Translate a PO file (gettext) into Japanese

What changes in a .po file when the target is Japanese

Japanese has a single CLDR plural category, other. The header becomes Plural-Forms: nplurals=1; plural=0;, and every entry with a msgid_plural gets exactly one msgstr[0] covering both source forms:

• msgid "%d file" / msgid_plural "%d files" → msgstr[0] "%d 件のファイル"

A number rarely stands alone in Japanese: it takes a counter word, such as 件 for items, 人 for people or 枚 for sheets and images. Picking the right one depends on what the string refers to, which is why msgctxt and #. comments matter.

Word order is also different: the verb comes last. "Delete %s?" becomes "%s を削除しますか？". With two placeholders the order may flip entirely: "Page %1$d of %2$d" becomes "%2$d ページ中 %1$d ページ目". That reordering only works if the source uses positional placeholders (%1$s, %2$d); plain %s %d forces the translation to keep the source order.

Politeness, punctuation and string length

Japanese user interfaces usually address the user in the polite desu/masu form: "ファイルを保存できませんでした。" for “The file could not be saved”. Buttons and menu items, by contrast, are short nouns: "Save" → "保存", "Cancel" → "キャンセル". One catalogue therefore mixes two registers, and each string's context decides which applies.

Punctuation switches to full-width forms: 。 for full stops, 、 for commas, plus ？ and （ ）, with no surrounding spaces. A source string ending in an ellipsis, like "Loading…", becomes "読み込み中…".

Japanese often uses fewer characters than English, but each glyph is roughly twice as wide as a Latin letter, so a button label can still overflow. With no spaces between words, line breaking is left to your application, and the font in use must include kanji or users will see empty boxes.

What Manuce returns, and what it does not do

• The same file: indentation, entry order, quoting, #. #: #, comments, line endings, encoding and BOM are preserved; only the values change.
• msgctxt and msgid_plural are kept, and the context guides the translation.
• A .pot comes back as a Japanese .po with Language: ja and Plural-Forms set; the number of msgstr[n] matches Japanese.
• Printf placeholders (%s, %d, %1$s) and tags are compared with the source for every value; if a mismatch persists, the source value is kept and the report says so.
• Not translated: keys, empty values, URLs, email addresses, numbers, colours, date formats, identifiers and anything marked as non-translatable.

Manuce does not compile the .mo, test the display in your application or choose your font. It does not accept glossaries or translation memories, and it does not produce sworn or certified translations.

Preparing the file before upload

1. Check the encoding. The header should read Content-Type: text/plain; charset=UTF-8. A file in ISO-8859-1 cannot hold Japanese characters, and Manuce keeps the original encoding, so convert it to UTF-8 first.
2. Disambiguate with msgctxt. "Open" may be an action (開く) or a status (公開中); "Post" may be a noun (投稿) or a verb (投稿する). A context such as msgctxt "button" is often enough.
3. Switch to positional placeholders wherever a string holds more than one, so Japanese word order is possible.
4. Add #. comments saying what %s stands for (a file name, a person), since that decides the counter word and particle.
5. Mark product names as non-translatable if they must stay in Latin script.

Checking the Japanese file on return

• Run msgfmt --check on the returned file: it validates printf formats and consistency with Plural-Forms.
• Read the report: a string left in the source language usually points to a placeholder or tag the translation could not respect without distorting the source.
• Compile and look at real screens: truncated buttons, line breaks in the middle of a katakana word, empty boxes from a font without kanji.
• Check sensitive terms: a brand name may be rendered in katakana or left in Latin script. Look at it in the free excerpt first, then in the full translation.

Japanese carries a surcharge (×1.6 coefficient); the count covers words in the source values only, with keys and tags excluded.

In practice

1. Make sure the .po or .pot header declares charset=UTF-8.
2. Upload the file and choose Japanese: the quote appears straight away.
3. Read the free excerpt to judge register and terminology.
4. Download the Japanese .po and its report, then run msgfmt --check.
5. Compile the .mo and check the strings in your application.