wrap bq -> duckdb REGEXP_EXTRACT SUBSTRING() call in NULLIF

Author: treysp

sqlglot/dialects/duckdb.py

@@ -1449,7 +1449,10 @@ def regexpextract_sql(self, expression: exp.RegexpExtract) -> str:
             position = expression.args.get("position")
             occurrence = expression.args.get("occurrence")
             if position and (not position.is_int or position.to_py() > 1):
-                this = exp.Substring(this=this, start=position)
+                # substring returns '' if position > len(string), but the '' shouldn't carry through to REGEXP_EXTRACT
+                this = exp.Nullif(
+                    this=exp.Substring(this=this, start=position), expression=exp.Literal.string("")
+                )
 
             # Do not render group if there is no following argument,
             # and it's the default value for this dialect

tests/dialects/test_bigquery.py

@@ -2820,7 +2820,7 @@ def test_regexp_extract(self):
             "SELECT REGEXP_EXTRACT(abc, 'pattern(group)', 2) FROM table",
             write={
                 "bigquery": "SELECT REGEXP_EXTRACT(abc, 'pattern(group)', 2) FROM table",
-                "duckdb": '''SELECT REGEXP_EXTRACT(SUBSTRING(abc, 2), 'pattern(group)', 1) FROM "table"''',
+                "duckdb": '''SELECT REGEXP_EXTRACT(NULLIF(SUBSTRING(abc, 2), ''), 'pattern(group)', 1) FROM "table"''',
             },
         )
 
@@ -2838,7 +2838,7 @@ def test_regexp_extract(self):
             "SELECT REGEXP_EXTRACT(abc, 'pattern(group)', 2, 3) FROM table",
             write={
                 "bigquery": "SELECT REGEXP_EXTRACT(abc, 'pattern(group)', 2, 3) FROM table",
-                "duckdb": '''SELECT ARRAY_EXTRACT(REGEXP_EXTRACT_ALL(SUBSTRING(abc, 2), 'pattern(group)', 1), 3) FROM "table"''',
+                "duckdb": '''SELECT ARRAY_EXTRACT(REGEXP_EXTRACT_ALL(NULLIF(SUBSTRING(abc, 2), ''), 'pattern(group)', 1), 3) FROM "table"''',
             },
         )