Fix support for Go 1.27 using unicode 17

.tables.json

@@ -4,29 +4,35 @@
         "unicode_version": "13.0.0",
         "cldr_version": "32",
         "case_fold_hash": "3ab0454d1a85064a4c401b9a0c7162bd08a306ebd0dc34065b25701e6d38dba9",
-        "gen_go_hash": "3c10a98fa2bf15161cfaf5db1abba335476a0583661c37ea204bddef6d61cba0",
+        "gen_go_hash": "603931ec54064e5bf46593d31672b552c4698d5bef01aefb5da507543e506ca2",
         "table_hashes": {
             "CaseFolds": 2422855,
             "FoldMap": 2521300993,
-            "UpperLower": 1793588629,
-            "_CaseFolds": 2422855,
-            "_FoldMap": 2521300993,
-            "_UpperLower": 1793588629
+            "UpperLower": 1793588629
         }
     },
     "tables_go121.go": {
         "filename": "tables_go121.go",
         "unicode_version": "15.0.0",
         "cldr_version": "32",
         "case_fold_hash": "26ed8b8eee3e8fb11d5bc828b898e9b714fed8d34dea08089dde133d0b10fb04",
-        "gen_go_hash": "3c10a98fa2bf15161cfaf5db1abba335476a0583661c37ea204bddef6d61cba0",
+        "gen_go_hash": "603931ec54064e5bf46593d31672b552c4698d5bef01aefb5da507543e506ca2",
         "table_hashes": {
             "CaseFolds": 4292873350,
             "FoldMap": 2521300993,
-            "UpperLower": 1793588629,
-            "_CaseFolds": 4292873350,
-            "_FoldMap": 4263480156,
-            "_UpperLower": 4290248538
+            "UpperLower": 1793588629
+        }
+    },
+    "tables_go127.go": {
+        "filename": "tables_go127.go",
+        "unicode_version": "17.0.0",
+        "cldr_version": "32",
+        "case_fold_hash": "dc6cc7a02620578ced5f7cff096043d463046a068304443aba325dfc5b3e3f03",
+        "gen_go_hash": "603931ec54064e5bf46593d31672b552c4698d5bef01aefb5da507543e506ca2",
+        "table_hashes": {
+            "CaseFolds": 4292873350,
+            "FoldMap": 935790141,
+            "UpperLower": 1333264157
         }
     }
 }

gen.go

@@ -155,7 +155,7 @@ func realMain(args []string) int {
 	var exitcode int
 	// Supporting Unicode version 12.0.0 is annoying since arm64 support
 	// is lacking on Go 1.15 and below.
-	for _, version := range []string{"13.0.0", "15.0.0"} {
+	for _, version := range []string{"13.0.0", "15.0.0", "17.0.0"} {
 		code := genCmd(append([]string{"-unicode", version}, args...)...)
 		if exitcode == 0 {
 			exitcode = code

internal/gen/gen/gen.go

@@ -89,7 +89,8 @@ var tags = []struct{ version, buildTags string }{
 	{"11.0.0", "go1.13,!go1.14"},
 	{"12.0.0", "go1.14,!go1.16"},
 	{"13.0.0", "go1.16"},
-	{"15.0.0", "go1.21"},
+	{"15.0.0", "go1.21,!go1.27"},
+	{"17.0.0", "go1.27"},
 }
 
 // buildTags reports the build tags used for the current Unicode version.

internal/gen/gentables/main.go

@@ -126,7 +126,8 @@ func loadCaseFolds() {
 
 var buildTags = map[string]struct{ version, buildTags, filename string }{
 	"13.0.0": {"13.0.0", "go1.16,!go1.21", "tables_go116.go"},
-	"15.0.0": {"15.0.0", "go1.21", "tables_go121.go"},
+	"15.0.0": {"15.0.0", "go1.21,!go1.27", "tables_go121.go"},
+	"17.0.0": {"17.0.0", "go1.27", "tables_go127.go"},
 }
 
 // tablesFileName is the names of the file to generate and is based off
@@ -753,10 +754,16 @@ func genFoldTable(w *bytes.Buffer, firstValidHash bool) {
 	runes := make(map[rune][]rune)
 	rangetable.Visit(categories, func(r rune) {
 		ff := folds(r)
-		if len(ff) > 2 {
-			runes[r] = append(runes[r], ff...)
-		}
-		if len(ff) == 1 && toUpper(r) != toLower(r) {
+		switch n := len(ff); {
+		case n > 2:
+			fallthrough
+		case n == 2 && ff[1] != toUpper(r) && ff[1] != toLower(r):
+			// A two-element fold orbit is normally covered by the _UpperLower
+			// table, but when the other rune is neither the upper nor lower
+			// case form of r it cannot be represented there and must be
+			// stored in _FoldMap.
+			fallthrough
+		case n == 1 && toUpper(r) != toLower(r):
 			runes[r] = append(runes[r], ff...)
 		}
 	})
@@ -1267,6 +1274,9 @@ var category = map[string]bool{
 	"S": true, // Sm Sc Sk So
 	"Z": true, // Zs Zl Zp
 	"C": true, // Cc Cf Cs Co Cn
+	// Multi-letter group categories added in go1.27.
+	"LC": true, // Lu Ll Lt (Cased_Letter)
+	"Cn": true, // Unassigned code points
 }
 
 // This contains only the properties we're interested in.
@@ -1363,10 +1373,24 @@ func loadCategoryTables() map[string]*unicode.RangeTable {
 		if _, ok := category[name]; !ok {
 			log.Fatal("unknown category", name)
 		}
+		// Previously, we only distinguished between "unified" and all other
+		// categories. Since go1.27, we need to also handle the "LC" (cased
+		// letter) and "Cn" (unassigned) categories. This code does that.
 		var rt *unicode.RangeTable
-		if len(name) == 1 { // unified categories
+		switch {
+		case name == "LC": // cased letter
+			rt = dumpRange(func(code rune) bool {
+				switch chars[code].category {
+				case "Lu", "Ll", "Lt":
+					return true
+				}
+				return false
+			})
+		case name == "Cn": // unassigned: no general category
+			rt = dumpRange(func(code rune) bool { return chars[code].category == "" })
+		case len(name) == 1: // unified categories
 			rt = dumpRange(func(code rune) bool { return categoryOp(code, name[0]) })
-		} else {
+		default:
 			rt = dumpRange(func(code rune) bool { return chars[code].category == name })
 		}
 		cats[name] = rt