syntax: basic language code (BCP-47) validation

atproto/syntax/language.go

···
+package syntax
+
+import (
+	"fmt"
+	"regexp"
+)
+
+// Represents a Language specifier in string format, as would pass Lexicon syntax validation.
+//
+// Always use [ParseLanguage] instead of wrapping strings directly, especially when working with network input.
+//
+// The syntax is BCP-47. This is a partial/naive parsing implementation, designed for fast validation and exact-string passthrough with no normaliztion. For actually working with BCP-47 language specifiers in atproto code bases, we recommend the golang.org/x/text/language package.
+type Language string
+
+func ParseLanguage(raw string) (Language, error) {
+	if len(raw) > 128 {
+		return "", fmt.Errorf("Language is too long (128 chars max)")
+	}
+	var langRegex = regexp.MustCompile(`^(i|[a-z]{2,3})(-[a-zA-Z0-9]+)*$`)
+	if !langRegex.MatchString(raw) {
+		return "", fmt.Errorf("Language syntax didn't validate via regex")
+	}
+	return Language(raw), nil
+}
+
+func (l Language) String() string {
+	return string(l)
+}
+
+func (l Language) MarshalText() ([]byte, error) {
+	return []byte(l.String()), nil
+}
+
+func (l *Language) UnmarshalText(text []byte) error {
+	lang, err := ParseLanguage(string(text))
+	if err != nil {
+		return err
+	}
+	*l = lang
+	return nil
+}

atproto/syntax/language_test.go

···
+package syntax
+
+import (
+	"bufio"
+	"fmt"
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestInteropLanguagesValid(t *testing.T) {
+	assert := assert.New(t)
+	file, err := os.Open("testdata/language_syntax_valid.txt")
+	assert.NoError(err)
+	defer file.Close()
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if len(line) == 0 || line[0] == '#' {
+			continue
+		}
+		_, err := ParseLanguage(line)
+		if err != nil {
+			fmt.Println("GOOD: " + line)
+		}
+		assert.NoError(err)
+	}
+	assert.NoError(scanner.Err())
+}
+
+func TestInteropLanguagesInvalid(t *testing.T) {
+	assert := assert.New(t)
+	file, err := os.Open("testdata/language_syntax_invalid.txt")
+	assert.NoError(err)
+	defer file.Close()
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if len(line) == 0 || line[0] == '#' {
+			continue
+		}
+		_, err := ParseLanguage(line)
+		if err == nil {
+			fmt.Println("BAD: " + line)
+		}
+		assert.Error(err)
+	}
+	assert.NoError(scanner.Err())
+}

atproto/syntax/testdata/language_syntax_invalid.txt

···
+jaja
+.
+123
+JA
+j
+ja-
+a-DE
+
+# technically not valid, but allowing in naive parser
+#de-419-DE

atproto/syntax/testdata/language_syntax_valid.txt

···
+ja
+ban
+pt-BR
+hy-Latn-IT-arevela
+en-GB
+zh-Hant
+sgn-BE-NL
+es-419
+en-GB-boont-r-extended-sequence-x-private
+
+# grandfathered
+zh-hakka
+i-default
+i-navajo
+
+# https://github.com/sebinsua/ietf-language-tag-regex/blob/master/test.js
+de-CH-1901
+qaa-Qaaa-QM-x-southern