Feat(executor): add support for TRIM, fix TRIM CSV-style parsing order (

#2342)

* Feat(executor): add support for TRIM, fix TRIM CSV-style parsing order

* Define trim inline using a lambda

* Comment cleanup

sqlglot/dialects/spark2.py

@@ -101,6 +101,8 @@ def _unqualify_pivot_columns(expression: exp.Expression) -> exp.Expression:
 
 class Spark2(Hive):
     class Parser(Hive.Parser):
+        TRIM_PATTERN_FIRST = True
+
         FUNCTIONS = {
             **Hive.Parser.FUNCTIONS,
             "AGGREGATE": exp.Reduce.from_arg_list,

sqlglot/executor/env.py

@@ -202,4 +202,5 @@ def interval(this, unit):
     "CURRENTTIME": datetime.datetime.now,
     "CURRENTDATE": datetime.date.today,
     "STRFTIME": null_if_any(lambda fmt, arg: datetime.datetime.fromisoformat(arg).strftime(fmt)),
+    "TRIM": null_if_any(lambda this, e=None: this.strip(e)),
 }

sqlglot/parser.py

@@ -889,9 +889,12 @@ class Parser(metaclass=_Parser):
     # Whether or not the table sample clause expects CSV syntax
     TABLESAMPLE_CSV = False
 
-    # Whether or not the SET command needs a delimiter (e.g. "=") for assignments.
+    # Whether or not the SET command needs a delimiter (e.g. "=") for assignments
     SET_REQUIRES_ASSIGNMENT_DELIMITER = True
 
+    # Whether the TRIM function expects the characters to trim as its first argument
+    TRIM_PATTERN_FIRST = False
+
     __slots__ = (
         "error_level",
         "error_message_context",
@@ -4411,16 +4414,18 @@ def _parse_trim(self) -> exp.Trim:
 
         position = None
         collation = None
+        expression = None
 
         if self._match_texts(self.TRIM_TYPES):
             position = self._prev.text.upper()
 
-        expression = self._parse_bitwise()
+        this = self._parse_bitwise()
         if self._match_set((TokenType.FROM, TokenType.COMMA)):
-            this = self._parse_bitwise()
-        else:
-            this = expression
-            expression = None
+            invert_order = self._prev.token_type == TokenType.FROM or self.TRIM_PATTERN_FIRST
+            expression = self._parse_bitwise()
+
+            if invert_order:
+                this, expression = expression, this
 
         if self._match(TokenType.COLLATE):
             collation = self._parse_bitwise()

tests/dialects/test_bigquery.py

@@ -134,6 +134,18 @@ def test_bigquery(self):
         self.validate_all('x <> """"""', write={"bigquery": "x <> ''"})
         self.validate_all("x <> ''''''", write={"bigquery": "x <> ''"})
         self.validate_all("CAST(x AS DATETIME)", read={"": "x::timestamp"})
+        self.validate_all(
+            "TRIM(item, '*')",
+            read={
+                "snowflake": "TRIM(item, '*')",
+                "spark": "TRIM('*', item)",
+            },
+            write={
+                "bigquery": "TRIM(item, '*')",
+                "snowflake": "TRIM(item, '*')",
+                "spark": "TRIM('*' FROM item)",
+            },
+        )
         self.validate_all(
             "CREATE OR REPLACE TABLE `a.b.c` COPY `a.b.d`",
             write={

tests/test_executor.py

@@ -624,6 +624,8 @@ def test_scalar_functions(self):
             ("LEFT('12345', 3)", "123"),
             ("RIGHT('12345', 3)", "345"),
             ("DATEDIFF('2022-01-03'::date, '2022-01-01'::TIMESTAMP::DATE)", 2),
+            ("TRIM(' foo ')", "foo"),
+            ("TRIM('afoob', 'ab')", "foo"),
         ]:
             with self.subTest(sql):
                 result = execute(f"SELECT {sql}")