Support the ability to pass in None for both get_column_plot and get_column_pair_plot (#2344)

Author: R-Palazzo

sdv/evaluation/_utils.py

@@ -0,0 +1,35 @@
+import pandas as pd
+
+
+def _prepare_data_visualization(data, metadata, column_names, sample_size):
+    """Prepare the data for a plot.
+
+    Args:
+        data (pd.DataFrame or None):
+            The data to be prepared.
+        metadata (Metadata):
+            The metadata of the data.
+        column_names (str or list[str]):
+            The column names to plot.
+        sample_size (int or None):
+            The number of samples to plot. If ``None``, use the whole dataset.
+
+    Returns:
+        pd.DataFrame or None:
+            The prepared data.
+    """
+    if data is None:
+        return None
+
+    col_names = column_names if isinstance(column_names, list) else [column_names]
+    data = data.copy()
+    for column_name in col_names:
+        sdtype = metadata.columns[column_name]['sdtype']
+        if sdtype == 'datetime':
+            datetime_format = metadata.columns[column_name].get('datetime_format')
+            data[column_name] = pd.to_datetime(data[column_name], format=datetime_format)
+
+    if sample_size and sample_size < len(data):
+        data = data.sample(n=sample_size)
+
+    return data

sdv/evaluation/multi_table.py

@@ -94,9 +94,9 @@ def get_column_pair_plot(
     """Get a plot of the real and synthetic data for a given column pair.
 
     Args:
-        real_data (dict):
+        real_data (dict or None):
             Dictionary containing the real table data.
-        synthetic_column (dict):
+        synthetic_column (dict or None):
             Dictionary containing the synthetic table data.
         metadata (Metadata):
             Metadata describing the data.

sdv/evaluation/single_table.py

@@ -1,11 +1,11 @@
 """Methods to compare the real and synthetic data for single-table."""
 
-import pandas as pd
 from sdmetrics import visualization
 from sdmetrics.reports.single_table.diagnostic_report import DiagnosticReport
 from sdmetrics.reports.single_table.quality_report import QualityReport
 
 from sdv.errors import VisualizationUnavailableError
+from sdv.evaluation._utils import _prepare_data_visualization
 from sdv.metadata.metadata import Metadata
 
 
@@ -68,9 +68,9 @@ def get_column_plot(real_data, synthetic_data, metadata, column_name, plot_type=
     """Get a plot of the real and synthetic data for a given column.
 
     Args:
-        real_data (pandas.DataFrame):
+        real_data (pandas.DataFrame or None):
             The real table data.
-        synthetic_data (pandas.DataFrame):
+        synthetic_data (pandas.DataFrame or None):
             The synthetic table data.
         metadata (Metadata):
             The table metadata.
@@ -103,14 +103,8 @@ def get_column_plot(real_data, synthetic_data, metadata, column_name, plot_type=
                 "'plot_type'."
             )
 
-    if sdtype == 'datetime':
-        datetime_format = metadata.columns.get(column_name).get('datetime_format')
-        real_data = pd.DataFrame({
-            column_name: pd.to_datetime(real_data[column_name], format=datetime_format)
-        })
-        synthetic_data = pd.DataFrame({
-            column_name: pd.to_datetime(synthetic_data[column_name], format=datetime_format)
-        })
+    real_data = _prepare_data_visualization(real_data, metadata, column_name, None)
+    synthetic_data = _prepare_data_visualization(synthetic_data, metadata, column_name, None)
 
     return visualization.get_column_plot(
         real_data, synthetic_data, column_name, plot_type=plot_type
@@ -147,8 +141,6 @@ def get_column_pair_plot(
     if isinstance(metadata, Metadata):
         metadata = metadata._convert_to_single_table()
 
-    real_data = real_data.copy()
-    synthetic_data = synthetic_data.copy()
     if plot_type is None:
         plot_type = []
         for column_name in column_names:
@@ -169,18 +161,9 @@ def get_column_pair_plot(
         else:
             plot_type = plot_type.pop()
 
-    for column_name in column_names:
-        sdtype = metadata.columns.get(column_name)['sdtype']
-        if sdtype == 'datetime':
-            datetime_format = metadata.columns.get(column_name).get('datetime_format')
-            real_data[column_name] = pd.to_datetime(real_data[column_name], format=datetime_format)
-            synthetic_data[column_name] = pd.to_datetime(
-                synthetic_data[column_name], format=datetime_format
-            )
-
-    require_subsample = sample_size and sample_size < min(len(real_data), len(synthetic_data))
-    if require_subsample:
-        real_data = real_data.sample(n=sample_size)
-        synthetic_data = synthetic_data.sample(n=sample_size)
+    real_data = _prepare_data_visualization(real_data, metadata, column_names, sample_size)
+    synthetic_data = _prepare_data_visualization(
+        synthetic_data, metadata, column_names, sample_size
+    )
 
     return visualization.get_column_pair_plot(real_data, synthetic_data, column_names, plot_type)

tests/unit/evaluation/test__utils.py

@@ -0,0 +1,36 @@
+import numpy as np
+import pandas as pd
+
+from sdv.evaluation._utils import _prepare_data_visualization
+from sdv.metadata import SingleTableMetadata
+
+
+def test__prepare_data_visualization():
+    """Test ``_prepare_data_visualization``."""
+    # Setup
+    np.random.seed(0)
+    metadata = SingleTableMetadata.load_from_dict({
+        'columns': {
+            'col1': {'sdtype': 'datetime', 'datetime_format': '%Y-%m-%d'},
+            'col2': {'sdtype': 'numerical'},
+        }
+    })
+    column_names = ['col1', 'col2']
+    sample_size = 2
+    data = pd.DataFrame({
+        'col1': ['2021-01-01', '2021-02-01', '2021-03-01'],
+        'col2': [4, 5, 6],
+    })
+
+    # Run
+    result = _prepare_data_visualization(data, metadata, column_names, sample_size)
+
+    # Assert
+    expected_result = pd.DataFrame(
+        {
+            'col1': pd.to_datetime(['2021-03-01', '2021-02-01']),
+            'col2': [6, 5],
+        },
+        index=[2, 1],
+    )
+    pd.testing.assert_frame_equal(result, expected_result)