.MO File Format – The strings

The header entry

The empty string is the original of the header entry, and its translation is the block of header fields: Content-Type: text/plain; charset=UTF-8\n and the others, each line ended by a line feed. It is the first entry of both tables, since the empty string sorts before every other string, and it carries the information that a loader needs: the encoding of the strings and the plural rule. The manual says the empty original is translated to system information attached to the catalogue. msgfmt 1.0 leaves the field POT-Creation-Date out of what it writes.

Contexts

The original of an entry with a context is the context, then one byte 0x04 (EOT), then the message:

menu 04 Open 00

A context that is empty (msgctxt "") gives an original that starts with the 0x04 byte, which is different from the message alone. The strings are sorted as they are stored, so the entries of one context are not together unless the contexts sort together.

Plural forms

The original of a plural entry is the singular, a NUL byte and the plural original; its stored length covers both, counting the NUL between them but not the one at the end. Only the singular takes part in the hash lookup, because the string that is hashed and compared ends at the first NUL.

The translation holds the translations of the forms, each ended by a NUL byte, one after the other, and the length covers all of them but the last NUL. A program that has the number of the form to use skips that many NULs and reads the string there. If it runs off the end, the first form is the answer. How the number of the form comes from n is the Plural-Forms rule of the header entry.

original:     One file 00 %d files 00
translation:  Un fichier 00 %d fichiers 00

An entry can have both a context and plural forms: the context, 0x04, the singular, a NUL and the plural original.

Order

The originals are sorted by the byte values of the whole stored string, context included, as strcmp compares them. The translation with the same number follows its original. Entries that msgfmt leaves out are the untranslated ones (an empty msgstr), the ones marked fuzzy unless it is told otherwise, and the obsolete ones.

Encoding

The strings are in any encoding that keeps the ASCII characters as they are, as long as the header entry names it in charset=. From gettext 0.22 msgfmt writes UTF-8 unless it is given --no-convert. Wide characters are not used, because their size depends on the platform. A reader decodes each string with the encoding named in the header; the header’s own text is ASCII.

Worked example

A catalogue written by msgfmt 1.0 from a file with these entries, without a header entry. The fuzzy “File %d” is left out; the entry with an empty context sorts first, then the plural one, then the entry with the context menu, and the one that starts with tab:

The file is little-endian, N is 4, the hash table has 5 entries, and the strings start at 0x70.

Not described

Known, not yet covered

• Reference: the conversion that msgfmt makes between encodings, and the encodings it accepts; the manual lists them in another chapter.