add tests for checking precision

Author: vvkh

data_diff/databases/base.py

@@ -201,6 +201,7 @@ class BaseDialect(abc.ABC):
     SUPPORTS_INDEXES: ClassVar[bool] = False
     PREVENT_OVERFLOW_WHEN_CONCAT: ClassVar[bool] = False
     TYPE_CLASSES: ClassVar[Dict[str, Type[ColType]]] = {}
+    DEFAULT_NUMERIC_PRECISION: ClassVar[int] = 0  # effective precision when type is just "NUMERIC"
 
     PLACEHOLDER_TABLE = None  # Used for Oracle
 

data_diff/databases/bigquery.py

@@ -75,6 +75,9 @@ class Dialect(BaseDialect):
     }
     TYPE_ARRAY_RE = re.compile(r"ARRAY<(.+)>")
     TYPE_STRUCT_RE = re.compile(r"STRUCT<(.+)>")
+    # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#parameterized_decimal_type
+    # The default scale is 9, which means a number can have up to 9 digits after the decimal point.
+    DEFAULT_NUMERIC_PRECISION = 9
 
     def random(self) -> str:
         return "RAND()"

data_diff/databases/duckdb.py

@@ -45,6 +45,10 @@ class Dialect(BaseDialect):
     SUPPORTS_PRIMARY_KEY = True
     SUPPORTS_INDEXES = True
 
+    # https://duckdb.org/docs/sql/data_types/numeric#fixed-point-decimals
+    # The default WIDTH and SCALE is DECIMAL(18, 3), if none are specified.
+    DEFAULT_NUMERIC_PRECISION = 3
+
     TYPE_CLASSES = {
         # Timestamps
         "TIMESTAMP WITH TIME ZONE": TimestampTZ,

data_diff/databases/postgresql.py

@@ -45,6 +45,12 @@ class PostgresqlDialect(BaseDialect):
     SUPPORTS_PRIMARY_KEY: ClassVar[bool] = True
     SUPPORTS_INDEXES = True
 
+    # https://www.postgresql.org/docs/current/datatype-numeric.html#DATATYPE-NUMERIC-DECIMAL
+    # without any precision or scale creates an “unconstrained numeric” column
+    # in which numeric values of any length can be stored, up to the implementation limits.
+    # https://www.postgresql.org/docs/current/datatype-numeric.html#DATATYPE-NUMERIC-TABLE
+    DEFAULT_NUMERIC_PRECISION = 16383
+
     TYPE_CLASSES: ClassVar[Dict[str, Type[ColType]]] = {
         # Timestamps
         "timestamp with time zone": TimestampTZ,

tests/test_database.py

@@ -134,3 +134,36 @@ def test_three_part_support(self):
             d = db.query_table_schema(part.path)
             assert len(d) == 1
             db.query(part.drop())
+
+
+@test_each_database
+class TestNumericPrecisionParsing(unittest.TestCase):
+    def test_specified_precision(self):
+        name = "tbl_" + random_table_suffix()
+        db = get_conn(self.db_cls)
+        tbl = table(name, schema={"value": "NUMERIC(10, 2)"})
+        db.query(tbl.create())
+        t = table(name)
+        raw_schema = db.query_table_schema(t.path)
+        schema = db._process_table_schema(t.path, raw_schema)
+        self.assertEqual(schema["value"].precision,  2)
+
+    def test_specified_zero_precision(self):
+        name = "tbl_" + random_table_suffix()
+        db = get_conn(self.db_cls)
+        tbl = table(name, schema={"value": "NUMERIC(10)"})
+        db.query(tbl.create())
+        t = table(name)
+        raw_schema = db.query_table_schema(t.path)
+        schema = db._process_table_schema(t.path, raw_schema)
+        self.assertEqual(schema["value"].precision, 0)
+
+    def test_default_precision(self):
+        name = "tbl_" + random_table_suffix()
+        db = get_conn(self.db_cls)
+        tbl = table(name, schema={"value": "NUMERIC"})
+        db.query(tbl.create())
+        t = table(name)
+        raw_schema = db.query_table_schema(t.path)
+        schema = db._process_table_schema(t.path, raw_schema)
+        self.assertEqual(schema["value"].precision, db.dialect.DEFAULT_NUMERIC_PRECISION)