Fix conversion bug when string data comes from binlog (#1571)

* Fix conversion bug when string data comes from binlog

Fixes #1568

When using gh-ost to migrate a table from latin1 to utf8mb3 character encoding, the initial data copy works correctly, but new data with special characters inserted during the migration via binlog replication fails with "Incorrect string value" errors.

The reason for this is that the data is a binary byte array when
converted from the binlog, so the character set conversion is not
applied.

This fix updates the character set conversion logic to apply to both
string and []uint8 types when the column has a character set conversion.

I added a new test for latin1 input to this method and confirmed that
the reproduction from the linked issue is fixed with this change.

* Update whitespace formatting

Author: mattbooks

go/sql/types.go

@@ -53,14 +53,22 @@ type Column struct {
 }
 
 func (this *Column) convertArg(arg interface{}, isUniqueKeyColumn bool) interface{} {
+	var arg2Bytes []byte
 	if s, ok := arg.(string); ok {
-		arg2Bytes := []byte(s)
-		// convert to bytes if character string without charsetConversion.
+		arg2Bytes = []byte(s)
+	} else if b, ok := arg.([]uint8); ok {
+		arg2Bytes = b
+	} else {
+		arg2Bytes = nil
+	}
+
+	if arg2Bytes != nil {
 		if this.Charset != "" && this.charsetConversion == nil {
 			arg = arg2Bytes
 		} else {
 			if encoding, ok := charsetEncodingMap[this.Charset]; ok {
-				arg, _ = encoding.NewDecoder().String(s)
+				decodedBytes, _ := encoding.NewDecoder().Bytes(arg2Bytes)
+				arg = string(decodedBytes)
 			}
 		}
 

go/sql/types_test.go

@@ -49,3 +49,19 @@ func TestBinaryToString(t *testing.T) {
 
 	require.Equal(t, "1b99", cv.StringColumn(0))
 }
+
+func TestConvertArgCharsetDecoding(t *testing.T) {
+	latin1Bytes := []uint8{0x47, 0x61, 0x72, 0xe7, 0x6f, 0x6e, 0x20, 0x21}
+
+	col := Column{
+		Charset: "latin1",
+		charsetConversion: &CharacterSetConversion{
+			FromCharset: "latin1",
+			ToCharset:   "utf8mb4",
+		},
+	}
+
+	// Should decode []uint8
+	str := col.convertArg(latin1Bytes, false)
+	require.Equal(t, "Garçon !", str)
+}