RuntimeWarning: invalid value encountered in cast. (#1380)

* fixing error by masking nans before converting to timedeltas

* fix tests

* trying to fix test again

sdv/constraints/tabular.py

@@ -40,7 +40,7 @@
 from sdv.constraints.errors import (
     AggregateConstraintsError, ConstraintMetadataError, FunctionError, InvalidFunctionError)
 from sdv.constraints.utils import cast_to_datetime64, logit, matches_datetime_format, sigmoid
-from sdv.utils import is_datetime_type
+from sdv.utils import convert_to_timedelta, is_datetime_type
 
 INEQUALITY_TO_OPERATION = {
     '>': np.greater,
@@ -512,8 +512,8 @@ def _reverse_transform(self, table_data):
 
         The transformation is reversed by computing an exponential of the difference value,
         subtracting 1 and converting it to the original dtype. Finally, the obtained column
-        is added to the ``low_column_name`` column to get back the original
-        ``high_column_name`` value.
+        is added to the ``low_column_name`` column to get back the original ``high_column_name``
+        value.
 
         Args:
             table_data (pandas.DataFrame):
@@ -528,7 +528,7 @@ def _reverse_transform(self, table_data):
             diff_column = diff_column.round()
 
         if self._is_datetime:
-            diff_column = pd.to_timedelta(diff_column)
+            diff_column = convert_to_timedelta(diff_column)
 
         low = table_data[self._low_column_name].to_numpy()
         if self._is_datetime and self._dtype == 'O':

sdv/utils.py

@@ -15,44 +15,6 @@ def cast_to_iterable(value):
     return [value]
 
 
-def display_tables(tables, max_rows=10, datetime_fmt='%Y-%m-%d %H:%M:%S', row=True):
-    """Display mutiple tables side by side on a Jupyter Notebook.
-
-    Args:
-        tables (dict[str, DataFrame]):
-            ``dict`` containing table names and pandas DataFrames.
-        max_rows (int):
-            Max rows to show per table. Defaults to 10.
-        datetime_fmt (str):
-            Format with which to display datetime columns.
-    """
-    # Import here to avoid making IPython a hard dependency
-    from IPython.core.display import HTML
-
-    names = []
-    data = []
-    for name, table in tables.items():
-        table = table.copy()
-        for column in table.columns:
-            column_data = table[column]
-            if column_data.dtype.kind == 'M':
-                table[column] = column_data.dt.strftime(datetime_fmt)
-
-        names.append(f'<td style="text-align:left"><b>{name}</b></td>')
-        data.append(f'<td>{table.head(max_rows).to_html(index=False)}</td>')
-
-    if row:
-        html = f"<table><tr>{''.join(names)}</tr><tr>{''.join(data)}</tr></table>"
-    else:
-        rows = [
-            f'<tr>{name}</tr><tr>{table}</tr>'
-            for name, table in zip(names, data)
-        ]
-        html = f"<table>{''.join(rows)}</table>"
-
-    return HTML(html)
-
-
 def get_datetime_format(value):
     """Get the ``strftime`` format for a given ``value``.
 
@@ -150,6 +112,27 @@ def validate_datetime_format(value, datetime_format):
     return True
 
 
+def convert_to_timedelta(column):
+    """Convert a ``pandas.Series`` to one with dtype ``timedelta``.
+
+    ``pd.to_timedelta`` does not handle nans, so this function masks the nans, converts and then
+    reinserts them.
+
+    Args:
+        column (pandas.Series):
+            Column to convert.
+
+    Returns:
+        pandas.Series:
+            The column converted to timedeltas.
+    """
+    nan_mask = pd.isna(column)
+    column[nan_mask] = 0
+    column = pd.to_timedelta(column)
+    column[nan_mask] = pd.NaT
+    return column
+
+
 def load_data_from_csv(filepath, pandas_kwargs=None):
     """Load DataFrame from a filepath.
 

tests/unit/test_utils.py

@@ -0,0 +1,31 @@
+from unittest.mock import patch
+
+import numpy as np
+import pandas as pd
+
+from sdv.utils import convert_to_timedelta
+from tests.utils import SeriesMatcher
+
+
+@patch('sdv.utils.pd.to_timedelta')
+def test_convert_to_timedelta(to_timedelta_mock):
+    """Test that nans and values are properly converted to timedeltas."""
+    # Setup
+    column = pd.Series([7200, 3600, np.nan])
+    to_timedelta_mock.return_value = pd.Series([
+        pd.Timedelta(hours=1),
+        pd.Timedelta(hours=2),
+        pd.Timedelta(hours=0)
+    ], dtype='timedelta64[ns]')
+
+    # Run
+    converted_column = convert_to_timedelta(column)
+
+    # Assert
+    to_timedelta_mock.assert_called_with(SeriesMatcher(pd.Series([7200, 3600, 0.0])))
+    expected_column = pd.Series([
+        pd.Timedelta(hours=1),
+        pd.Timedelta(hours=2),
+        pd.NaT
+    ], dtype='timedelta64[ns]')
+    pd.testing.assert_series_equal(converted_column, expected_column)

tests/utils.py

@@ -15,6 +15,17 @@ def __eq__(self, other):
         return True
 
 
+class SeriesMatcher:
+    """Match a given Pandas Series in a mock function call."""
+
+    def __init__(self, series):
+        self.series = series
+
+    def __eq__(self, other):
+        pd.testing.assert_series_equal(self.series, other)
+        return True
+
+
 def get_multi_table_metadata():
     """Return a ``MultiTableMetadata`` object to be used with tests."""
     dict_metadata = {