gamnit: intro support for Angel Code bitmap font

[xymus]

The BMFont format supports packed textures, varying advance per character and even kernings. It can be generated by a number of tools, including:

• BMFont, free software Windows app, http://www.angelcode.com/products/bmfont/
• Littera, a web app, http://kvazars.com/littera/

This PR also generalizes the Font and TextSprites concepts, previously limited to tile set bitmap fonts. And add a new shortcut service to the excellent DOM parser used to parse the XML description.

The result is much nicer than what was previously available, you can see it in a preview of my latest project:

The parsing method crash easily on malformatted font description file. However, since the description file should be generated, loaded at game start and not often modified, all expected crash errors should occur in dev only. Errors that can happen in production, like a truncated file or a misplaced file, should be detected by the basic sanity checks.

If an error occurs while loading a font, there is currently no automatic fallback and it will probably lead to an unusable game. It could be remedied by adding a default font, similar to the default material and mesh. It would be very useful, even at the beginning of any projects, but there is some technical limitations to implementing it as of now.

In the future, I plan on moving up more services from TileSetFont and add text alignment features. The flat module needs to be split in a group to include more services, including fonts.

[privat]

I would use a name that suggests that the map is a copy and is recreated each time. Maybe attributes_to_map as to assume there is a copy&conversion whereas nothing (or as) assume a view.

[privat]

in duplicated attributes, only the last on is kept.

[xymus]

Since attributes should be unique in well-formed XML, I would prefer to not document this behavior.

[lbajolet]

Agreed, this job should be left to the parser, I'd suggest adding a strict mode to it if there is a need for that, but that's outside the scope of this PR.
Maybe open an issue for that if such a behaviour has not already been implemented?

[privat]

Maybe 0x8B.code_point?
Nit should have literal unicode char '\u008B' or '\U0000080B' added in unescape_nit @R4PaSs

[lbajolet]

Agreed for the unicode char escapes to char, is there an issue for that? If not, could you create one and eventually assign it to me?

[xymus]

I'll leave it as is for now since the commit just move this code from another module. But you can update it with #2461!

[lbajolet]

I just have a little nitpick on the unicode attribute of BMFont, but aside from that, it seems reasonable to me!

[lbajolet]

Agreed, this job should be left to the parser, I'd suggest adding a strict mode to it if there is a need for that, but that's outside the scope of this PR.
Maybe open an issue for that if such a behaviour has not already been implemented?

[lbajolet]

I may miss something, but as this attribute is hard-set to 1, is never used within the scope of this code, and pretty much every font supported also supports Unicode (albeit a subset of it), is it really useful to keep it as an attribute?

[xymus]

The value comes from the font description file, if unicode != 1 then we should look at charset. But since charset is not implemented, unicode it is not very useful as of now.

However, it is important because the charset is used to identify characters by their code point! In a previous version there was an assert unicode == true, however, I removed it since the tool that I use does not set it correctly in the generated file.

So yeah, it's half implemented but it would be worth completing when we have a better understanding of what different tools can generate.