From 0c49f18829ee89cbabb64e4b2788af5f15f5ea91 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Dani=C3=ABl=20van=20Eeden?= <git@myname.nl>
Date: Wed, 14 Jun 2023 12:39:08 +0200
Subject: [PATCH] charset: alias utf8 and utf8mb3 (#44655)

close pingcap/tidb#26226
---
 parser/charset/BUILD.bazel     |  2 +-
 parser/charset/charset.go      | 30 ++++++++++++++++++++++++++----
 parser/charset/charset_test.go | 28 ++++++++++++++++++++++++++++
 3 files changed, 55 insertions(+), 5 deletions(-)

diff --git a/parser/charset/BUILD.bazel b/parser/charset/BUILD.bazel
index f8fa92e88dc82..c437ee959cf28 100644
--- a/parser/charset/BUILD.bazel
+++ b/parser/charset/BUILD.bazel
@@ -42,7 +42,7 @@ go_test(
     ],
     embed = [":charset"],
     flaky = True,
-    shard_count = 7,
+    shard_count = 8,
     deps = [
         "@com_github_stretchr_testify//require",
         "@org_golang_x_text//transform",
diff --git a/parser/charset/charset.go b/parser/charset/charset.go
index 6067e4f623424..668acefcac239 100644
--- a/parser/charset/charset.go
+++ b/parser/charset/charset.go
@@ -106,8 +106,8 @@ func GetSupportedCollations() []*Collation {
 // and returns a boolean.
 func ValidCharsetAndCollation(cs string, co string) bool {
 	// We will use utf8 as a default charset.
-	if cs == "" {
-		cs = "utf8"
+	if cs == "" || cs == CharsetUTF8MB3 {
+		cs = CharsetUTF8
 	}
 	chs, err := GetCharsetInfo(cs)
 	if err != nil {
@@ -117,7 +117,7 @@ func ValidCharsetAndCollation(cs string, co string) bool {
 	if co == "" {
 		return true
 	}
-	co = strings.ToLower(co)
+	co = utf8Alias(strings.ToLower(co))
 	_, ok := chs.Collations[co]
 	return ok
 }
@@ -125,6 +125,8 @@ func ValidCharsetAndCollation(cs string, co string) bool {
 // GetDefaultCollationLegacy is compatible with the charset support in old version parser.
 func GetDefaultCollationLegacy(charset string) (string, error) {
 	switch strings.ToLower(charset) {
+	case CharsetUTF8MB3:
+		return GetDefaultCollation(CharsetUTF8)
 	case CharsetUTF8, CharsetUTF8MB4, CharsetASCII, CharsetLatin1, CharsetBin:
 		return GetDefaultCollation(charset)
 	default:
@@ -148,6 +150,10 @@ func GetDefaultCharsetAndCollate() (defaultCharset string, defaultCollationName
 
 // GetCharsetInfo returns charset and collation for cs as name.
 func GetCharsetInfo(cs string) (*Charset, error) {
+	if strings.ToLower(cs) == CharsetUTF8MB3 {
+		cs = CharsetUTF8
+	}
+
 	if c, ok := CharacterSetInfos[strings.ToLower(cs)]; ok {
 		return c, nil
 	}
@@ -180,9 +186,23 @@ func GetCollations() []*Collation {
 	return collations
 }
 
+func utf8Alias(csname string) string {
+	switch csname {
+	case "utf8mb3_bin":
+		csname = "utf8_bin"
+	case "utf8mb3_unicode_ci":
+		csname = "utf8_unicode_ci"
+	case "utf8mb3_general_ci":
+		csname = "utf8_general_ci"
+	default:
+	}
+	return csname
+}
+
 // GetCollationByName returns the collation by name.
 func GetCollationByName(name string) (*Collation, error) {
-	collation, ok := collationsNameMap[strings.ToLower(name)]
+	csname := utf8Alias(strings.ToLower(name))
+	collation, ok := collationsNameMap[csname]
 	if !ok {
 		return nil, ErrUnknownCollation.GenWithStackByArgs(name)
 	}
@@ -225,6 +245,8 @@ const (
 	CharsetLatin1 = "latin1"
 	// CharsetUTF8 is the default charset for string types.
 	CharsetUTF8 = "utf8"
+	// CharsetUTF8MB3 is 3 bytes utf8, a MySQL legacy encoding. "utf8" and "utf8mb3" are aliases.
+	CharsetUTF8MB3 = "utf8mb3"
 	// CharsetUTF8MB4 represents 4 bytes utf8, which works the same way as utf8 in Go.
 	CharsetUTF8MB4 = "utf8mb4"
 	//revive:disable:exported
diff --git a/parser/charset/charset_test.go b/parser/charset/charset_test.go
index 6de594c68c54d..9f17b9588b37c 100644
--- a/parser/charset/charset_test.go
+++ b/parser/charset/charset_test.go
@@ -44,6 +44,10 @@ func TestValidCharset(t *testing.T) {
 		{"UTF8MB4", "UTF8MB4_bin", true},
 		{"UTF8MB4", "UTF8MB4_general_ci", true},
 		{"Utf8", "uTf8_bIN", true},
+		{"utf8mb3", "", true},
+		{"utf8mb3", "utf8mb3_bin", true},
+		{"utf8mb3", "utf8mb3_general_ci", true},
+		{"utf8mb3", "utf8mb3_unicode_ci", true},
 	}
 	for _, tt := range tests {
 		testValidCharset(t, tt.cs, tt.co, tt.succ)
@@ -145,6 +149,30 @@ func TestValidCustomCharset(t *testing.T) {
 	}
 }
 
+func TestUTF8MB3(t *testing.T) {
+	colname, err := GetDefaultCollationLegacy("utf8mb3")
+	require.NoError(t, err)
+	require.Equal(t, colname, "utf8_bin")
+
+	csinfo, err := GetCharsetInfo("utf8mb3")
+	require.NoError(t, err)
+	require.Equal(t, csinfo.Name, "utf8")
+
+	tests := []struct {
+		cs    string
+		alias string
+	}{
+		{"utf8mb3_bin", "utf8_bin"},
+		{"utf8mb3_general_ci", "utf8_general_ci"},
+		{"utf8mb3_unicode_ci", "utf8_unicode_ci"},
+	}
+	for _, tt := range tests {
+		col, err := GetCollationByName(tt.cs)
+		require.NoError(t, err)
+		require.Equal(t, col.Name, tt.alias)
+	}
+}
+
 func BenchmarkGetCharsetDesc(b *testing.B) {
 	b.ResetTimer()
 	charsets := []string{CharsetUTF8, CharsetUTF8MB4, CharsetASCII, CharsetLatin1, CharsetBin}