* Add HashPointPen from psautohint (with changes)
* Decompose components
* Use format string and disambiguate critical changes
* Remove "getHash()" in favour of "hash" property
* Add transformation for composites
* Omit base glyph name from component hash
* Use untransformed coords for components
* Add tests
* Add example code
Co-authored-by: Cosimo Lupo <cosimo@anthrotype.com>
Fixes https://github.com/googlefonts/fontmake/issues/558
When drawing a composite glyph with a scaled component using the TTGlyphPen, the bounding
box coordinates may change depending on whether one computes them *before* compiling or
*after* decompiling. Before compiling, component.transform holds double precision floats,
but after compiling and decompiling, these are necessarily clamped to F2Dot14 fixed precision.
The TTGlyphPen needs to quantize transform floats to F2Dot14 so that the values don't
change after compilation.
Without this change, it may happen that round-tripping fonts through ttx (which by default
recalcBBoxes) will produce different bounding boxes for composite glyphs that have
scaled or transformed components.
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 https://github.com/googlefonts/fontmake/issues/572.
This test is not normally run. It is skipped when ufoLib is not importable,
as it's the case from the test runner's virtual environment.
The only reason it exists is so that I could check that the
ReverseContourPen I was writing behaves the same as using ufoLib's
ReverseContourPointPen when the latter is run through the
SegmentToPointPen and PointToSegmentPen converters.
The two methods for reversing contours now diverge since
https://github.com/fonttools/fonttools/pull/1080,
but only nominally because both produce effectively the same results.
The only difference is that, when using the point pens with
outputImpliedClosingLine=True, the closing lineTo is always there even
when it's redundant. In our implmentation, we only output the closing
lineTo when it is the same as moveTo (this was necessary in order to
fix https://github.com/googlei18n/cu2qu/issues/51)
Nevertheless, the total number of points is the same in both cases.
Maybe this test should not be here, as it's testing functionalities
from a different package.
Closes https://github.com/fonttools/fonttools/issues/1081
One way to work around https://github.com/googlei18n/cu2qu/issues/51
when using the ufoLib's ReverseConturPointPen via the converter pens
would be to pass the outputImpliedClosingLine=True argument
(False by default) to the PointToSegmentPen.
This way, all the final lineTos are explicitly outputted, even when
they are redundant and could be implied (i.e. when they are not
duplicate points).
Note that this test is skipped by the CI, because ufoLib is not a
dependency of fonttools; you can run locally if you wish.
