contours of zero width can't be exported into a variable font #572

jansindl3r · 2019-08-19T21:16:58Z

Hello, I have a master that has all points at the same x position. It cannot be exported as variable font with fontmake -o variable -m font.designspace. Source is in ufo. I am attaching demo that reproduces this. Can I turn off compatibility check and force the export? See letter A Thanks, Jan
bug.zip

--keep-direction and --keep-overlaps didn't help
CLI:
fontmake -o variable -m bug.designspace --verbose WARNING
Error:
WARNING:fontTools.varLib:glyph A has incompatible masters; skipping

The text was updated successfully, but these errors were encountered:

jansindl3r · 2019-08-23T16:56:14Z

What is uncompatible in this?

  <outline>
    <contour>
      <point type="line" x="72" y="651" />
      <point type="line" x="72" y="101" />
      <point type="line" x="476" y="101" />
      <point type="line" x="476" y="651" />
    </contour>
  </outline>

vs

  <outline>
    <contour>
      <point type="line" x="0" y="651" />
      <point type="line" x="0" y="101" />
      <point type="line" x="0" y="101" />
      <point type="line" x="0" y="651" />
    </contour>
  </outline>

change the second contour to this, makes it compatible... >>>

  <outline>
    <contour>
      <point type="line" x="1" y="651" />
      <point type="line" x="2" y="101" />
      <point type="line" x="3" y="101" />
      <point type="line" x="4" y="651" />
    </contour>
  </outline>

madig · 2019-08-23T18:16:48Z

Cursory look: Maybe the points

      <point type="line" x="0" y="101" />
      <point type="line" x="0" y="101" />

get merged into one point?

jansindl3r · 2019-08-24T06:03:08Z

Most likely, I dont want them to get merged, what could be done, please?
I also believe that only overlapped starting point gets merged - will check tomorrow.

jansindl3r · 2019-08-25T09:30:36Z

this is what Glyphs are able to export into a variable font

master 1
x:83.0 y:129.0 type:line
x:518.0 y:129.0 type:line
x:518.0 y:607.0 type:line
x:83.0 y:607.0 type:line

Master 2 (everything at the same position)
x:300.0 y:350.0 type:line
x:300.0 y:350.0 type:line
x:300.0 y:350.0 type:line
x:300.0 y:350.0 type:line

jansindl3r · 2019-08-26T08:42:34Z

same problem with arg -o ttf-interpolatable ...overlapped startpoint gets merged

UFO:

     <point type="line" x="0" y="651" />
      <point type="line" x="0" y="101" />
      <point type="line" x="0" y="101" />
      <point type="line" x="0" y="651" />

opening this master in GLYPHS, after exporting 'compatible ttfs':

x:0.0 y:101.0 type:line
x:0.0 y:101.0 type:line
x:0.0 y:651.0 type:line

This works as expected though..

-o variable-cff2

madig · 2019-08-26T10:53:18Z

Hm. It werks if I do fontmake.exe -m bug.designspace -o ttf --keep-overlaps but fails with ttf-interpolatable.

madig · 2019-08-26T11:44:33Z

Tracking this down to cu2qu.ufo.fonts_to_quadratic...

TTFPreProcessor seems to use Cu2QuPointPen per individual glyph, whereas TTFInterpolatablePreProcessor uses cu2qu.ufo.fonts_to_quadratic on the same glyphs from all masters.

madig · 2019-08-26T13:23:56Z

This seems to be a feature (?) of fontTools' ReverseContourPen, which qu2cu uses for the conversion to change the winding order. Making the https://github.com/fonttools/fonttools/blob/63fb3fb881440b636267dc43082aab134e40f8f9/Lib/fontTools/pens/reverseContourPen.py#L64 unconditionally true and commenting out https://github.com/fonttools/fonttools/blob/63fb3fb881440b636267dc43082aab134e40f8f9/Lib/fontTools/pens/reverseContourPen.py#L79-L83 turns the input points

