Contributing COLRv1

[Dominik Röttsches]

Hi all,

in previous posts to this list, I have sent updates regarding our work
on COLRv1:
https://lists.gnu.org/archive/html/freetype-devel/2020-07/msg00041.html

Under the keyword COLRv1 we subsume a set of proposed specification
extensions which add support for linear and radial gradients,
compositing and glyph transformation and shape reuse to the COLR
table, increasing the table version number from 0 to 1. These
extensions come with support for Variations, which means gradient and
transformation properties can be varied using font variation axes and
bring a compact and expressive color vector font format to OpenType
and the ISO Open Font Format.

We’ve been continuously improving the specification - the latest
version is found here:
https://github.com/googlefonts/colr-gradients-spec/#a2-changes-to-off-5711---color-table

At this point, we believe that the proposed COLRv1 specification is in
good shape and the main concepts of it are stable. Work is progressing
on integrating the specification proposal from the link above into
OpenType and the ISO Open Font Format.

We successfully implemented and tested a prototype of COLRv1 rendering
based on experimental COLRv1 color fonts we made
(https://github.com/googlefonts/color-fonts), a COLRv1 implementation
in FreeType, and a rasterization implementation in Skia
(https://skia-review.googlesource.com/c/skia/+/300558).

With your agreement — and I am happy to hear feedback and answer
questions — I would like to start upstreaming our implementation of
COLRv1 support into FreeType. This contribution is meant as an
addition to the current experimental support of COLRv0.

It brings support to FreeType for understanding the COLRv1 additions
and returns parsed COLRv1 information from the font to a client of the
FreeType API in structured form. We believe this is a useful
abstraction level, as the COLRv1 additions map to concepts of 2D
graphics libraries (Cairo, Skia, etc.) in a straightforward way. So
that it is an effective strategy to implement a COLRv1 rasterizer
using the drawing primitives from the 2D graphics libraries based on
COLRv1 information returned from FreeType.

An example on how the API is used can be found here:
https://gist.github.com/drott/83bca80294eec7db308231f95655d1c0 and in
the Skia work-in-progress implementation at
https://skia-review.googlesource.com/c/skia/+/300558 - Pseudocode for
the drawing algorithm is also part of the spec here:
https://github.com/googlefonts/colr-gradients-spec/#pseudocode

Among the things I would like to hear your feedback on: Would you
prefer a separate feature flag to be used to distinguish between
COLRv0 and COLRv1, or would we keep this new implementation under the
same feature flag TT_CONFIG_OPTION_COLOR_LAYERS?

I published a set of the first 8 proposed changes. Below you find a
list of these. The changes are broken down into addition of new
structs and enums to the FreeType public API, adding new public API
methods for parsing and extracting COLRv1 information, separate
changes for incremental addition of the COLRv1 implementation in
ttcolr.h/cc, and finally the registration of the new API to the sfnt
driver.

* [COLRv1] Add types required for COLRv1 to public API (#59703)
  https://github.com/drott/freetype2-colr/pull/4
* [COLRv1] Add new methods required for COLRv1 to public API (#59703)
  https://github.com/drott/freetype2-colr/pull/5
* [COLRv1] Add implementation to retrieve root paint to ttcolr (#59703)
  https://github.com/drott/freetype2-colr/pull/6
* [COLRv1] Add API to resolve FT_OpaquePaint to FT_COLR_Paint (#59703)
  https://github.com/drott/freetype2-colr/pull/7
* [COLRv1] Add COLR API to iterate color stops (#59703)
  https://github.com/drott/freetype2-colr/pull/8
* [COLRv1] Add COLR API to retrieve color layers (#59703)
  https://github.com/drott/freetype2-colr/pull/9
* [COLRv1] Register new COLR API in sfnt driver (#59703)
  https://github.com/drott/freetype2-colr/pull/10
* [COLRv1] Connect API from SFNT driver to public API (#59703)
  https://github.com/drott/freetype2-colr/pull/11

Support for variations is defined in the specification, but this
initial implementation does not implement support for variable COLRv1
fonts yet. However, we are committed to contributing that as well,
same as we are committed to keeping this implementation in sync should
there be further changes to the spec.

In order to track the upstreaming effort, I filed
https://savannah.nongnu.org/bugs/?59703

If this sounds good to you, can I ask you - Werner and others - to
start reviewing the above pull requests as a first step towards
merging them?

I am looking forward to collaborating with you on this effort, hearing
your feedback and hopefully making good progress on bringing COLRv1
support to FreeType.

Dominik