Merge branch 'main' into issue-2719-context-order-par-synthesizer

Author: sarahmish

HISTORY.md

@@ -1,5 +1,25 @@
 # Release Notes
 
+## v1.28.0 - 2025-10-17
+
+### New Features
+
+* Unable to validate just 1 table of a multi-table schema - Issue [#2678](https://github.com/sdv-dev/SDV/issues/2678) by @frances-h
+* Allow users to validate the DayZ parameters - Issue [#2667](https://github.com/sdv-dev/SDV/issues/2667) by @frances-h
+* Allow users to estimate parameters for DayZSynthesizer - Issue [#2666](https://github.com/sdv-dev/SDV/issues/2666) by @R-Palazzo
+
+### Bugs Fixed
+
+* Minimum tests failing - OSError: [WinError 1114] A dynamic link library (DLL) initialization routine failed - Issue [#2725](https://github.com/sdv-dev/SDV/issues/2725) by @amontanez24
+* [DayZ Parameters] `'missing_values_proportion'` must be zero for any key columns - Issue [#2708](https://github.com/sdv-dev/SDV/issues/2708) by @frances-h
+* [DayZ Parameters] Validation results in unexpected errors for some edge cases - Issue [#2703](https://github.com/sdv-dev/SDV/issues/2703) by @fealho
+* [DayZ Parameters] `create_parameters` should fall back to default parameters if parameters cannot be detected - Issue [#2702](https://github.com/sdv-dev/SDV/issues/2702) by @fealho
+* [DayZ Parameters] DayZ parameter validation does not validate DAYZ_SPEC_VERSION - Issue [#2701](https://github.com/sdv-dev/SDV/issues/2701) by @R-Palazzo
+* [DayZParameters] `KeyError` when creating parameters with empty data and metadata - Issue [#2700](https://github.com/sdv-dev/SDV/issues/2700) by @fealho
+* Unable to load the DayZSynthesizer after saving it - Issue [#2698](https://github.com/sdv-dev/SDV/issues/2698) by @R-Palazzo
+* `DayZSynthesizer.create_parameters` errors in Colab with numeric columns - Issue [#2683](https://github.com/sdv-dev/SDV/issues/2683) by @frances-h
+* PARSynthesizer: `FutureWarnings` in `groupby.apply` and `Series.__getitem__` from pandas - Issue [#2682](https://github.com/sdv-dev/SDV/issues/2682) by @R-Palazzo
+
 ## v1.27.0 - 2025-09-15
 
 ### New Features

pyproject.toml

@@ -143,7 +143,7 @@ namespaces = false
 version = {attr = 'sdv.__version__'}
 
 [tool.bumpversion]
-current_version = "1.27.1.dev0"
+current_version = "1.28.1.dev0"
 parse = '(?P<major>\d+)\.(?P<minor>\d+)\.(?P<patch>\d+)(\.(?P<release>[a-z]+)(?P<candidate>\d+))?'
 serialize = [
     '{major}.{minor}.{patch}.{release}{candidate}',

sdv/__init__.py

@@ -6,7 +6,7 @@
 
 __author__ = 'DataCebo, Inc.'
 __email__ = 'info@sdv.dev'
-__version__ = '1.27.1.dev0'
+__version__ = '1.28.1.dev0'
 
 
 import sys

sdv/_utils.py

@@ -273,7 +273,7 @@ def check_sdv_versions_and_warn(synthesizer):
     """
     current_community_version = getattr(version, 'community', None)
     current_enterprise_version = getattr(version, 'enterprise', None)
-    if synthesizer._fitted:
+    if getattr(synthesizer, '_fitted', False):
         fitted_community_version = getattr(synthesizer, '_fitted_sdv_version', None)
         fitted_enterprise_version = getattr(synthesizer, '_fitted_sdv_enterprise_version', None)
         community_mismatch = current_community_version != fitted_community_version

sdv/multi_table/_dayz_utils.py

@@ -1,8 +1,16 @@
 """Multi-Table DayZ parameter detection and creation."""
 
+import json
+
+import pandas as pd
+
+from sdv.cag._utils import _is_list_of_type
 from sdv.errors import SynthesizerInputError, SynthesizerProcessingError
-from sdv.multi_table._dayz_utils import create_parameters_multi_table
-from sdv.single_table.dayz import _validate_parameter_structure, _validate_tables_parameter
+from sdv.single_table.dayz import (
+    _validate_parameter_structure,
+    _validate_tables_parameter,
+    create_parameters,
+)
 
 REQUIRED_RELATIONSHIP_KEYS = [
     'parent_table_name',
@@ -18,6 +26,53 @@
 DEFAULT_NUM_ROWS = 1000
 
