Optimize i18n to save space

[stoecker]

Currently we use the gettext functions to have i18n. Each language copies all english strings which is a terrible waste of space. We have 21 copies of each string. This could be reduced to 2.

As we anyway have a wrapper we can change it like that:

• store English strings in a hash containing a array index as value
• store i18n strings in arrays
• on tr() parse get the array index from hash and get string from array with the index

Problem: How to handle plural forms?

[xeen]

Plural Forms: I suggest implementing the same method that is used by Firefox:
​https://developer.mozilla.org/en/Localization_and_Plurals

(I followed development at that time and it took quite a while until the devs found something suitable that works for all languages, see ​https://bugzilla.mozilla.org/show_bug.cgi?id=394516 ).

[stoecker]

gettext already has a system (using a numeric syntax to define the forms I believe). The question is how we should handle that when we replace gettext:

• use gettext for plurals also in future
• copy gettext plural stuff into josm
• reimplement it
• something else

Above described hash/array method works as is only for non-plurals.

[stoecker]

Fixed in r2754.

[Gubaer]

Seems to be fixed. Closing.