From 5933f3bf2671f2091eaf6ecfe08323b8468ef4b0 Mon Sep 17 00:00:00 2001
From: HolyLow <jiaming.xie7@gmail.com>
Date: Tue, 11 Feb 2025 19:04:32 +0800
Subject: [PATCH] fix: regexp_split fails in empty match pattern

---
 velox/functions/lib/Re2Functions.h             | 18 ++++++++++++------
 velox/functions/lib/tests/Re2FunctionsTest.cpp | 11 +++++++++++
 2 files changed, 23 insertions(+), 6 deletions(-)

diff --git a/velox/functions/lib/Re2Functions.h b/velox/functions/lib/Re2Functions.h
index ef6712a73401..510b9ac09597 100644
--- a/velox/functions/lib/Re2Functions.h
+++ b/velox/functions/lib/Re2Functions.h
@@ -408,13 +408,14 @@ struct Re2RegexpSplit {
 
     const auto re2String = re2::StringPiece(string.data(), string.size());
 
-    size_t pos = 0;
+    size_t currPos = 0;
+    size_t lastPos = 0;
     const char* start = string.data();
 
     re2::StringPiece subMatches[1];
     while (re->Match(
         re2String,
-        pos,
+        currPos,
         string.size(),
         RE2::Anchor::UNANCHORED,
         subMatches,
@@ -423,16 +424,21 @@ struct Re2RegexpSplit {
       const auto offset = fullMatch.data() - start;
       const auto size = fullMatch.size();
 
-      out.add_item().setNoCopy(StringView(string.data() + pos, offset - pos));
+      out.add_item().setNoCopy(
+          StringView(string.data() + lastPos, offset - lastPos));
 
-      pos = offset + size;
+      currPos = offset + size;
+      lastPos = currPos;
+      // Only change currPos if encounters empty string match. The lastPos is
+      // not changed to make sure the produced string has correct position
+      // when encountering empty string matches in next match.
       if (UNLIKELY(size == 0)) {
-        ++pos;
+        ++currPos;
       }
     }
 
     out.add_item().setNoCopy(
-        StringView(string.data() + pos, string.size() - pos));
+        StringView(string.data() + lastPos, string.size() - lastPos));
   }
 
  private:
diff --git a/velox/functions/lib/tests/Re2FunctionsTest.cpp b/velox/functions/lib/tests/Re2FunctionsTest.cpp
index a7d349243db9..6512bc6acb56 100644
--- a/velox/functions/lib/tests/Re2FunctionsTest.cpp
+++ b/velox/functions/lib/tests/Re2FunctionsTest.cpp
@@ -1529,6 +1529,17 @@ TEST_F(Re2FunctionsTest, split) {
       {"a", "b"},
   });
   assertEqualVectors(expected, result);
+
+  input = makeRowVector({
+      makeFlatVector<std::string>({
+          "abcd",
+      }),
+  });
+  result = evaluate("regexp_split(c0, '')", input);
+  expected = makeArrayVector<std::string>({
+      {"", "a", "b", "c", "d", ""},
+  });
+  assertEqualVectors(expected, result);
 }
 
 TEST_F(Re2FunctionsTest, parseSubstrings) {