Fix UDF publisher script (#330)

script/publish_persistent_udfs

@@ -24,7 +24,7 @@ from bigquery_etl.parse_udf import (
 )
 
 
-UDF_RE = re.compile(r"udf_(?:js_)?([a-zA-z0-9_]+)")
+UDF_RE = re.compile(r"udf_(?:js_|legacy_)?([a-zA-z0-9_]+)")
 
 OPTIONS_LIB_RE = re.compile(r'library = "gs://[^"]+/([^"]+)"')
 
@@ -75,11 +75,12 @@ def main():
         udfs_to_publish.append(raw_udf)
         for dep in udfs_to_publish:
             if dep not in published_udfs:
-                publish_persistent_udf(raw_udfs[dep], client, args.dataset, args.project_id)
+                publish_persistent_udf(raw_udfs[dep], client, args.dataset, args.project_id,
+                                       args.gcs_bucket, args.gcs_path)
                 published_udfs.append(dep)
 
 
-def publish_persistent_udf(raw_udf, client, dataset, project_id):
+def publish_persistent_udf(raw_udf, client, dataset, project_id, gcs_bucket, gcs_path):
     # transforms temporary UDF to persistent UDFs and publishes them
     for definition in raw_udf.definitions:
         # Within a standard SQL function, references to other entities require explicit project IDs
@@ -87,6 +88,10 @@ def publish_persistent_udf(raw_udf, client, dataset, project_id):
             "`" + project_id + "`." + dataset + "." + r"\1", definition
         )
 
+        query_with_renamed_udfs = query_with_renamed_udfs.replace(
+            "CREATE TEMP FUNCTION", "CREATE OR REPLACE FUNCTION"
+        )
+
         # adjust paths for dependencies stored in GCS
         query = OPTIONS_LIB_RE.sub(
             fr'library = "gs://{gcs_bucket}/{gcs_path}\1"',
@@ -97,7 +102,7 @@ def publish_persistent_udf(raw_udf, client, dataset, project_id):
 
 def push_dependencies_to_gcs(bucket, path, dep_dir):
     client = storage.Client()
-    bucket = storage_client.get_bucket(bucket)
+    bucket = client.get_bucket(bucket)
 
     for root, dirs, files in os.walk(dep_dir):
         for filename in files: