Merge pull request #1708 from dolthub/daylon/fix-issue-5699

Allow uppercase collations

Author: Hydrocharged

enginetest/queries/charset_collation_engine.go

@@ -43,6 +43,19 @@ type CharsetCollationEngineTestQuery struct {
 // CharsetCollationEngineTests are used to ensure that character sets and collations have the correct behavior over the
 // engine. Return values should all have the `utf8mb4` encoding, as it's returning the internal encoding type.
 var CharsetCollationEngineTests = []CharsetCollationEngineTest{
+	{
+		Name: "Uppercase and lowercase collations",
+		Queries: []CharsetCollationEngineTestQuery{
+			{
+				Query:    "CREATE TABLE test1 (v1 VARCHAR(255) COLLATE utf16_unicode_ci, v2 VARCHAR(255) COLLATE UTF16_UNICODE_CI);",
+				Expected: []sql.Row{{types.NewOkResult(0)}},
+			},
+			{
+				Query:    "CREATE TABLE test2 (v1 VARCHAR(255) CHARACTER SET utf16, v2 VARCHAR(255) CHARACTER SET UTF16);",
+				Expected: []sql.Row{{types.NewOkResult(0)}},
+			},
+		},
+	},
 	{
 		Name: "Insert multiple character sets",
 		SetUpScript: []string{

enginetest/queries/charset_collation_wire.go

@@ -36,6 +36,19 @@ type CharsetCollationWireTestQuery struct {
 // CharsetCollationWireTests are used to ensure that character sets and collations have the correct behavior over the
 // wire. Return values should all have the table encoding, as it's returning the table's encoding type.
 var CharsetCollationWireTests = []CharsetCollationWireTest{
+	{
+		Name: "Uppercase and lowercase collations",
+		Queries: []CharsetCollationWireTestQuery{
+			{
+				Query:    "CREATE TABLE test1 (v1 VARCHAR(255) COLLATE utf16_unicode_ci, v2 VARCHAR(255) COLLATE UTF16_UNICODE_CI);",
+				Expected: []sql.Row{{types.NewOkResult(0)}},
+			},
+			{
+				Query:    "CREATE TABLE test2 (v1 VARCHAR(255) CHARACTER SET utf16, v2 VARCHAR(255) CHARACTER SET UTF16);",
+				Expected: []sql.Row{{types.NewOkResult(0)}},
+			},
+		},
+	},
 	{
 		Name: "Insert multiple character sets",
 		SetUpScript: []string{

sql/charactersets.go

@@ -14,7 +14,11 @@
 
 package sql
 
-import "github.com/dolthub/go-mysql-server/sql/encodings"
+import (
+	"strings"
+
+	"github.com/dolthub/go-mysql-server/sql/encodings"
+)
 
 // CharacterSet represents the character set of a string.
 type CharacterSet struct {
@@ -204,7 +208,7 @@ var SupportedCharsets = []CharacterSetID{
 // ParseCharacterSet takes in a string representing a CharacterSet and returns the result if a match is found, or an
 // error if not.
 func ParseCharacterSet(str string) (CharacterSetID, error) {
-	if cs, ok := characterSetStringToID[str]; ok {
+	if cs, ok := characterSetStringToID[strings.ToLower(str)]; ok {
 		return cs, nil
 	}
 	// It is valid recognize an empty string as the invalid charset, as some analyzer steps may temporarily use the

sql/collations.go

@@ -17,6 +17,7 @@ package sql
 import (
 	"fmt"
 	"io"
+	"strings"
 	"unicode/utf8"
 
 	"github.com/cespare/xxhash"
@@ -33,9 +34,16 @@ type Collation struct {
 	IsCompiled   bool
 	SortLength   uint8
 	PadAttribute string
-	Sorter       func(r rune) int32
+	Sorter       CollationSorter
 }
 
+// CollationSorter is a collation's sort function. When given a rune, an integer is returned that represents that rune's
+// order when sorted against all other runes. That integer is referred to as a sort order. When two runes have the same
+// sort order, they are considered equivalent. For example, case-insensitive collations return the same sort order for
+// uppercase and lowercase variants of a character, while case-sensitive collations return different sort orders.
+// Comparing sort orders from different collations is meaningless, and therefore represents a logical error.
+type CollationSorter func(r rune) int32
+
 // CollationsIterator iterates over every collation available, ordered by their ID (ascending).
 type CollationsIterator struct {
 	idx int
@@ -758,7 +766,7 @@ func ParseCollation(characterSetStr *string, collationStr *string, binaryAttribu
 		if collationStr == nil || len(*collationStr) == 0 {
 			return Collation_Unspecified, nil
 		}
-		if collation, ok := collationStringToID[*collationStr]; ok {
+		if collation, ok := collationStringToID[strings.ToLower(*collationStr)]; ok {
 			if binaryAttribute {
 				return collation.CharacterSet().BinaryCollation(), nil
 			}
@@ -776,7 +784,7 @@ func ParseCollation(characterSetStr *string, collationStr *string, binaryAttribu
 			}
 			return characterSet.DefaultCollation(), nil
 		}
-		collation, exists := collationStringToID[*collationStr]
+		collation, exists := collationStringToID[strings.ToLower(*collationStr)]
 		if !exists {
 			return Collation_Unspecified, ErrCollationUnknown.New(*collationStr)
 		}
@@ -911,7 +919,7 @@ func (c CollationID) HashToBytes(str string) ([]byte, error) {
 
 // Sorter returns this collation's sort function. As collations are a work-in-progress, it is recommended to avoid
 // using any collations that return a nil sort function.
-func (c CollationID) Sorter() func(r rune) int32 {
+func (c CollationID) Sorter() CollationSorter {
 	return collationArray[c].Sorter
 }