prophecy-libs 2.1.14.dev101__tar.gz → 2.1.14.dev102__tar.gz

{prophecy_libs-2.1.14.dev101/prophecy_libs.egg-info → prophecy_libs-2.1.14.dev102}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: prophecy_libs
-Version: 2.1.14.dev101
+Version: 2.1.14.dev102
 Summary: Helper library for prophecy generated code
 Home-page: https://github.com/SimpleDataLabsInc/prophecy-python-libs
 License: GPL-3.0

{prophecy_libs-2.1.14.dev101 → prophecy_libs-2.1.14.dev102}/prophecy/utils/diff.py

@@ -97,6 +97,12 @@ def needs_complex_comparison(data_type):
     return False
 
 
+# When every column is a join key, value structs would otherwise be empty. Spark 3.5+
+# fails in ExpressionEncoder with: "each serializer expression should contain at least
+# one BoundReference". A single bound column avoids an empty `struct()`.
+_EMPTY_VALUE_STRUCT_FIELD = "__prophecy_diff_empty_values_anchor__"
+
+
 class DataFrameDiff:
     COMPUTED_DIFFS = {}
 
@@ -370,6 +376,20 @@ class DataFrameDiff:
             for col in value_columns
         ]
 
+        if not left_value_exprs:
+            # No value columns (all columns are keys): empty struct() breaks Spark's encoder.
+            anchor_col = key_columns[0]
+            left_value_exprs = [
+                F.col(DataFrameDiff.prophecy_sc("left", anchor_col)).alias(
+                    _EMPTY_VALUE_STRUCT_FIELD
+                )
+            ]
+            right_value_exprs = [
+                F.col(DataFrameDiff.prophecy_sc("right", anchor_col)).alias(
+                    _EMPTY_VALUE_STRUCT_FIELD
+                )
+            ]
+
         left_struct = F.struct(*left_value_exprs).alias("left_values")
         right_struct = F.struct(*right_value_exprs).alias("right_values")
 
@@ -489,7 +509,11 @@ class DataFrameDiff:
 
         comparison_struct = "compared_values"
 
-        fields = joined_df.select(f"{comparison_struct}.*").columns
+        fields = [
+            c
+            for c in joined_df.select(f"{comparison_struct}.*").columns
+            if c != _EMPTY_VALUE_STRUCT_FIELD
+        ]
 
         # Build aggregators for match and mismatch counts for each specified column
         per_column_aggregators = []

{prophecy_libs-2.1.14.dev101 → prophecy_libs-2.1.14.dev102/prophecy_libs.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: prophecy_libs
-Version: 2.1.14.dev101
+Version: 2.1.14.dev102
 Summary: Helper library for prophecy generated code
 Home-page: https://github.com/SimpleDataLabsInc/prophecy-python-libs
 License: GPL-3.0

{prophecy_libs-2.1.14.dev101 → prophecy_libs-2.1.14.dev102}/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup, find_packages
 
 setup(
     name="prophecy_libs",
-    version="2.1.14.dev101",
+    version="2.1.14.dev102",
     url="https://github.com/SimpleDataLabsInc/prophecy-python-libs",
     packages=find_packages(exclude=["test.*", "test"]),
     package_data={