pull/52081
-
-    """
-    a = a.fillna(np.nan)
-    b = b.fillna(np.nan)
-    pd.testing.assert_frame_equal(a, b)
-
-
 @pytest.fixture
 def duckdb_source():
     root = os.path.dirname(__file__)
@@ -63,7 +42,7 @@ def test_duckdb_get_tables(duckdb_source, source_tables):
     tables = duckdb_source.get_tables()
     assert not len(set(tables) - set(source_tables.keys()))
     for table in tables:
-        assert_frame_equal_ignore_null_like(
+        pd.testing.assert_frame_equal(
             duckdb_source.get(table),
             source_tables[table],
         )
@@ -131,9 +110,10 @@ def test_duckdb_filter(duckdb_source, table_column_value_type, dask, expected_fi
     table, column, value, _ = table_column_value_type
     kwargs = {column: value}
     filtered = duckdb_source.get(table, __dask=dask, **kwargs)
-    assert_frame_equal_ignore_null_like(filtered, expected_filtered_df.reset_index(drop=True))
+    pd.testing.assert_frame_equal(filtered, expected_filtered_df.reset_index(drop=True))
 
 
+@pytest.mark.flaky(reruns=3)
 def test_duckdb_transforms(duckdb_source, source_tables):
     df_test_sql = source_tables['test_sql']
     transforms = [SQLGroupBy(by=['B'], aggregates={'SUM': 'A'})]
@@ -142,6 +122,7 @@ def test_duckdb_transforms(duckdb_source, source_tables):
     pd.testing.assert_frame_equal(transformed, expected)
 
 
+@pytest.mark.flaky(reruns=3)
 def test_duckdb_transforms_cache(duckdb_source, source_tables):
     df_test_sql = source_tables['test_sql']
     transforms = [SQLGroupBy(by=['B'], aggregates={'SUM': 'A'})]

setup.py

@@ -31,18 +31,19 @@ def get_setup_version(reponame):
     'sql': [
         'duckdb',
         'intake-sql',
-        'sqlalchemy <2',  # Don't work with pandas yet
+        'sqlalchemy',
     ],
     'tests': [
         'pytest',
+        'pytest-rerunfailures',
         'flake8',
         'intake',
         'fastparquet',
         'msgpack-python',
         'toolz',
         'pre-commit',
         'matplotlib >=3.4',  # Ubuntu + Python 3.9 installs old version matplotlib (3.3.2)
-        'pandas <2.2',
+        'pandas',
     ],
     'tests_ci' : [
         'pytest-github-actions-annotate-failures',