Implement anchor extraction from OpenType files

[rimas-kudelis]

Fixes #67

Disclaimer: I'm not really a Python developer, so it's possible that I didn't do some things the right way. Fell free to help me improve this code. :)

[rimas-kudelis]

Hmmm, I just noticed that my code failed to extract two anchor groups (is this the correct term to use?) from a TTF I'm working on. I have twelve of them in the font, but only the first ten (0-9) got extracted. Weird.

[rimas-kudelis]

Hmmm, I just noticed that my code failed to extract two anchor groups (is this the correct term to use?) from a TTF I'm working on. I have twelve of them in the font, but only the first ten (0-9) got extracted. Weird.

False alarm. It appears the TTF fonts I'm working on use incorrect feature tags for the last two lookups (zz11 and zz12 instead of mark). Fixing that bit in a ttx dump and recompiling the font causes the anchors to be exported correctly.

So, will this PR be accepted or do I need to do something more?

[jenskutilek]

You can rewrite this without using the defcon.Anchor object and thus get rid of the defcon dependency:

        for base in baseAnchors.keys():
            try:
                anchor = {
                    "x": baseAnchors[base]["x"],
                    "y": baseAnchors[base]["y"],
                    "name": f"Anchor-{groupIndex}",
                }
                destination[base].appendAnchor(anchor)
            except:
                continue
        for mark in markAnchors.keys():
            try:
                anchor = {
                    "x": markAnchors[mark]["x"],
                    "y": markAnchors[mark]["y"],
                    "name": f"_Anchor-{groupIndex}",
                }
                destination[mark].appendAnchor(anchor)
            except:
                continue

[jenskutilek]

I wonder about the try...except statements here. In what case could it go wrong, and would we want to know about it instead of silently ignoring the anchors?

[rimas-kudelis]

I guess you're right. I didn't want my additions to potentially break the code, but I suppose it's better to fix it properly if something breaks instead of pretending like nothing happened.

[rimas-kudelis]

You can rewrite this without using the defcon.Anchor object and thus get rid of the defcon dependency:

Done. Also rebased on the current master and force-pushed everything in a single commit.

[jenskutilek]

False alarm. It appears the TTF fonts I'm working on use incorrect feature tags for the last two lookups (zz11 and zz12 instead of mark). Fixing that bit in a ttx dump and recompiling the font causes the anchors to be exported correctly.

IIRC those are unofficial feature tags used by MS VOLT to store its "source code".

[rimas-kudelis]

False alarm. It appears the TTF fonts I'm working on use incorrect feature tags for the last two lookups (zz11 and zz12 instead of mark). Fixing that bit in a ttx dump and recompiling the font causes the anchors to be exported correctly.

IIRC those are unofficial feature tags used by MS VOLT to store its "source code".

Ah, I see. Indeed, all the anchors were added to these fonts using MS VOLT. Apparently, the original designer created the two lookups in question, but did not reference them from the mark feature under latn/dflt (unlike other lookups), so they were/are basically unused. I guess he wasn't satisfied with the result but opted not to waste time on a feature they didn't plan to use (using the combining solidus characters U+0337 and U+0338). It all makes sense now.

[benkiel]

This looks good to me. Thanks for the nudge and the update.