Post BA3EK11FHkOKbrbWvg by moody@hj.9fs.net
 (DIR) More posts by moody@hj.9fs.net
 (DIR) Post #BA3EK11FHkOKbrbWvg by moody@hj.9fs.net
       0 likes, 0 repeats
       
       @dalias@hachyderm.io Oh hey something I actually know about this time! I haven't worked on UCA but I did do normalization for 9front, both NFC and NFD[0]. I wrote a paper and did a talk about my experiences working on this stuff for 9front to try and get other folks there to be more interested in Unicode. As of right now the NFD code I wrote passes all the test vectors (and I've been testing against the v18 beta right now, still no issues). I don't fully recall the license conditions for musl but all the code I did for 9front is MIT if you wanted to compare notes :).[0] https://git.9front.org/plan9front/9front/5b902d52edaa25e5f430f024c1a1b9b56f94e3e0/sys/src/libc/ucd/runenorm.cI can link the paper if you think that would help too, would just have to dig it up.
       
 (DIR) Post #BA3Fe1kkMlc0dDzegi by moody@hj.9fs.net
       0 likes, 0 repeats
       
       @dalias@hachyderm.io ah sorry I misinterpreted, sounds similar to the lookahead for recomposition but I'll stop myself before saying anything more stupid.
       
 (DIR) Post #BA3HTtj5YRDKkXW1HE by moody@hj.9fs.net
       0 likes, 0 repeats
       
       @dalias@hachyderm.io recomposition had something similar where everything between two characters with ccc of 0 are candidates and technically you have to loop over it, potentially skipping intermediates until it stops compacting. The spec says this area between them could be unbounded but also provided a "stream safe" variant with some reasonable bounds and a special code point for "continuing potential recomposition".