Prevent PatchMake panic on invalid UTF-8

diffmatchpatch/patch.go

@@ -16,6 +16,7 @@ import (
 	"regexp"
 	"strconv"
 	"strings"
+	"unicode/utf8"
 )
 
 // Patch represents one patch operation.
@@ -145,6 +146,16 @@ func (dmp *DiffMatchPatch) patchMake2(text1 string, diffs []Diff) []Patch {
 		return patches // Get rid of the null case.
 	}
 
+	// DiffMain replaces invalid UTF-8 with the Unicode replacement character.
+	// Use that normalized source when it is exactly what the diffs describe,
+	// while preserving raw bytes in caller-supplied byte-level diffs.
+	if !utf8.ValidString(text1) {
+		diffText1 := dmp.DiffText1(diffs)
+		if string([]rune(text1)) == diffText1 {
+			text1 = diffText1
+		}
+	}
+
 	patch := Patch{}
 	charCount1 := 0 // Number of characters into the text1 string.
 	charCount2 := 0 // Number of characters into the text2 string.

diffmatchpatch/patch_test.go

@@ -362,3 +362,25 @@ func TestPatchMakeOutOfRangePanic(t *testing.T) {
 	patches := dmp.PatchMake(text1, text2)
 	assert.Equal(t, 6, len(patches), "TestPatchMakeOutOfRangePanic")
 }
+
+func TestPatchMakeInvalidUTF8(t *testing.T) {
+	text1 := string([]byte{0xe0})
+	normalizedText1 := string([]rune(text1))
+	dmp := New()
+	diffs := dmp.DiffMain(text1, "", true)
+
+	for _, patches := range [][]Patch{
+		dmp.PatchMake(text1, ""),
+		dmp.PatchMake(text1, diffs),
+	} {
+		actual, applied := dmp.PatchApply(patches, normalizedText1)
+		assert.Equal(t, "", actual)
+		assert.Equal(t, []bool{true}, applied)
+	}
+
+	// Handcrafted byte-level diffs should continue to use the original bytes.
+	patches := dmp.PatchMake(text1, []Diff{{DiffDelete, text1}})
+	actual, applied := dmp.PatchApply(patches, text1)
+	assert.Equal(t, "", actual)
+	assert.Equal(t, []bool{true}, applied)
+}