Make dask an optional dependency

.circleci/config.yml

@@ -32,6 +32,8 @@ jobs:
     parameters:
       image_tag:
         type: string
+      optional_libraries:
+        type: string
     executor:
       name: python
       image_tag: << parameters.image_tag >>
@@ -45,8 +47,20 @@ jobs:
           source venv/bin/activate
           python -m pip config --site set global.progress_bar off
           python -m pip install --upgrade pip
-          python -m pip install -e unpacked_sdist/
-          python -m pip install -r unpacked_sdist/test-requirements.txt
+      - when:
+          condition:
+            equal: [ "optional", << parameters.optional_libraries >> ]
+          steps:
+            - run: |
+                python -m pip install -e unpacked_sdist/[dask]
+                python -m pip install -r unpacked_sdist/test-requirements.txt
+      - unless:
+          condition:
+            equal: ["optional", << parameters.optional_libraries >> ]
+          steps:
+            - run: |
+                python -m pip install -e unpacked_sdist/
+                python -m pip install -r unpacked_sdist/test-requirements.txt
       - persist_to_workspace:
           root: ~/woodwork
           paths:
@@ -58,6 +72,8 @@ jobs:
     parameters:
       image_tag:
         type: string
+      optional_libraries:
+        type: string
     executor:
       name: python
       image_tag: << parameters.image_tag >>
@@ -67,7 +83,9 @@ jobs:
           at: ~/woodwork
       - when:
           condition:
-            equal: [ "3.6", << parameters.image_tag >> ]
+            and:
+              - equal: [ "3.6", << parameters.image_tag >> ]
+              - equal: ["optional", << parameters.optional_libraries >>]
           steps:
             - run: |
                 source venv/bin/activate
@@ -81,7 +99,9 @@ jobs:
                 codecov --required
       - unless:
           condition:
-            equal: [ "3.6", << parameters.image_tag >> ]
+            and:
+              - equal: [ "3.6", << parameters.image_tag >> ]
+              - equal: ["optional", << parameters.optional_libraries >>]
           steps:
             - run: |
                 source venv/bin/activate
@@ -143,23 +163,25 @@ workflows:
           matrix:
             parameters:
               image_tag: ["3.8", "3.7", "3.6"]
-          name: << matrix.image_tag >> install woodwork
+              optional_libraries: ["minimal", "optional"]
+          name: << matrix.image_tag >> install woodwork << matrix.optional_libraries >>
           context: dockerhub
       - unit_tests:
           matrix:
             parameters:
               image_tag: ["3.8", "3.7", "3.6"]
-          name: << matrix.image_tag >> unit tests
+              optional_libraries: ["minimal", "optional"]              
+          name: << matrix.image_tag >> unit tests << matrix.optional_libraries >>
           requires:
-            - << matrix.image_tag >> install woodwork
+            - << matrix.image_tag >> install woodwork << matrix.optional_libraries >>
           context: dockerhub
       - lint_test:
           matrix:
             parameters:
               image_tag: ["3.8", "3.7", "3.6"]
           name: << matrix.image_tag >> lint test
           requires:
-            - << matrix.image_tag >> install woodwork
+            - << matrix.image_tag >> install woodwork optional
           context: dockerhub
       - release_notes_updated:
           name: "release notes updated"

dask-requirements.txt

@@ -0,0 +1 @@
+dask[dataframe]>=2.30.0

dev-requirements.txt

@@ -12,4 +12,5 @@ ipython==7.18.1; python_version>'3.6'
 pygments==2.7.0
 jupyter==1.0.0
 pandoc==1.0.2
-ipykernel==5.3.4
+ipykernel==5.3.4
+-r dask-requirements.txt

docs/source/guides/using_woodwork_with_dask.ipynb

@@ -8,6 +8,11 @@
     "\n",
     "Woodwork enables DataTables to be created from Dask DataFrames when working with datasets that are too large to easily fit in memory. Although creating a DataTable from a Dask DataFrame follows the same process as one would follow when creating a DataTable from a pandas DataFrame, there are a few limitations to be aware of. This guide will provide a brief overview of creating a DataTable starting with a Dask DataFrame, and will outline several key items to keep in mind when using a Dask DataFrame as input.\n",
     "\n",
+    "Dask DataTables require the installation of the Dask library, which can be installed directly with the following command:\n",
+    "```bash\n",
+    "pip install \"woodwork[dask]\"\n",
+    "```\n",
+    "\n",
     "First we will create a Dask DataFrame to use in our example. Normally you would create the DataFrame directly by reading in the data from saved files, but we will create it from a demo pandas DataFrame."
    ]
   },
@@ -162,7 +167,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.3"
+   "version": "3.6.8"
   }
  },
  "nbformat": 4,

