Fix(tokenizer): handle empty hex strings (#5763)

Author: paulolieuthier

sqlglot/tokens.py

@@ -1427,7 +1427,7 @@ def _scan_string(self, start: str) -> bool:
         self._advance(len(start))
         text = self._extract_string(end, raw_string=token_type == TokenType.RAW_STRING)
 
-        if base:
+        if base and text:
             try:
                 int(text, base)
             except Exception:

sqlglotrs/src/tokenizer.rs

@@ -468,7 +468,7 @@ impl<'a> TokenizerState<'a> {
             self.extract_string(&end, false, token_type == self.token_types.raw_string, true)?;
 
         if let Some(b) = base {
-            if u128::from_str_radix(&text, b).is_err() {
+            if !text.is_empty() && u128::from_str_radix(&text, b).is_err() {
                 return self.error_result(format!(
                     "Numeric string contains invalid characters from {}:{}",
                     self.line, self.start

tests/dialects/test_spark.py

@@ -1029,6 +1029,16 @@ def test_string(self):
                 query = parse_one("STRING(a)", dialect=dialect)
                 self.assertEqual(query.sql(dialect), "CAST(a AS STRING)")
 
+    def test_binary_string(self):
+        for dialect in ("spark2", "spark", "databricks"):
+            with self.subTest(f"Testing HEX strings for {dialect}"):
+                query = parse_one("X'ab'", dialect=dialect)
+                self.assertEqual(query.sql(dialect), "X'ab'")
+
+            with self.subTest(f"Testing empty HEX strings for {dialect}"):
+                query = parse_one("X''", dialect=dialect)
+                self.assertEqual(query.sql(dialect), "X''")
+
     def test_analyze(self):
         self.validate_identity("ANALYZE TABLE tbl COMPUTE STATISTICS NOSCAN")
         self.validate_identity("ANALYZE TABLE tbl COMPUTE STATISTICS FOR ALL COLUMNS")