NoStringConcat [principle] v1.0.0

Translatable strings must never be assembled by concatenation in code. Word order, plural forms, gender agreement, and grammatical case differ across languages — concatenation hardcodes English grammar and produces broken translations in 6500+ other locales.

Every user-facing string must be a single, complete sentence template owned by the translation pipeline. Variables interpolate via named placeholders ({count}, {name}, {gender}); plurals use ICU MessageFormat plural rules; gender uses ICU select. Code that concatenates two translatable fragments (t('hello') + ' ' + name + t('today')) is broken — translators cannot reorder words, cannot inflect adjectives, cannot insert language-specific punctuation.

Attributed To

Mozilla L10n Best Practices; ICU Project (IBM/Unicode); Mark Davis, 'Why We Need Real Localization' (Unicode Conference 2009).

Applies To

• All translatable strings in any product UI
• Server-rendered emails and notifications
• Push notification payloads (title + body)
• Voice assistant prompts and TTS output
• Marketing copy that ships to multiple locales
• Error messages, validation messages, accessibility labels

Counter Examples

• WRONG: t('You have ') + count + t(' new messages') — Russian needs three plural forms, Arabic six; the template fragments cannot encode this.
• WRONG: ${count} ${count === 1 ? t('item') : t('items')} — hardcodes English plural rule; Polish has 'jeden / dwa / pięć' three-way agreement.
• WRONG: t('Welcome') + ', ' + name + '!' — Japanese punctuation is 「 」, German often capitalizes nouns, Arabic reverses order.
• WRONG: building lists via arr.join(', ') + ' and ' + last — English-only conjunction; use Intl.ListFormat which handles 'A, B, et C' (French) and Arabic RTL.

NoStringConcat [principle] v1.0.0

Translatable strings must never be assembled by concatenation in code. Word order, plural forms, gender agreement, and grammatical case differ across languages — concatenation hardcodes English grammar and produces broken translations in 6500+ other locales.

Every user-facing string must be a single, complete sentence template owned by the translation pipeline. Variables interpolate via named placeholders ({count}, {name}, {gender}); plurals use ICU MessageFormat plural rules; gender uses ICU select. Code that concatenates two translatable fragments (t('hello') + ' ' + name + t('today')) is broken — translators cannot reorder words, cannot inflect adjectives, cannot insert language-specific punctuation.

Attributed To

Mozilla L10n Best Practices; ICU Project (IBM/Unicode); Mark Davis, 'Why We Need Real Localization' (Unicode Conference 2009).

Applies To

• All translatable strings in any product UI
• Server-rendered emails and notifications
• Push notification payloads (title + body)
• Voice assistant prompts and TTS output
• Marketing copy that ships to multiple locales
• Error messages, validation messages, accessibility labels

Counter Examples

• WRONG: t('You have ') + count + t(' new messages') — Russian needs three plural forms, Arabic six; the template fragments cannot encode this.
• WRONG: ${count} ${count === 1 ? t('item') : t('items')} — hardcodes English plural rule; Polish has 'jeden / dwa / pięć' three-way agreement.
• WRONG: t('Welcome') + ', ' + name + '!' — Japanese punctuation is 「 」, German often capitalizes nouns, Arabic reverses order.
• WRONG: building lists via arr.join(', ') + ' and ' + last — English-only conjunction; use Intl.ListFormat which handles 'A, B, et C' (French) and Arabic RTL.

Sources

Examples

• Correct: t('items.cart.summary', { count: 3 }) → ICU template '{count, plural, one {# item} other {# items}} in cart'. Translators handle Slavic 3-form plurals (one/few/many).
• Correct: t('user.greeting', { name: user.name, gender: user.gender }) → '{gender, select, female {Welcome back, {name}!} male {Welcome back, {name}!} other {Welcome back, {name}!}}' — Spanish/French inflect adjectives; English does not, but the template supports both.
• Correct: dates via Intl.DateTimeFormat, numbers via Intl.NumberFormat, lists via Intl.ListFormat — never manual ' and ' joining.

Relations

requires: @community/pattern-icu-message-format

Source

• ICU MessageFormat 2.0 specification (Unicode Technical Standard 35, Part 2)
• Mozilla Localization Best Practices — 'Don't concatenate strings'
• GNU gettext manual — Section on plural forms and message contexts
• W3C Internationalization Working Group — 'Localizable Strings' best practice

Requires

@community/pattern-icu-message-format