[Support] Fix behavior of StringRef::count with overlapping occurrences, add tests
authorJohannes Doerfert <johannes@jdoerfert.de>
Tue, 24 Dec 2019 23:25:37 +0000 (17:25 -0600)
committerJohannes Doerfert <johannes@jdoerfert.de>
Wed, 25 Dec 2019 00:30:41 +0000 (18:30 -0600)
Summary:
Fix the behavior of StringRef::count(StringRef) to not count overlapping occurrences, as is stated in the documentation.
Fixes bug https://bugs.llvm.org/show_bug.cgi?id=44072

I added Krzysztof Parzyszek to review this change because a use of this function in HexagonInstrInfo::getInlineAsmLength might depend on the overlapping-behavior. I don't have enough domain knowledge to tell if this change could break anything there.

All other uses of this method in LLVM (besides the unit tests) only use single-character search strings. In those cases, search occurrences can not overlap anyway.

Patch by Benno (@Bensge)

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D70585

llvm/lib/Support/StringRef.cpp
llvm/unittests/ADT/StringRefTest.cpp

index 4bafc4e..d7fa99d 100644 (file)
@@ -374,9 +374,14 @@ size_t StringRef::count(StringRef Str) const {
   size_t N = Str.size();
   if (N > Length)
     return 0;
-  for (size_t i = 0, e = Length - N + 1; i != e; ++i)
-    if (substr(i, N).equals(Str))
+  for (size_t i = 0, e = Length - N + 1; i < e;) {
+    if (substr(i, N).equals(Str)) {
       ++Count;
+      i += N;
+    }
+    else
+      ++i;
+  }
   return Count;
 }
 
index 2e5159d..cbb2a30 100644 (file)
@@ -509,6 +509,13 @@ TEST(StringRefTest, Count) {
   EXPECT_EQ(1U, Str.count("hello"));
   EXPECT_EQ(1U, Str.count("ello"));
   EXPECT_EQ(0U, Str.count("zz"));
+
+  StringRef OverlappingAbba("abbabba");
+  EXPECT_EQ(1U, OverlappingAbba.count("abba"));
+  StringRef NonOverlappingAbba("abbaabba");
+  EXPECT_EQ(2U, NonOverlappingAbba.count("abba"));
+  StringRef ComplexAbba("abbabbaxyzabbaxyz");
+  EXPECT_EQ(2U, ComplexAbba.count("abba"));
 }
 
 TEST(StringRefTest, EditDistance) {