docs/source/install.ipynb

@@ -16,6 +16,18 @@
     "python -m pip install woodwork\n",
     "```\n",
     "\n",
+    "Woodwork allows users to install add-ons individually or all at once. In order to install all add-ons, run:\n",
+    "```bash\n",
+    "python -m pip install \"woodwork[complete]\"\n",
+    "```\n",
+    "\n",
+    "You can use Woodwork to create Dask DataTables by running:\n",
+    "\n",
+    "```bash\n",
+    "python -m pip install \"woodwork[dask]\"\n",
+    "```\n",
+    "\n",
+    "\n",
     "\n",
     "## Conda\n",
     "\n",
@@ -25,6 +37,9 @@
     "conda install -c conda-forge woodwork\n",
     "```\n",
     "\n",
+    ".. note::\n",
+    "    In order to create Dask DataTables, the following command must be run prior to installing Woodwork with conda: `conda install dask`\n",
+    "\n",
     "\n",
     "\n",
     "## Source\n",
@@ -61,7 +76,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.5"
+   "version": "3.6.8"
   }
  },
  "nbformat": 4,

docs/source/release_notes.rst

@@ -25,6 +25,7 @@ Release Notes
         * Remove ``copy`` parameter from ``DataTable.to_dataframe`` and ``DataColumn.to_series`` (:pr:`338`)
         * Allow pandas ExtensionArrays as inputs to DataColumn (:pr:`343`)
         * Move warnings to a separate exceptions file and call via UserWarning subclasses (:pr:`348`)
+        * Make Dask an optional dependency installable with woodwork[dask] (:pr:`357`)
     * Documentation Changes
         * Create a guide for using Woodwork with Dask (:pr:`304`)
         * Add conda install instructions (:pr:`305`, :pr:`309`)

requirements.txt

@@ -1,6 +1,5 @@
 numpy>=1.19.1
 pandas>=1.1.0
 click>=7.1.2
-dask[dataframe]>=2.30.0
 scikit-learn>=0.21.3
 pyarrow>=2.0.0

setup.py

@@ -6,6 +6,8 @@
 with open(path.join(dirname, 'README.md')) as f:
     long_description = f.read()
 
+extras_require = {'dask': open('dask-requirements.txt').readlines()}
+extras_require['complete'] = sorted(set(sum(extras_require.values(), [])))
 
 setup(
     name='woodwork',
@@ -27,6 +29,7 @@
     install_requires=open('requirements.txt').readlines(),
     tests_require=open('test-requirements.txt').readlines(),
     python_requires='>=3.6, <4',
+    extras_require=extras_require,
     keywords='data science machine learning typing',
     include_package_data=True,
     entry_points={

woodwork/data_column.py

@@ -1,6 +1,5 @@
 import warnings
 
-import dask.dataframe as dd
 import pandas as pd
 import pandas.api.types as pdtypes
 
@@ -26,9 +25,12 @@
 from woodwork.utils import (
     _convert_input_to_set,
     _get_ltype_class,
-    col_is_datetime
+    col_is_datetime,
+    import_or_none
 )
 
+dd = import_or_none('dask.dataframe')
+
 
 class DataColumn(object):
     def __init__(self, series,
@@ -94,7 +96,7 @@ def _update_dtype(self):
             # Update the underlying series
             try:
                 if _get_ltype_class(self.logical_type) == Datetime:
-                    if isinstance(self._series, dd.Series):
+                    if dd and isinstance(self._series, dd.Series):
                         name = self._series.name
                         self._series = dd.to_datetime(self._series, format=self.logical_type.datetime_format)
                         self._series.name = name
@@ -132,7 +134,9 @@ def set_logical_type(self, logical_type, retain_index_tags=True):
         return new_col
 
     def _set_series(self, series):
-        if not (isinstance(series, pd.Series) or isinstance(series, dd.Series) or isinstance(series, pd.api.extensions.ExtensionArray)):
+        if not (isinstance(series, pd.Series) or
+                (dd and isinstance(series, dd.Series)) or
+                isinstance(series, pd.api.extensions.ExtensionArray)):
             raise TypeError('Series must be a pandas Series, Dask Series, or a pandas ExtensionArray')
 
         # pandas ExtensionArrays should be converted to pandas.Series
@@ -323,7 +327,7 @@ def infer_logical_type(series):
     Args:
         series (pd.Series): Input Series
     """
-    if isinstance(series, dd.Series):
+    if dd and isinstance(series, dd.Series):
         series = series.get_partition(0).compute()
     natural_language_threshold = config.get_option('natural_language_threshold')
 

woodwork/data_table.py

@@ -1,6 +1,5 @@
 import warnings
 
-import dask.dataframe as dd
 import pandas as pd
 from sklearn.metrics.cluster import normalized_mutual_info_score
 
@@ -19,9 +18,12 @@
     _get_ltype_class,
     _get_mode,
     _is_numeric_series,
-    col_is_datetime
+    col_is_datetime,
+    import_or_none
 )
 
