wmlxgettext does not recognize plural strings in some contexts

[CelticMinstrel]

I had a line similar to the following:

local var = wesnoth.format(_("one thing","$count things",count), {count = count})

When I ran wmlxgettext, the string was not gathered. Adding a space before the underscore allowed wmlxgettext to recognize it.

[AncientLich]

It was intentional that wmlxgettext requires a space before the underscore (_), but I never figured this particular usage case.

The explaination why wmlxgettext can't recognize it without space:
The underscore symbol could be a part of a variable name (example: variable_name, var_) and I also though that sytuations like
something_ ("one thing", "$count things",count)
shouldn't be recognized as a valid translatable string

since wmlxgettext uses regular expressions in order to parse lua/wml code it is necessary to define some clear rules in order to avoid possible unwanted behaviors.

So... what should I consider a valid translatable, other than
_ at the beginning of the line
_ preceded by space and other things (like: something = _ "translatable string")
also
[ I figure: _ preceded by parenthesis and other things.... like something(_ "translatable string") ]
other cases?

[CelticMinstrel]

You could just say _ at the beginning of a line or preceded by [^a-zA-Z0-9], couldn't you? That said, the cases you listed are probably enough (I doubt people would use translatable strings as a table key, so no need to account for square brackets).

[jostephd]

I think you want to exclude not A-Za-z0-9 but A-Za-z0-9_. You can also do r'\b_\b' to match _ only as a whole word.

[AncientLich]

boundaries seems a very good suggestion. It would make the current regex also simplier, I will follow your suggestion (testing if regex will create or not new bugs... since the regex chain is complex a test is always mandatory before committing any change)

[CelticMinstrel]

By the way, while I have your attention @AncientLich - is it correct that all the wmlxgettext code (apart from the main file) is gathered into that single pywmlx directory?

[AncientLich]

Exactly. pywmlx directory contains all the code of wmlxgettext (imported in main file with 'import pywmlx').
The reason of this choice is that, splitting code in more files with a specific target can make easier to fix or improve the code.

Note:
Committed bugfix (very short change): #3481
I did some test and it seems it works fine. You are welcome to test it too :)