[
    (0, 651),
    (0, 101),
    (0, 101),
    (0, 651),
]

into

[
    (0, 651),
    (0, 651),
    (0, 101),
    (0, 101),
]

The associated cu2qu test case would be

    def test_overlapping_points(self):
        glyph = Glyph()
        pen = glyph.getPointPen()
        pen.beginPath()
        pen.addPoint((0, 651), segmentType="line")
        pen.addPoint((0, 101), segmentType="line")
        pen.addPoint((0, 101), segmentType="line")
        pen.addPoint((0, 651), segmentType="line")
        pen.endPath()
        assert glyph_to_quadratic(glyph, reverse_direction=True)
        assert [(p.x, p.y) for p in glyph[0]] == [
            (0, 651),
            (0, 651),  # is this reordering to be expected?
            (0, 101),
            (0, 101),
        ]

which passes with the modification mentioned above. Not sure what to do about this one. I'd expect the contour to not change? @anthrotype

jansindl3r · 2019-08-28T09:03:33Z

@madig thanks for having look on that! I tried to track it, but I wasn't able to go that deep. The reordering is wrong as long as it happens per master, none of the apps out there is able to export my variable font correctly, only FontLab VI for Mac can do it. I don't know, probably they use their own version of FontTools. They have option to trun on and off these "smart features" that shift or remove points randomly

jansindl3r · 2019-09-04T20:40:02Z

@madig I have just checked and FontLab VI exports it correctly and is using the same ReverseContourPen.py as you pointed to. It must be somewhere else, then?

The PointToSegmentPen translates between PointPen and (Segment)Pen protocol. In the SegmentPen protocol, closed contours always imply a final 'lineTo' segment from the last oncurve point to the starting point. So the PointToSegmentPen omits the final 'lineTo' segment for closed contours -- unless the option 'outputImpliedClosingLine' is True (it is False by default, and defcon.Glyph.draw method initializes the converter pen without this option). However, if the last oncurve point is on a "line" segment and has same coordinates as the starting point of a closed contour, the converter pen must always output the closing 'lineTo' explicitly (regardless of the value of the 'outputImpliedClosingLine' option) in order to disambiguate this case from the implied closing 'lineTo'. If it doesn't do that, a duplicate 'line' point at the end of a closed contour gets lost in the conversion. See googlefonts/fontmake#572.

anthrotype · 2019-09-11T13:45:15Z

@jansindl3r I fixed the issue with duplicate points being dropped in fonttools PointToSegmentPen. The fix is available in the latest fontmake v2.0.2 (which requires fonttools v4.0.1).
Please let us know if you find other issues, thanks.

jansindl3r · 2019-09-11T17:17:23Z

@anthrotype Thanks, I have just tested it on my font

using fontmake of version 2.0.2 in virtual enviroment is raising this error.
ERROR:cu2qu.ufo:Glyphs named 'A' have different number of segments Some of the glyphs went though and are displayed correctly, so it is on good way! About half of the glyphs returned error, I can't really discover the pattern. I am using this arguments command fontmake -o variable --keep-overlaps -m bug.designspace --keep-direction also tried with the same result
I have attached source causing it, you find there one glyph that gets exported and one that gets not.

ERROR:cu2qu.ufo:Glyphs named 'A' have different number of segments

bug.zip

anthrotype · 2019-09-13T10:44:22Z

@jansindl3r thanks for the files. I think I found the root of the issue. I will push a fix to cu2qu soon.
Nice rotation effect, btw 👍

When collecting a glyph's segments, we can't simply call the glyphs' draw method with the GetSegmentsPen, but we must initialize the PointToSegmentPen explicitly with outputImpliedClosingLine=True. By default PointToSegmentPen does not outputImpliedClosingLine -- unless last and first point on closed contour are duplicated. Because we are converting multiple glyphs at the same time, we want to make sure the _get_segments function returns the same number of segments, whether or not the last and first point overlap. Fixes googlefonts/fontmake#572 Also see: fonttools/fonttools#1720

anthrotype · 2019-09-13T13:46:16Z

@jansindl3r I have released cu2qu 1.6.6 with the fix and verified that your test font builds successfully with latest fontmake and updated cu2qu.

jansindl3r · 2019-09-13T15:38:01Z

@anthrotype thanks! It works in full version of the font as well.. I am very happy that I can start implementing fontmake into my workflow..

