Hunspell affixes and the regular expressions in the rules like "SFX ^ ати ну (^|[^б])жати"

[ssvb]

The word зажну and some of the others are not recognized by Hunspell, but they can be found at https://vesum.nlp.net.ua/?w=зажати

зажати/^

SFX ^ ати ну (^|[^б])жати
SFX ^ ати неш (^|[^б])жати
SFX ^ ати не (^|[^б])жати
SFX ^ ати немо (^|[^б])жати
SFX ^ ати нете (^|[^б])жати
SFX ^ ати нуть (^|[^б])жати
SFX ^ ати ни (^|[^б])жати
SFX ^ ати німо (^|[^б])жати
SFX ^ ати нім (^|[^б])жати
SFX ^ ати ніть (^|[^б])жати

[arysin]

Yes, there's a problem with regular experessions becoming too complicated in this case and зажну is relatively rare word.
But we could add its forms just as a list (we do that for some other words).

[ssvb]

Is there a full uncompressed list of words available for download somewhere?

[ssvb]

I mean, building everything with gradlew produces out/words.txt and out/words_spell.txt. Which of these two is closer to what needs to be inside of the Hunspell's dictionary? The file name "_spell" implies that it's supposed to be a better choice for a spellchecker. But I see that the Hunspell's uk_UA.dic file includes the word "ангел-хранитель", which is present in out/words.txt and absent in out/words_spell.txt.

Additionally, there's the distr/hunspell/src/main/resources/extra_words.txt file too. What is its purpose? I presume that its content is also added to uk_UA.dic. Have I missed anything else?

Building uk_UA.aff and uk_UA.dic files from the v6.5.4 tag using the instructions from distr/hunspell/README.md produces the results, which differ from the content of hunspell-uk_UA_6.5.4.zip. These files have numerous glitches, such as the Hunspell-incompatible regular expressions mentioned in the title of this bugreport, which cause additional discrepancies. The uk_UA.dic file contains references to some non-existing affix flag 8, which is not defined in uk_UA.aff:

вволити/XNK8
вивіситися/XK8
виглянути/X8
визволити/XK8N
визволитися/XK8

With all of these things, tens of thousands of words are either missing or incorrect. For example, the released Hunspell's dictionary hunspell-uk_UA_6.5.4.zip incorrectly recognizes "+cs=найсправжнісінька" and "матусе" as valid words. And I already mentioned the word "зажну", which is not recognized by Hunspell as a valid word despite being present in both out/words.txt and out/words_spell.txt.

What's the best plan to address these issues?

[ssvb]

I'm developing https://github.com/ssvb/hunaftool. Even though it's still at an early stage, it already can construct a decently compressed and also correct dictionary for Hunspell from a huge list of words. No false positives and no false negatives. If out/words_spell.txt is the actually desired list of words that the Ukrainian spellchecker needs to treat as correct, then please let me know. The affix file needs error corrections and some minor tuning to work efficiently. And common Ukrainian prefixes can be also added for further size reduction.

[arysin]

Yes, the out/words_spell.txt is the words the speller should recognize (there's a full word list but includes many forms that should not be included in spellers).

[ssvb]

Thank you. My hands are a bit full at the moment and I'm busy with other tasks, but I'll build a proper dictionary from out/words_spell.txt in a few days.