Difference to pyftsubset? about dehinter HOT 3 CLOSED

FYI, the table__g_l_y_f class has a removeHinting method, which the subsetter is also using. I recommend you do the same
https://github.com/fonttools/fonttools/blob/197b36fef4a71fa967cb81ec20a63de1f34b6fd7/Lib/fontTools/ttLib/tables/_g_l_y_f.py#L1161

from dehinter.

hmmm. Did not know that this feature existed in the tool. Too late now :)

Will take a look.

from dehinter.

With default execution of pyftsubset using:

$ pyftsubset  Ubuntu-Regular.ttf --glyphs="*" --no-hinting

vs. dehinter with:

$ dehinter Ubuntu-Regular.ttf

I get a very large diff. pyftsubset appears to change glyph names, remove post table values, remove nameID records, and more by default. The docs state that that it performs optimizations in preparation for use in web font compiles. There is none of that in this project. This simply removes glyf table TrueType instruction set assembly and "hinting associated" tables. The latter is broadly defined and follows the approach used during the execution of ttfautohint -d [FONT].

I am also clearing bit 4 of the head table flags when LTSH and hdmx tables are removed which according to my read of the OT spec appears to follow guidance. Based on n=1 testing with pyftsubset I don't think that is happening. Likely a minor Microsoft only issue? I still need to add the source for VDMX table removals (#3) which will be in the next release. Once that is in place this project is nearly a reverse engineered ttfautohint -d in Python. There are a handful of head table flags bits outside of the bit 4 issue that are not perfect matches to what takes place with ttfautohint -d execution but otherwise there is a clean diff across a number of fonts that I've tested. Need to investigate further...

from dehinter.