 
+def _detect_relationship_parameters(data, metadata):
+    """Detect all relationship-level for the DayZ parameters.
+
+    The relationship-level parameters are:
+    - The min and max cardinality
+
+    Args:
+        data (dict[str, pd.DataFrame]): The input data.
+        metadata (Metadata): The metadata object.
+
+    Returns:
+        dict: A list containing the detected parameters.
+    """
+    relationship_parameters = []
+    for relationship in metadata.relationships:
+        rel_tuple = (
+            relationship['parent_table_name'],
+            relationship['child_table_name'],
+            relationship['parent_primary_key'],
+            relationship['child_foreign_key'],
+        )
+        cardinality_table = pd.DataFrame(index=data[rel_tuple[0]][rel_tuple[2]].copy())
+        cardinality_table['cardinality'] = data[rel_tuple[1]][rel_tuple[3]].value_counts()
+        cardinality_table = cardinality_table.fillna(0)
+        relationship_parameters.append({
+            'parent_table_name': rel_tuple[0],
+            'child_table_name': rel_tuple[1],
+            'parent_primary_key': rel_tuple[2],
+            'child_foreign_key': rel_tuple[3],
+            'min_cardinality': cardinality_table['cardinality'].min(),
+            'max_cardinality': cardinality_table['cardinality'].max(),
+        })
+
+    return relationship_parameters
+
+
+def create_parameters_multi_table(data, metadata, output_filename):
+    """Create parameters for the DayZSynthesizer."""
+    parameters = create_parameters(data, metadata, None)
+    parameters['relationships'] = _detect_relationship_parameters(data, metadata)
+    if output_filename:
+        with open(output_filename, 'w') as f:
+            json.dump(parameters, f, indent=4)
+
+    return parameters
+
+
 def _validate_min_cardinality(relationship):
     min_cardinality = relationship['min_cardinality']
     if not isinstance(min_cardinality, int) or min_cardinality < 0:
@@ -48,8 +103,10 @@ def _validate_cardinality_bounds(relationship):
 
 
 def _validate_relationship_structure(dayz_parameters):
-    if not isinstance(dayz_parameters.get('relationships', []), list):
-        raise SynthesizerProcessingError("The 'relationships' parameter value must be a list.")
+    if not _is_list_of_type(dayz_parameters.get('relationships', []), dict):
+        raise SynthesizerProcessingError(
+            "The 'relationships' parameter value must be a list of dictionaries."
+        )
 
     for relationship in dayz_parameters.get('relationships', []):
         unknown_relationship_parameters = relationship.keys() - set(RELATIONSHIP_PARAMETER_KEYS)
@@ -160,18 +217,18 @@ def __init__(self, metadata, locales=['en_US']):
         )
 
     @classmethod
-    def create_parameters(cls, data, metadata, output_filename=None):
+    def create_parameters(cls, data, metadata, filepath=None):
         """Create parameters for the DayZSynthesizer.
 
         Args:
             data (dict[str, pd.DataFrame]): The input data.
             metadata (Metadata): The metadata object.
-            output_filename (str, optional): The output filename for the parameters.
+            filepath (str, optional): The output filename for the parameters.
 
         Returns:
             dict: The created parameters.
         """
-        return create_parameters_multi_table(data, metadata, output_filename)
+        return create_parameters_multi_table(data, metadata, filepath)
 
     @staticmethod
     def validate_parameters(metadata, parameters):

sdv/multi_table/dayz.py

@@ -11,9 +11,10 @@
 import tqdm
 from rdt.transformers import FloatFormatter
 
-from sdv._utils import MODELABLE_SDTYPES, _cast_to_iterable, _groupby_list
+from sdv._utils import MODELABLE_SDTYPES, _cast_to_iterable, _groupby_list, _is_datetime_type
 from sdv.cag import ProgrammableConstraint
 from sdv.cag._utils import _validate_constraints_single_table
+from sdv.constraints.utils import cast_to_datetime64
 from sdv.errors import SamplingError, SynthesizerInputError
 from sdv.metadata.errors import InvalidMetadataError
 from sdv.metadata.metadata import Metadata
@@ -37,6 +38,11 @@
 LOGGER = logging.getLogger(__name__)
 
 
+def _diff_and_bfill(series):
+    """Compute the diff of a pandas Series and backfill the first NaN."""
+    return series.diff().bfill()
+
+
 class PARSynthesizer(LossValuesMixin, MissingModuleMixin, BaseSynthesizer):
     """Synthesizer for sequential data.
 
@@ -310,20 +316,25 @@ def _transform_sequence_index(self, data):
         sequence_index_context = sequence_index_context.rename(
             columns={self._sequence_index: f'{self._sequence_index}.context'}
         )
+
+        if _is_datetime_type(sequence_index[self._sequence_index]):
+            sequence_index[self._sequence_index] = cast_to_datetime64(
+                sequence_index[self._sequence_index]
+            ).astype(np.int64)
+
         if all(sequence_index[self._sequence_key].nunique() == 1):
-            sequence_index_sequence = sequence_index[[self._sequence_index]].diff().bfill()
+            diff_series = sequence_index[self._sequence_index].diff().bfill()
         else:
-            sequence_index_sequence = (
-                sequence_index.groupby(self._sequence_key)
-                .apply(lambda x: x[self._sequence_index].diff().bfill())
-                .droplevel(1)
-                .reset_index()
-            )
+            diff_series = sequence_index.groupby(self._sequence_key, group_keys=False)[
+                self._sequence_index
+            ].transform(_diff_and_bfill)
 
+        sequence_index_sequence = diff_series.to_frame(name=self._sequence_index)
         if all(sequence_index_sequence[self._sequence_index].isna()):
             fill_value = 0
         else:
             fill_value = min(sequence_index_sequence[self._sequence_index].dropna())
+
         sequence_index_sequence = sequence_index_sequence.fillna(fill_value)
 
         data[self._sequence_index] = sequence_index_sequence[self._sequence_index].to_numpy()
@@ -579,7 +590,7 @@ def _sample_from_par(self, context, sequence_length=None):
                     pd.DataFrame({self._sequence_index: diffs})
                 )[self._sequence_index].to_numpy()
                 start_index = context_columns.index(f'{self._sequence_index}.context')
-                start = context_values[start_index]
+                start = context_values.iloc[start_index]
                 sequence[sequence_index_idx] = np.cumsum(diffs) - diffs[0] + start
 
             # Reformat as a DataFrame