Let function validation fail for lookaround pattern in RE2-based impl…

…ementation (facebookincubator#124)
liujiayi771 · Feb 27, 2023 · b5ae610 · b5ae610
1 parent 74de41d
commit b5ae610
Showing 1 changed file with 8 additions and 0 deletions.
diff --git a/velox/functions/sparksql/RegexFunctions.cpp b/velox/functions/sparksql/RegexFunctions.cpp
@@ -53,6 +53,14 @@ void ensureRegexIsCompatible(
       // instead adds the character ].
     } else if (*c == ']' && charClassStart + 1 != c) {
       charClassStart = nullptr;
+    } else if (*c == '(' && c + 3 < pattern.end() && *(c + 1) == '?') {
+      // RE2 doesn't support lookaround (lookahead or lookbehind), so we should exclude such patterns:
+      // (?=), (?!), (?<=), (?<!).
+      if (*(c + 2) == '=' ||  *(c + 2) == '!') {
+        VELOX_USER_FAIL("{} does NOT support lookahead in RE2-based implementation.", functionName)
+      } else if (*(c + 2) == '<' && (*(c + 3) == '=' ||  *(c + 3) == '!')) {
+        VELOX_USER_FAIL("{} does NOT support lookbehind in RE2-based implementation.", functionName)
+      }
     }
   }
 }