feat: cast trim input to varchar implicitly

fakesnow/fakes.py

@@ -174,6 +174,8 @@ def _execute(
             .transform(transforms.tag)
             .transform(transforms.semi_structured_types)
             .transform(transforms.try_parse_json)
+            # NOTE: trim_cast_varchar must be before json_extract_cast_as_varchar
+            .transform(transforms.trim_cast_varchar)
             # indices_to_json_extract must be before regex_substr
             .transform(transforms.indices_to_json_extract)
             .transform(transforms.json_extract_cast_as_varchar)

fakesnow/transforms.py

@@ -1095,6 +1095,21 @@ def timestamp_ntz_ns(expression: exp.Expression) -> exp.Expression:
     return expression
 
 
+def trim_cast_varchar(expression: exp.Expression) -> exp.Expression:
+    """Snowflake's TRIM casts input to VARCHAR implicitly."""
+
+    if not (isinstance(expression, exp.Trim)):
+        return expression
+
+    operand = expression.this
+    if isinstance(operand, exp.Cast) and operand.to.this in [exp.DataType.Type.VARCHAR, exp.DataType.Type.TEXT]:
+        return expression
+
+    return exp.Trim(
+        this=exp.Cast(this=operand, to=exp.DataType(this=exp.DataType.Type.VARCHAR, nested=False, prefix=False))
+    )
+
+
 def try_parse_json(expression: exp.Expression) -> exp.Expression:
     """Convert TRY_PARSE_JSON() to TRY_CAST(... as JSON).
 

tests/test_fakes.py

@@ -1393,6 +1393,24 @@ def test_transactions(conn: snowflake.connector.SnowflakeConnection):
         assert cur.fetchall() == [("Statement executed successfully.",)]
 
 
+def test_trim_cast_varchar(conn: snowflake.connector.SnowflakeConnection):
+    with conn.cursor() as cur:
+        cur.execute("create or replace table trim_cast_varchar(id number, name varchar);")
+        cur.execute("insert into trim_cast_varchar(id, name) values (1, '  name 1  '), (2, 'name2   ');")
+        cur.execute("select trim(id), trim(name) from trim_cast_varchar;")
+
+        assert cur.fetchall() == [("1", "name 1"), ("2", "name2")]
+
+    with conn.cursor() as cur:
+        cur.execute("create or replace table trim_cast_varchar_variant_field(data variant);")
+        cur.execute(
+            """insert into trim_cast_varchar_variant_field(data) values ('{"k1": "   v11  "}'),('{"k1": 21}');"""
+        )
+        cur.execute("select trim(data:k1) from trim_cast_varchar_variant_field;")
+
+        assert cur.fetchall() == [("v11",), ("21",)]
+
+
 def test_unquoted_identifiers_are_upper_cased(dcur: snowflake.connector.cursor.SnowflakeCursor):
     dcur.execute("create table customers (id int, first_name varchar, last_name varchar)")
     dcur.execute("insert into customers values (1, 'Jenny', 'P')")

tests/test_transforms.py

@@ -44,6 +44,7 @@
     to_decimal,
     to_timestamp,
     to_timestamp_ntz,
+    trim_cast_varchar,
     try_parse_json,
     try_to_decimal,
     upper_case_unquoted_identifiers,
@@ -770,6 +771,23 @@ def test_try_parse_json() -> None:
     )
 
 
+def test_trim_cast_varchar() -> None:
+    assert (
+        sqlglot.parse_one("SELECT TRIM(col) FROM table1").transform(trim_cast_varchar).sql(dialect="duckdb")
+        == "SELECT TRIM(CAST(col AS TEXT)) FROM table1"
+    )
+
+    assert (
+        sqlglot.parse_one("SELECT TRIM(col::varchar) FROM table1").transform(trim_cast_varchar).sql(dialect="duckdb")
+        == "SELECT TRIM(CAST(col AS TEXT)) FROM table1"
+    )
+
+    assert (
+        sqlglot.parse_one("SELECT TRIM(col::TEXT) FROM table1").transform(trim_cast_varchar).sql(dialect="duckdb")
+        == "SELECT TRIM(CAST(col AS TEXT)) FROM table1"
+    )
+
+
 def test_upper_case_unquoted_identifiers() -> None:
     assert (
         sqlglot.parse_one("select name, name as fname from table1").transform(upper_case_unquoted_identifiers).sql()