rakhithjk / kss Goto Github PK

Khitan Small Stacked is a PUA-encoded Khitan Small Script font based on BabelStone Khitan Small Glyphs. It is a proof of concept of the CGJ-based encoding proposed in L2/18-121 using OpenType to stack glyphs into clusters of unbounded length. Only vertical clusters embedded in horizontal text are supported, but a true vertical font would follow most of the same principles.

Each Khitan Small Script character has six glyph variants:

• the nominal glyph, not in a cluster
• the initial glyph of a cluster, spanning the entire row (kss1init)
• the initial glyph of a cluster on the left side of the row (kss2init)
• the final glyph of a cluster, spanning the entire row (kss1)
• a non-initial glyph on the left side of the row (kss2)
• a non-initial glyph on the right side of the row (kss3)

The first three are base glyphs and the last three are mark glyphs.

The main logic happens in two contextual chaining substitution lookups.

feature ccmp {
    substitute [@kss1init @kss1 @kss2init] @kss' lookup kss1;
    substitute @kss' lookup kss1init uni034F' @kss;
    substitute @kss' lookup kss2init @kss;
} ccmp;

feature ccmp {
    substitute [@kss2init @kss2] @kss1' lookup kss3;
    substitute @kss1' lookup kss2 @kss1;
} ccmp;

Any two adjacent nominal glyphs must be in the same cluster, so the first lookup converts any initial nominal glyph to its kss1init or kss2init form, depending on whether U+034F COMBINING GRAPHEME JOINER follows, and any trailing nominal glyphs to kss1. The second lookup corrects those kss1 glyphs to kss2 or kss3 depending on the context.

Two more lookups attach the marks to bases and other marks in the usual way.

When a multi-row cluster is embedded in horizontal text, the extra rows should go above the line, so a contextual chaining position lookup raises kss1init and kss2init glyphs depending on how many glyphs follow them in the same cluster. It only handles up to four rows, but it would be simple to extend it if a longer cluster were ever needed.