feat(duckdb): transpile snowflake's JAROWINKLER_SIMILARITY properly (#7501)

* fix(snowflake): transpile JAROWINKLER_SIMILARITY to DuckDB with correct scale

Snowflake's JAROWINKLER_SIMILARITY returns values on a 0-100 integer scale,
while DuckDB's JARO_WINKLER_SIMILARITY returns values on a 0.0-1.0 float scale.

Changes:
- Added integer_scale flag to JarowinklerSimilarity expression
- Snowflake parser sets integer_scale=True when parsing JAROWINKLER_SIMILARITY
- DuckDB generator multiplies by 100 and casts to INTEGER when flag is set
- Preserves DuckDB roundtrip (no scale conversion when flag is not set)

This ensures semantic equivalence when transpiling from Snowflake to DuckDB.

* chore: update integration tests submodule

Author: fivetran-ashashankar

sqlglot-integration-tests

@@ -108,7 +108,12 @@ class EuclideanDistance(Expression, Func):
 
 
 class JarowinklerSimilarity(Expression, Func):
-    arg_types = {"this": True, "expression": True, "case_insensitive": False}
+    arg_types = {
+        "this": True,
+        "expression": True,
+        "case_insensitive": False,
+        "integer_scale": False,
+    }
 
 
 class ManhattanDistance(Expression, Func):

sqlglot/expressions/math.py

@@ -24,7 +24,6 @@
     getbit_sql,
     groupconcat_sql,
     inline_array_unless_query,
-    jarowinkler_similarity,
     months_between_sql,
     no_datetime_sql,
     no_comment_column_constraint_sql,
@@ -1609,7 +1608,6 @@ class DuckDBGenerator(generator.Generator):
         ),
         exp.Ceil: _ceil_floor,
         exp.Floor: _ceil_floor,
-        exp.JarowinklerSimilarity: jarowinkler_similarity("JARO_WINKLER_SIMILARITY"),
         exp.JSONBExists: rename_func("JSON_EXISTS"),
         exp.JSONExtract: _arrow_json_extract_sql,
         exp.JSONExtractArray: _json_extract_value_array_sql,
@@ -2285,6 +2283,21 @@ def parseip_sql(self, expression: exp.ParseIp) -> str:
         self.unsupported("PARSE_IP is not supported in DuckDB")
         return self.function_fallback_sql(expression)
 
+    def jarowinklersimilarity_sql(self, expression: exp.JarowinklerSimilarity) -> str:
+        this = expression.this
+        expr = expression.expression
+
+        if expression.args.get("case_insensitive"):
+            this = exp.Upper(this=this)
+            expr = exp.Upper(this=expr)
+
+        result = exp.func("JARO_WINKLER_SIMILARITY", this, expr)
+
+        if expression.args.get("integer_scale"):
+            result = exp.cast(result * 100, "INTEGER")
+
+        return self.sql(result)
+
     def nthvalue_sql(self, expression: exp.NthValue) -> str:
         from_first = expression.args.get("from_first", True)
         if not from_first:

sqlglot/generators/duckdb.py

@@ -539,6 +539,7 @@ class SnowflakeParser(parser.Parser):
             this=seq_get(args, 0),
             expression=seq_get(args, 1),
             case_insensitive=True,
+            integer_scale=True,
         ),
         "MD5_HEX": exp.MD5.from_arg_list,
         "MD5_BINARY": exp.MD5Digest.from_arg_list,

sqlglot/parsers/snowflake.py

@@ -221,7 +221,7 @@ def test_snowflake(self):
             "JAROWINKLER_SIMILARITY('hello', 'world')",
             write={
                 "snowflake": "JAROWINKLER_SIMILARITY('hello', 'world')",
-                "duckdb": "JARO_WINKLER_SIMILARITY(UPPER('hello'), UPPER('world'))",
+                "duckdb": "CAST(JARO_WINKLER_SIMILARITY(UPPER('hello'), UPPER('world')) * 100 AS INT)",
                 "clickhouse": "jaroWinklerSimilarity(UPPER('hello'), UPPER('world'))",
             },
         )