Re: [Tuxpaint-devel] Font scoring method

[Pere P. i C. <per...@gm...>]

El ds. 17 de 12 de 2022 a les 11:03 +0900, en/na Shin-ichi TOYAMA va escriure:
> Hi!
> 
> I've looked into the font scoring method in 'dirwalk.c' and 'fonts.c'.
> 
> The function 'charset_works()' scores up the given font if ALL 
> characters passed to it are correctely rendered.
> 
> However, it seems to have nothing to do with checking if the font 
> renders given characters distinctly. I was not able to understand what 
> qsort() in charset_works() do.

charset_works() passes or fails a font given all characters passed to 
it render and render different.
Sometimes a font substitutes lacking characters by "unknown rectangle", 
that dont count as a fail as the font renders the character, 
to catch this, you need to make 2 characters rendering as "unknow rectangle"


> 
> 
> I confirmed that adding Japanese characters to the translations for the 
> strings used for scoring works to place the font supporting Japanese 
> characters on top.
> 
> The result is good but I am not so confident it is a correct way as 
> originaly intended in font scoring method.

In plus of the font quality translation scoring, there is more score in 
fonts.c based on font name and bold/italic, so is harder for a font that 
has the needed character set but lacks bold/italic variants to raise on top.



> 
> I suppose it is more reasonable to have additional string for scoreing 
> to score up the fonts supporting locale specific characters.
> 
> For example;
> 
>    user_font_styles[num_font_styles]->score += 10 *
>      charset_works(
>        font,
>        /* put locale specific characters to the translation for this
>           string to score up the fonts supporting those characters. */
>        gettext("ABCXYZ")
>    );
> 

All in all I am for this addition, it would raise fonts supporting the locale charset
and left place for quality font scoring with the current system.

Pere