fix: stop all-rows-fail on numeric column type mismatch (22018)

Root cause: Snowflake DESCRIBE TABLE returns FIXED(38,0) for INT columns
and REAL for FLOAT columns, but convert_value only checked for NUMBER/INT/FLOAT.
So string values (even valid ones like "0.035") passed through uncoerced to
Snowflake, which rejected them with 22018 on every row.

Three-part fix:
1. generator.py entity check: store float(entity_value) not the raw string,
and strip currency/unit chars before parsing. Add FIXED/REAL/DOUBLE/MONEY
to the is_numeric_col check.
2. legitdata_bridge.py convert_value: add FIXED/REAL/DOUBLE/MONEY to numeric
type check, and strip non-numeric chars from strings before float().
3. generic.py generate_key + generator.py _infer_strategy: add same types
so integer PKs and inferred strategies use numeric returns.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

legitdata_bridge.py

@@ -232,12 +232,18 @@ class KeyPairSnowflakeWriter:
                     if isinstance(value, str) and len(value) > max_len:
                         value = value[:max_len]
                 
-                # Clamp numbers to fit DECIMAL precision
-                elif 'NUMBER' in data_type or 'DECIMAL' in data_type or 'NUMERIC' in data_type or 'INT' in data_type or 'FLOAT' in data_type:
-                    # If a string landed in a numeric column (AI misclassification), coerce or null it
+                # Clamp numbers to fit DECIMAL precision.
+                # Includes Snowflake's internal DESCRIBE TABLE names: FIXED (integers), REAL (floats).
+                elif any(t in data_type for t in (
+                    'NUMBER', 'DECIMAL', 'NUMERIC', 'INT', 'FLOAT',
+                    'FIXED', 'REAL', 'DOUBLE', 'MONEY',
+                )):
+                    # If a string landed in a numeric column, strip formatting and coerce
                     if isinstance(value, str):
+                        import re as _re
+                        cleaned = _re.sub(r'[^\d.\-+eE]', '', value.replace(',', ''))
                         try:
-                            value = float(value)
+                            value = float(cleaned) if cleaned else None
                         except (ValueError, TypeError):
                             return None  # Can't coerce — use NULL rather than crash
                     precision = info.get('precision', 38)

legitdata_project/legitdata/generator.py

@@ -935,16 +935,21 @@ class LegitGenerator:
                         entity_value = entity[col_name]
                         # Check if column is numeric but AI gave us text
                         data_type_upper = (column.data_type or '').upper()
-                        is_numeric_col = any(t in data_type_upper for t in ('INT', 'NUMBER', 'NUMERIC', 'DECIMAL', 'BIGINT', 'SMALLINT', 'FLOAT'))
-                        
+                        is_numeric_col = any(t in data_type_upper for t in (
+                            'INT', 'NUMBER', 'NUMERIC', 'DECIMAL', 'BIGINT', 'SMALLINT',
+                            'FLOAT', 'REAL', 'DOUBLE', 'FIXED', 'MONEY',
+                        ))
+
                         if is_numeric_col:
-                            # Validate that the value is actually numeric
+                            # Coerce to float and store as Python numeric, not string.
+                            # Strip currency symbols / units before trying.
+                            import re as _re
+                            cleaned = _re.sub(r'[^\d.\-+eE]', '', str(entity_value).replace(',', ''))
                             try:
-                                float(entity_value)
-                                row[col_name] = entity_value
+                                row[col_name] = float(cleaned)
                                 continue
                             except (ValueError, TypeError):
-                                # AI gave us text for a numeric column - fall through to generic generation
+                                # AI gave us unparseable text — fall through to generic
                                 if i == 0:
                                     print(f"    [WARN] {col_name}: AI gave '{entity_value}' for numeric column, using inferred instead")
                                 pass  # Fall through to generic generation below
@@ -1649,8 +1654,8 @@ class LegitGenerator:
         if semantic_strategy:
             return semantic_strategy
         
-        # Check if data type is numeric
-        is_numeric = any(t in data_type for t in ('INT', 'NUMBER', 'NUMERIC', 'DECIMAL', 'BIGINT', 'SMALLINT'))
+        # Check if data type is numeric (includes Snowflake internal names FIXED/REAL)
+        is_numeric = any(t in data_type for t in ('INT', 'NUMBER', 'NUMERIC', 'DECIMAL', 'BIGINT', 'SMALLINT', 'FLOAT', 'REAL', 'DOUBLE', 'FIXED', 'MONEY'))
         
         # Key columns (CustomerKey, ProductKey, etc.)
         if 'key' in col_lower and 'keyboard' not in col_lower:

legitdata_project/legitdata/sourcer/generic.py

@@ -189,7 +189,7 @@ class GenericSourcer:
         # Check if column expects a numeric type
         if data_type:
             data_type_upper = data_type.upper()
-            if any(t in data_type_upper for t in ('INT', 'NUMBER', 'NUMERIC', 'DECIMAL', 'BIGINT', 'SMALLINT')):
+            if any(t in data_type_upper for t in ('INT', 'NUMBER', 'NUMERIC', 'DECIMAL', 'BIGINT', 'SMALLINT', 'FIXED', 'REAL', 'DOUBLE')):
                 return self._key_counters[prefix]
         
         return f"{prefix}_{self._key_counters[prefix]:05d}"