Independent Sampler always create a sample size of 1 no matter how sm…

…all the scale (#2102)

sdv/sampling/independent_sampler.py

@@ -1,6 +1,7 @@
 """Independent Samplers."""
 
 import logging
+import warnings
 
 LOGGER = logging.getLogger(__name__)
 
@@ -144,8 +145,12 @@ def _sample(self, scale=1.0):
                 sampled data tables as ``pandas.DataFrame``.
         """
         sampled_data = {}
+        send_min_sample_warning = False
         for table in self.metadata.tables:
             num_rows = int(self._table_sizes[table] * scale)
+            if num_rows <= 0:
+                send_min_sample_warning = True
+                num_rows = 1
             synthesizer = self._table_synthesizers[table]
             self._sample_table(
                 synthesizer=synthesizer,
@@ -154,5 +159,12 @@ def _sample(self, scale=1.0):
                 sampled_data=sampled_data,
             )
 
+        if send_min_sample_warning:
+            warn_msg = (
+                "The 'scale' parameter is too small. Some tables may have 1 row."
+                ' For better quality data, please choose a larger scale.'
+            )
+            warnings.warn(warn_msg)
+
         self._connect_tables(sampled_data)
         return self._finalize(sampled_data)

tests/unit/sampling/test_independent_sampler.py

@@ -1,3 +1,4 @@
+import re
 from unittest.mock import Mock, call, patch
 
 import numpy as np
@@ -404,3 +405,36 @@ def _connect_tables(sampled_data):
             'transactions': DataFrameMatcher(connected_transactions),
         })
         assert result == instance._finalize.return_value
+
+    def test__sample_too_small(self):
+        instance = Mock()
+        metadata = Mock()
+        metadata.tables = {
+            'guests': {
+                'METADATA_SPEC_VERSION': 'SINGLE_TABLE_V1',
+                'primary_key': 'key_1',
+                'columns': {
+                    'key_1': {
+                        'sdtype': 'numerical',
+                    }
+                },
+            },
+            'hotels': {
+                'METADATA_SPEC_VERSION': 'SINGLE_TABLE_V1',
+                'primary_key': 'key_2',
+                'columns': {
+                    'key_2': {
+                        'sdtype': 'numerical',
+                    }
+                },
+            },
+        }
+        instance._table_sizes = {'hotels': 10, 'guests': 658}
+        instance._table_synthesizers = {'hotels': Mock(), 'guests': Mock()}
+        instance.metadata = metadata
+        warning_msg = re.escape(
+            "The 'scale' parameter is too small. Some tables may have 1 row."
+            ' For better quality data, please choose a larger scale.'
+        )
+        with pytest.warns(Warning, match=warning_msg):
+            BaseIndependentSampler._sample(instance, 0.01)