table-validator 0.1.1__tar.gz → 0.1.2__tar.gz

{table_validator-0.1.1/table_validator.egg-info → table_validator-0.1.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: table-validator
-Version: 0.1.1
+Version: 0.1.2
 Summary: CLI tool for validating data migrations between Azure (Blob Storage / SQL Database) and Databricks Delta Lake catalogs
 License: MIT
 Requires-Python: >=3.9

{table_validator-0.1.1 → table_validator-0.1.2}/table_validator/validators/catalog_validator.py

@@ -88,6 +88,30 @@ class CatalogValidator:
         self.partition_prompt = partition_prompt
         logger.debug("CatalogValidator initialised")
 
+    @staticmethod
+    def _lookup_primary_key(
+        request: "CatalogValidationRequest",
+        schema_name: str,
+        table_name: str,
+    ) -> Optional[List[str]]:
+        """
+        Look up request.primary_keys (keyed by "schema.table" or bare
+        table name, as typed by the user into config.yaml/the wizard) for
+        the given schema_name/table_name.
+
+        schema_name/table_name here come from whatever casing Databricks'
+        information_schema actually returns, matched case-insensitively
+        against the catalog in compare_schemas/compare_tables - which
+        does not necessarily match the exact casing the user configured.
+        An exact-match dict lookup would silently miss a configured key
+        and fall through to the much more expensive ROW_NUMBER() fallback
+        with no indication why, so this normalizes both sides to
+        lowercase before comparing.
+        """
+        lowered = {k.lower(): v for k, v in request.primary_keys.items()}
+        key_lookup = f"{schema_name}.{table_name}".lower()
+        return lowered.get(key_lookup) or lowered.get(table_name.lower())
+
     # ------------------------------------------------------------------
     # Stage 1 + top-level orchestration
     # ------------------------------------------------------------------
@@ -697,10 +721,7 @@ class CatalogValidator:
 
         # Configured PK column missing from either side is also BLOCKING -
         # every later tier depends on being able to resolve a usable key.
-        key_lookup = f"{schema_name}.{table_name}"
-        configured_key = request.primary_keys.get(key_lookup) or request.primary_keys.get(
-            table_name
-        )
+        configured_key = self._lookup_primary_key(request, schema_name, table_name)
         if configured_key:
             common_lower = {c.lower() for c in common_cols}
             if any(k.lower() not in common_lower for k in configured_key):
@@ -1040,8 +1061,7 @@ class CatalogValidator:
             )
             return
 
-        key_lookup = f"{schema_name}.{table_name}"
-        key_columns = request.primary_keys.get(key_lookup) or request.primary_keys.get(table_name)
+        key_columns = self._lookup_primary_key(request, schema_name, table_name)
         candidates = self._partition_candidates(common_cols, key_columns)
 
         context = PartitionPromptContext(
@@ -1184,10 +1204,7 @@ class CatalogValidator:
             result.tier_reached = ValidationTier.ROW_HASH
 
             if not using_row_number_fallback:
-                key_lookup = f"{schema_name}.{table_name}"
-                key_columns = request.primary_keys.get(key_lookup) or request.primary_keys.get(
-                    table_name
-                )
+                key_columns = self._lookup_primary_key(request, schema_name, table_name)
                 mismatched_keys = [m.primary_key for m in mismatches if m.status == "MISMATCH"]
                 if key_columns and mismatched_keys:
                     self._tier5_column_diff(
@@ -1238,10 +1255,7 @@ class CatalogValidator:
         written straight to `result.data`, so multiple bucket calls
         aggregate instead of each overwriting the last.
         """
-        key_lookup = f"{schema_name}.{table_name}"
-        row_hash_key_columns = request.primary_keys.get(key_lookup) or request.primary_keys.get(
-            table_name
-        )
+        row_hash_key_columns = self._lookup_primary_key(request, schema_name, table_name)
 
         using_row_number_fallback = not row_hash_key_columns
         if using_row_number_fallback:
@@ -1470,10 +1484,7 @@ class CatalogValidator:
     ) -> DataValidationResult:
 
         mode = request.data_compare_mode
-        key = f"{schema_name}.{table_name}"
-        key_columns = request.primary_keys.get(key) or request.primary_keys.get(
-            table_name
-        )
+        key_columns = self._lookup_primary_key(request, schema_name, table_name)
 
         logger.info(
             "[compare_data] table=%s.%s | mode=%s | resolved_key_columns=%s",

{table_validator-0.1.1 → table_validator-0.1.2/table_validator.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: table-validator
-Version: 0.1.1
+Version: 0.1.2
 Summary: CLI tool for validating data migrations between Azure (Blob Storage / SQL Database) and Databricks Delta Lake catalogs
 License: MIT
 Requires-Python: >=3.9

{table_validator-0.1.1 → table_validator-0.1.2}/table_validator.egg-info/scm_version.json

@@ -1,7 +1,7 @@
 {
-  "tag": "0.1.1",
+  "tag": "0.1.2",
   "distance": 0,
-  "node": "gd0c901c9dfc750527e33620c5dea6b15e43e4443",
+  "node": "g43988f9f51a097e239c49384ab59f00fec5851ab",
   "dirty": false,
   "branch": "HEAD",
   "node_date": "2026-08-27"

{table_validator-0.1.1 → table_validator-0.1.2}/tests/test_catalog_validator.py

@@ -991,6 +991,46 @@ def test_tier5_column_diff_runs_for_mismatched_key_and_names_exact_column():
     assert detail.verified is True
 
 
+def test_primary_key_lookup_is_case_insensitive_against_catalog_metadata():
+    """request.primary_keys is keyed by whatever casing the user typed
+    into config.yaml/the wizard, but schema_name/table_name at lookup
+    time come from Databricks' information_schema (matched
+    case-insensitively against the catalog elsewhere in this file). A
+    primary key configured as "Bronze.Customers" must still be found for
+    the real "bronze.customers" table - an exact-match lookup would
+    silently miss it and fall through to the much slower ROW_NUMBER()
+    fallback instead of using the real configured key."""
+    connector = _mismatched_fingerprint_connector()
+    connector.get_row_hashes.side_effect = lambda catalog, schema, table, cols, pk, bucket_predicate=None: (
+        _hash_df([(1, "aaa"), (2, "bbb")])
+        if catalog == "cat_source"
+        else _hash_df([(1, "aaa"), (2, "zzz")])
+    )
+    connector.get_row_detail_for_keys.return_value = [
+        {
+            "key": {"id": 2},
+            "mismatched_columns": ["name"],
+            "source_values": {"name": "old-value"},
+            "target_values": {"name": "new-value"},
+            "source_row_hash": "bbb",
+            "target_row_hash": "zzz",
+        }
+    ]
+    validator = CatalogValidator(connector)
+
+    result = validator.compare_catalogs(
+        _request(primary_keys={"Bronze.Customers": ["id"]})
+    )
+    table = result.schemas[0].tables[0]
+
+    connector.get_row_hashes_by_row_number.assert_not_called()
+    assert connector.get_row_hashes.call_count == 2  # once per side (source, target)
+    connector.get_row_detail_for_keys.assert_called_once()
+    assert table.data.key_columns == ["id"]
+    assert table.tier_reached == ValidationTier.COLUMN_DIFF
+    assert table.data.sample_changed_detail[0].verified is True
+
+
 def test_tier5_column_diff_runs_for_row_number_fallback_when_mismatched():
     """No primary key configured -> row-number fallback is used for Tier
     4, but a real mismatch must still let Tier 5 attempt best-effort