code optimization

Author: sycai

bigframes/bigquery/_operations/ai.py

@@ -113,9 +113,8 @@ def generate(
     if output_schema is None:
         output_schema_str = None
     else:
-        sorted_fields = sorted(tuple(output_schema.items()), key=lambda t: t[0])
         output_schema_str = ", ".join(
-            [f"{name} {sql_type}" for name, sql_type in sorted_fields]
+            [f"{name} {sql_type}" for name, sql_type in output_schema.items()]
         )
         # Validate user input
         output_schemas.parse_sql_fields(output_schema_str)

bigframes/operations/output_schemas.py

@@ -49,7 +49,7 @@ def parse_sql_type(sql: str) -> pa.DataType:
 
     if sql.startswith("STRUCT<") and sql.endswith(">"):
         inner_fields = parse_sql_fields(sql[len("STRUCT<") : -1])
-        return pa.struct(sorted(inner_fields, key=lambda f: f.name))
+        return pa.struct(inner_fields)
 
     raise ValueError(f"Unsupported SQL type: {sql}")
 
@@ -76,7 +76,7 @@ def parse_sql_fields(sql: str) -> tuple[pa.Field]:
     # Append the last field
     fields.append(parse_sql_field(sql[start_idx:]))
 
-    return tuple(fields)
+    return tuple(sorted(fields, key=lambda f: f.name))
 
 
 def parse_sql_field(sql: str) -> pa.Field:

tests/unit/operations/test_output_schemas.py

@@ -36,7 +36,11 @@
             pa.struct((pa.field("x", pa.int64()), pa.field("y", pa.float64()))),
         ),
         (
-            "ARRAY<STRUCT<x INT64, y INT64>>",
+            "STRUCT<y INT64,  x  FLOAT64>",
+            pa.struct((pa.field("x", pa.float64()), pa.field("y", pa.int64()))),
+        ),
+        (
+            "ARRAY<STRUCT<y INT64, x INT64>>",
             pa.list_(pa.struct((pa.field("x", pa.int64()), pa.field("y", pa.int64())))),
         ),
         (