[ZeroOps bug bash] Use fixed-point decimals in the schema-drift example

knowledge_base/failing_job_examples/src/schema_drift.py

@@ -1,18 +1,25 @@
 # Databricks notebook source
-from pyspark.sql.types import NumericType, StringType, StructField, StructType
+from pyspark.sql.types import DecimalType, NumericType, StringType, StructField, StructType
 
 
+# Simulated monetary amounts use a fixed-point decimal (not float) to avoid binary floating-point rounding drift.
 observed_source_schema = StructType(
     [
         StructField("transaction_id", StringType(), False),
-        StructField("amount", StringType(), False),
+        StructField("amount", DecimalType(12, 2), False),
     ]
 )
-amount_type = observed_source_schema["amount"].dataType
 
-if not isinstance(amount_type, NumericType):
-    raise TypeError(
-        "Schema contract violation in simulated source schema (no source table): "
-        f"expected amount to be numeric, found {amount_type.simpleString()}; "
-        f"observed schema {observed_source_schema.simpleString()}"
-    )
+
+def validate_amount_schema(schema):
+    """Raise TypeError if the schema's amount field is not numeric."""
+    amount_type = schema["amount"].dataType
+    if not isinstance(amount_type, NumericType):
+        raise TypeError(
+            "Schema contract violation in simulated source schema (no source table): "
+            f"expected amount to be numeric, found {amount_type.simpleString()}; "
+            f"observed schema {schema.simpleString()}"
+        )
+
+
+validate_amount_schema(observed_source_schema)