Gettext without a filesystem

Question

For an embedded system with GUI, we are using our custom translation system. The strings are stored in the code ROM of a microcontroller.

Currently we have up to 10 languages and about 400 translated strings (varies depending on the product variant).

Our current system is rather cumbersome and I believe that gettext would be a better solution.

As far as I understand gettext, it requires the use of bindtextdomain to set the directory containing the compiled translations files (*.mo).

Is there a way to instead read the translation from memory? That is I would like to include the compiled *.mo files in the binary, and set up gettext to use these. Alternatively, the translation data would be stored in a data EEPROM without a filesystem.

Or can you recommend a different translation system for use in a microcontroller system (16 or 32 bit, 256 to 512 kbyte ROM) with a C interface?

Edit: I should add that being able to maintain translations apart from the microcontroller firmware would be a reason to switch to gettext. The appropriate translation data would be loaded by the user with a configuration software that we already supply with our systems.

Answer 1

This is what I would do: I would include the binary ".mo" in constant variables in the code. You can write a simple converter from binary to a char array and have that ".mo" file compiled inside your program. You would have an array of ".mo" file data, each with a different language.

I would modify libintl sourcecode to access one of those arrays. Check the file loadmsgcat.c, the function _nl_load_domain. See there how it tries to use mmap() (so it's prepared to have everything in memory). Just add some code there to decide which element of your ".mo" array to use based on the language requested.

I haven't tried this, but this is what I would try given your situation. It doesn't look too hard.