Fixing unicode string split issue (#1364)

https://github.com/bazelbuild/buildtools/pull/1361 caused an issue when splitting a string if the character before the split would be a unicode character since the character was inserted using byte offset.
New version only modifies string content which is known since it matches the regex separator

Co-authored-by: Tim Malmström <oreflow@google.com>

edit/buildozer.go

diff --git a/edit/buildozer.go b/edit/buildozer.go
index 9227878..b4ae487 100644
--- a/edit/buildozer.go
+++ b/edit/buildozer.go
@@ -1043,14 +1043,18 @@
 	if !ok {
 		panic(fmt.Sprintf("splitOnNonEscaped not implemented for %c", sep))
 	}
-	split := re.Split(input, n)
-	offset := 0
-	for i, s := range split[:len(split)-1] {
-		offset += len(s) + 2 // Index starting the following segment.
-		val := split[i] + string(input[offset-2])
-		split[i] = val
+	inputBytes := []byte(input)
+	var splits []string
+	previousIndex := 0
+	// Split n-1 times since it's limiting number of divisions, not the number of segments.
+	for _, find := range re.FindAllIndex(inputBytes, n-1) {
+		splitVal := string(inputBytes[previousIndex:find[1]])
+		splitVal = strings.TrimSuffix(splitVal, string(sep))
+		splits = append(splits, splitVal)
+		previousIndex = find[1]
 	}
-	return split
+	splits = append(splits, string(inputBytes[previousIndex:]))
+	return splits
 }
 
 func removeEscapes(values []string, sep byte) []string {

edit/buildozer_test.go

diff --git a/edit/buildozer_test.go b/edit/buildozer_test.go
index d6c81ee..c1814a4 100644
--- a/edit/buildozer_test.go
+++ b/edit/buildozer_test.go
@@ -1175,6 +1175,13 @@
 			lim:   -1,
 			want:  []string{`one\|two`, "three"},
 		},
+		{
+			name:  "split_with_unicode",
+			input: `😉|❤️|😁`,
+			sep:   '|',
+			lim:   -1,
+			want:  []string{"😉", "❤️", "😁"},
+		},
 	}
 
 	for _, tc := range tests {