Translate a PO file (gettext) into Simplified Chinese

What changes in a PO file when the target is Chinese

Under CLDR, Chinese has a single plural category, other. An entry that carries msgstr[0] and msgstr[1] in an English source therefore keeps only msgstr[0], and the header becomes Plural-Forms: nplurals=1; plural=0;. If you upload a .pot template, Manuce returns a .po for Chinese with the Language and Plural-Forms headers set, and the number of msgstr[n] forms matches the target language.

For example, msgid "%d file" / msgid_plural "%d files" becomes a single string, msgstr[0] "%d 个文件". Chinese does not inflect nouns for number; it inserts a measure word (here 个) between the numeral and the noun. The msgid_plural line stays in the file so that ngettext keeps working.

Context matters more than with most European targets, because one English word maps to different characters depending on its role. “Open” on a button is 打开; as the status of a ticket it is more likely 已打开 or 开放. A msgctxt "button" or a translator comment #. is read and steers the translation.

Punctuation, word order and placeholders

Chinese uses full-width punctuation with no space before it: ， 。 ： ？ ！. “File name:” becomes 文件名： and “Delete %s?” becomes 删除 %s？. Mainland usage puts quotations in “ ” and titles of works in 《 》. If your code appends its own colon after a translated label, you end up with a mixed 文件名:, so keep punctuation inside the string.

Word order often shifts. “Deleted %d items from %s” reads naturally as 已从 %2$s 删除 %1$d 个项目, with the location before the verb. That reordering is only possible if the source uses positional placeholders (%1$d, %2$s); with plain %d and %s, the Chinese sentence has to follow the source order and may read stiffly.

Every printf placeholder (%s, %d, %1$s) is checked against the source. If a translated value drops or alters one and the mismatch persists, Manuce keeps the source value for that entry and says so in the report, so you get an untranslated string rather than a crashing printf.

One style point to check: many Chinese interfaces put a space between hanzi and digits or Latin words (%d 个文件, 使用 Google 登录), others do not. Look at the free sample to see whether the result suits your product.

What Manuce returns, and what it does not do

You get the same file in the same format, ready to drop back into your project's locale folder:

• indentation, entry order, quoting, line endings, encoding and BOM preserved; only the msgstr values change;
• msgctxt, msgid_plural and #., #:, #, comments kept, including the c-format flag;
• left untranslated: msgid keys, empty values, URLs, email addresses, numbers, colours, date formats (%Y-%m-%d stays as it is), identifiers and anything marked as non-translatable;
• a report listing translated strings, strings left as they were, and the reason.

What Manuce does not do: it does not compile the .mo, it does not adjust fonts or your UI layout, and it does not produce Traditional Chinese in the same order, as that is a separate language. If you want a Chinese date pattern such as %Y年%m月%d日, that is a decision for your code or a dedicated string. Manuce does not take glossaries: for a product name that must stay in Latin script, check the free sample and then the delivered file. Nor does it provide certified or sworn translation, which is not expected for interface strings anyway.

Preparing the file and checking the result

1. Encoding. The encoding is preserved, so a .po declared as charset=ISO-8859-1 cannot hold Chinese characters. Make sure the header reads Content-Type: text/plain; charset=UTF-8 and that the file really is UTF-8 before uploading.
2. Positional placeholders. For sentences with several variables, switch the source to %1$s, %2$s so the Chinese can follow natural word order.
3. Short, ambiguous strings. “Back”, “Clear”, “Post”: add a msgctxt or a #. comment, otherwise the engine picks the most common sense.
4. Technical check. After delivery, run msgfmt -c: it validates c-format strings and consistency with Plural-Forms. Then compile.
5. Display. Latin-only fonts lack Chinese glyphs, so make sure your font stack includes one that has them (Noto Sans CJK SC, PingFang SC, Microsoft YaHei). Chinese text is two to three times shorter in characters, but each character is wider, so the space taken barely changes: test fixed-width buttons. With no spaces between words, lines break between characters, which can split a two-character term at the end of a line.

In practice

1. Check that the .po or .pot is UTF-8 and that ambiguous strings have a msgctxt.
2. Upload the file, select Simplified Chinese and review the quote and free sample.
3. Download the translated .po and the report of strings left unchanged.
4. Run msgfmt -c, compile the .mo and test the UI with a Chinese-capable font.