Changelog

v0.10.2

Support for Unicode codepoints outside of the Basic Multilingual Plane. Things like Hieroglyphs should work now! (Courtesy of @flying-kestrel!)

There has been a large overhaul of the public API. Most notably the function sft_char has been replaced by a set of more fine-grained functions, namely sft_lookup, sft_gmetrics, and sft_render. Explanations of these new functions' semantics can be found in the man page.
struct types are now typedef'd, so you can now drop the struct keyword before libschrift's compound types.
The demo program received a complete rewrite courtesy of Andor Badi.

For any user of the library, nothing should have changed at all since v0.9.0, except that performance should be noticeably improved.

Cleaned up the internals. For example, the struct buffer row pointer array is now gone. Instead, we just use a flat cell array.
More suitable tesselation heuristic. The old one was based on the distance of the middle control point to the halfway point between the two other control points. The new one is instead based on the area of the triangles that the three points generate.
A lot of minor performance improvements building up to a relatively significant speedup. Mostly achieved by reducing branch mispredictions and reordering or removing pipeline hazards in the raycaster and the post-processor.

Pushed quantization to a later phase. Should get rid of any quantization errors (i.e. not quite black backgrounds).
Overall performance optimizations.

Fixed a left side bearing issue.
Fixed a situation where the assert in decode_contours could be triggered.
Fields in SFT_Char that are not set by sft_char will now always be initialized with zeros.
Documented sft_linemetrics and sft_kerning in the man page.
All outlines are now decoded before tesselation and rasterization are done in a single pass each.
Optimized clip_points
Changed the internal outline data-structure to be index-based, which also allowed untangling the critical path into a series of small data transformations. This has a huge amount of benefits:
- The internal architecture is now much nicer
- Much better debuggability all-round
- Many new optimization possibilities have opened up
- Less memory consumption than before
- A pleasant 15% speed-boost for sft_char!

This is the first version to have a changelog.