64 scale factor should only apply to FT funcs about harfpy HOT 12 CLOSED

Looks like all glyph positions are divided by 64, but the code that set font scale does not multiply it by 64. I think the division by 64 should be kept only for when FreeType font functions are used (that I assume multiply the font size by 64).

from harfpy.

Or I guess the Font.scale setter can just call HARFBUZZ.to_position_t() on the value being set.

from harfpy.

I forgot about non-FreeType fonts. Looks like I need to distinguish FreeType and non-FreeType cases. This will have ramifications through the Face, Font and FontFuncs classes.

I may need to make a backward-incompatible change to the API for FontFuncs.

Got to think some more about this.

from harfpy.

OK, I have done a change to the Font.scale attribute to be consistent with the other metrics. Here is the output I now get from running your script:

gid43=0@1540,0+0
gid72=1@1260,0+0
gid79=2@569,0+0
gid79=3@569,0+0
gid82=4@1253,0+0
gid3=5@651,0+0
gid58=6@2025,0+0
gid82=7@1253,0+0
gid85=8@842,0+0
gid79=9@569,0+0
gid71=10@1300,0+0
gid4=11@821,0+0

I still plan to make a backward-incompatible API change to allow more control over the metrics scaling.

from harfpy.

OK, I have committed in d88fbb2 the necessary reworking of all the handling of metrics. Note the new autoscale setting which is automatically set to True for FreeType faces, and which you now have to specify for non-FreeType ones and for FontFunc objects.

I think this properly solves the problem. Unless I misunderstood something about FreeType fonts...

from harfpy.

and which you now have to specify for non-FreeType ones and for FontFunc objects.

Maybe that argument could have a default value of False?

from harfpy.

Is this whole autoscale business appropriate at all? Or should I get rid of it?

from harfpy.

Can be removed I think. I'm not using it, I don't use FT funcs.

from harfpy.

I think it is surprising at least, especially if one is familiar with the C API.

from harfpy.

The 26.6 scaling is done to be consistent with FreeType. When you call FreeType’s FT_Set_Char_Size, you set the size in 26.6-scaled units. Therefore it makes sense to get back glyph metrics in the same units. This is how a high-level API should work.

I think I will leave the autoscale system as it is.

from harfpy.

But would it be possible to make autoscale constructor args default to False?

Right now I have to pass an additional False argument:

face = hb.Face.create_for_tables(
            _spitLayoutTable, layoutTables, None, False)

funcs = hb.FontFuncs.create(False)

So if Face and Font get their autoscale set to True when created from freetype object, why not make it default to False otherwise?

In FontFuncs, the autoscale argument can be inferred when Font.set_funcs(funcs, ..) is called (if FontFuncs.autoscale is None the font sets its autoscale value to the FontFuncs).

from harfpy.

The Font.autoscale and FontFuncs.autoscale settings are independent, and serve different purposes.

In fact, I’ve been thinking whether it might not be useful to generalize to other scale factors besides the shift of 6 used by FreeType.

But maybe I’ll just leave it this way for now.

from harfpy.