Add translate anchors filter #514
Conversation
| class OptimizeAnchorsFilter(TransformationsFilter): | ||
| def set_context(self, font, glyphSet): | ||
| # Skip over transformations filter to base filter | ||
| return super(TransformationsFilter, self).set_context(font, glyphSet) |
There was a problem hiding this comment.
better to explicitly call the BaseFilter.set_context(self, font, glyphSet) since that is what you are after.
|
Note to self: the warn should be a debug |
We definitely should not assume that by default. It's legit to leave marks to have correction placement for a bunch of base glyphs and hence not require attachment. |
|
In general, moving glyphs around like that is not something a build tool should do by default; and this particular one feels rather fragile to me. |
|
This does nothing by default. It’s a filter that’s activated via a UFO’s lib, which can also include/exclude specific glyphs. |
I understand. Thanks. |
simoncozens
force-pushed
the
translate-anchors
branch
from
80629c2
to
626701a
Compare
| return False | ||
| # Also skip over marks which are deliberately positioned over the | ||
| # previous glyphs | ||
| if glyph.getBounds().xMax < 0: |
There was a problem hiding this comment.
as a heuristic as to guess whether a mark was "deliberately positioned over the previous glyph", I think this is a bit too fragile. It's conceivable that the xMax still be >= 0 while the mark being intentionally designed to fall on top of another (e.g. narrow) base glyph
|
We should probably update TransformationsFilter at some point so it updates composite glyphs using transformed glyphs as components. Possibly with a new flag that could be enabled in OptimizeAnchorsFilter. |
This is another experimental filter.
Mark glyphs with anchor attachment points can kind of be positioned anywhere; they're going to be repositioned when the mark attachment is done. (There's an assumption here that you probably need to test - what if a mark glyph "gets through" without being attached to anything?)
Because of this, we can translate the whole glyph such that the position of their first attaching anchor (
_whatever) is (0,0). If a lot of anchors end up being zero, they all end up sharing a very small Anchor subtable. This naturally leads to big space savings.