The PointToSegmentPen translates between PointPen and (Segment)Pen protocol. In the SegmentPen protocol, closed contours always imply a final 'lineTo' segment from the last oncurve point to the starting point. So the PointToSegmentPen omits the final 'lineTo' segment for closed contours -- unless the option 'outputImpliedClosingLine' is True (it is False by default, and defcon.Glyph.draw method initializes the converter pen without this option). However, if the last oncurve point is on a "line" segment and has same coordinates as the starting point of a closed contour, the converter pen must always output the closing 'lineTo' explicitly (regardless of the value of the 'outputImpliedClosingLine' option) in order to disambiguate this case from the implied closing 'lineTo'. If it doesn't do that, a duplicate 'line' point at the end of a closed contour gets lost in the conversion. See googlefonts/fontmake#572.

anthrotype mentioned this issue Sep 10, 2019

PointToSegmentPen: preserve duplicate last point fonttools/fonttools#1720

Merged

anthrotype mentioned this issue Sep 13, 2019

cu2qu/ufo: must use PointToSegmentPen with outputImpliedClosingLine=True googlefonts/cu2qu#185

Merged

anthrotype closed this as completed in googlefonts/cu2qu#185 Sep 13, 2019

anthrotype mentioned this issue Feb 17, 2023

Implement a subset of FontTools Pens for write-fonts, leaning on kurbo for things like transform googlefonts/fontations#237

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

contours of zero width can't be exported into a variable font #572

contours of zero width can't be exported into a variable font #572

jansindl3r commented Aug 19, 2019 •

edited

jansindl3r commented Aug 23, 2019 •

edited

madig commented Aug 23, 2019 •

edited

jansindl3r commented Aug 24, 2019

jansindl3r commented Aug 25, 2019

jansindl3r commented Aug 26, 2019 •

edited

madig commented Aug 26, 2019

madig commented Aug 26, 2019 •

edited

madig commented Aug 26, 2019 •

edited

jansindl3r commented Aug 28, 2019

jansindl3r commented Sep 4, 2019 •

edited

anthrotype commented Sep 11, 2019

jansindl3r commented Sep 11, 2019 •

edited

anthrotype commented Sep 13, 2019

anthrotype commented Sep 13, 2019

jansindl3r commented Sep 13, 2019

contours of zero width can't be exported into a variable font #572

contours of zero width can't be exported into a variable font #572

Comments

jansindl3r commented Aug 19, 2019 • edited

jansindl3r commented Aug 23, 2019 • edited

madig commented Aug 23, 2019 • edited

jansindl3r commented Aug 24, 2019

jansindl3r commented Aug 25, 2019

jansindl3r commented Aug 26, 2019 • edited

madig commented Aug 26, 2019

madig commented Aug 26, 2019 • edited

madig commented Aug 26, 2019 • edited

jansindl3r commented Aug 28, 2019

jansindl3r commented Sep 4, 2019 • edited

anthrotype commented Sep 11, 2019

jansindl3r commented Sep 11, 2019 • edited

anthrotype commented Sep 13, 2019

anthrotype commented Sep 13, 2019

jansindl3r commented Sep 13, 2019

jansindl3r commented Aug 19, 2019 •

edited

jansindl3r commented Aug 23, 2019 •

edited

madig commented Aug 23, 2019 •

edited

jansindl3r commented Aug 26, 2019 •

edited

madig commented Aug 26, 2019 •

edited

madig commented Aug 26, 2019 •

edited

jansindl3r commented Sep 4, 2019 •

edited

jansindl3r commented Sep 11, 2019 •

edited