+dd = import_or_none('dask.dataframe')
+
 
 class DataTable(object):
     def __init__(self, dataframe,
@@ -72,7 +74,7 @@ def __init__(self, dataframe,
             self._dataframe = dataframe
 
         if make_index:
-            if isinstance(self._dataframe, dd.DataFrame):
+            if dd and isinstance(self._dataframe, dd.DataFrame):
                 self._dataframe[index] = 1
                 self._dataframe[index] = self._dataframe[index].cumsum() - 1
             else:
@@ -545,7 +547,7 @@ def describe(self, include=None):
 
         results = {}
 
-        if isinstance(self._dataframe, dd.DataFrame):
+        if dd and isinstance(self._dataframe, dd.DataFrame):
             df = self._dataframe.compute()
         else:
             df = self._dataframe
@@ -628,7 +630,7 @@ def value_counts(self, ascending=False, top_n=10, dropna=False):
         val_counts = {}
         valid_cols = [col for col, column in self.columns.items() if column._is_categorical()]
         data = self._dataframe[valid_cols]
-        if isinstance(data, dd.DataFrame):
+        if dd and isinstance(data, dd.DataFrame):
             data = data.compute()
 
         for col in valid_cols:
@@ -725,7 +727,7 @@ def get_mutual_information(self, num_bins=10, nrows=None):
                                                       _get_ltype_class(column.logical_type) == Boolean)
                                                      )]
         data = self._dataframe[valid_columns]
-        if isinstance(data, dd.DataFrame):
+        if dd and isinstance(data, dd.DataFrame):
             data = data.compute()
 
         # cut off data if necessary
@@ -814,7 +816,8 @@ def to_parquet(self, path, compression=None, profile_name=None):
 
 def _validate_params(dataframe, name, index, time_index, logical_types, semantic_tags, make_index):
     """Check that values supplied during DataTable initialization are valid"""
-    if not isinstance(dataframe, (pd.DataFrame, dd.DataFrame)):
+    if not ((dd and isinstance(dataframe, dd.DataFrame)) or
+            isinstance(dataframe, pd.DataFrame)):
         raise TypeError('Dataframe must be one of: pandas.DataFrame, dask.DataFrame')
     _check_unique_column_names(dataframe)
     if name and not isinstance(name, str):

woodwork/deserialize.py

@@ -6,15 +6,14 @@
 from itertools import zip_longest
 from pathlib import Path
 
-import dask.dataframe as dd
 import pandas as pd
 
 from woodwork import DataTable
 from woodwork.exceptions import OutdatedSchemaWarning, UpgradeSchemaWarning
 from woodwork.logical_types import str_to_logical_type
 from woodwork.s3_utils import get_transport_params, use_smartopen
 from woodwork.serialize import FORMATS, SCHEMA_VERSION
-from woodwork.utils import _is_s3, _is_url
+from woodwork.utils import _is_s3, _is_url, import_or_raise
 
 
 def read_table_metadata(path):
@@ -60,7 +59,13 @@ def metadata_to_datatable(table_metadata, **kwargs):
     table_type = loading_info.get('table_type', 'pandas')
 
     if table_type == 'dask':
-        lib = dd
+        DASK_ERR_MSG = (
+            'Cannot load Dask DataTable - unable to import Dask.\n\n'
+            'Please install with pip or conda:\n\n'
+            'python -m pip install "woodwork[dask]"\n\n'
+            'conda install dask'
+        )
+        lib = import_or_raise('dask.dataframe', DASK_ERR_MSG)
     else:
         lib = pd
 

woodwork/serialize.py

@@ -4,16 +4,17 @@
 import tarfile
 import tempfile
 
-import dask.dataframe as dd
-
 from woodwork.s3_utils import get_transport_params, use_smartopen
 from woodwork.utils import (
     _get_ltype_class,
     _get_specified_ltype_params,
     _is_s3,
-    _is_url
+    _is_url,
+    import_or_none
 )
 
+dd = import_or_none('dask.dataframe')
+
 SCHEMA_VERSION = '1.0.0'
 FORMATS = ['csv', 'pickle', 'parquet']
 
@@ -40,7 +41,7 @@ def datatable_to_metadata(datatable):
         for col in datatable.columns.values()
     ]
 
-    if isinstance(df, dd.DataFrame):
+    if dd and isinstance(df, dd.DataFrame):
         table_type = 'dask'
     else:
         table_type = 'pandas'
@@ -114,7 +115,7 @@ def write_table_data(datatable, path, format='csv', **kwargs):
     dt_name = datatable.name or 'data'
     df = datatable.to_dataframe()
 
-    if isinstance(df, dd.DataFrame) and format == 'csv':
+    if dd and isinstance(df, dd.DataFrame) and format == 'csv':
         basename = "{}-*.{}".format(dt_name, format)
     else:
         basename = '.'.join([dt_name, format])
@@ -131,7 +132,7 @@ def write_table_data(datatable, path, format='csv', **kwargs):
         )
     elif format == 'pickle':
         # Dask currently does not support to_pickle
-        if isinstance(df, dd.DataFrame):
+        if dd and isinstance(df, dd.DataFrame):
             msg = 'Cannot serialize Dask DataTable to pickle'
             raise ValueError(msg)
         df.to_pickle(file, **kwargs)