an option to output unsorted txt

[375gnu]

Could you add an option to disable output sorting? This will allow to use it as a drop-in unmunch replacement. This is low-priority request because I personally switched to csv output format.

[ssvb]

unmunch itself is pretty much unusable and I have no idea why Hunspell still keeps it in its own source tree. This only causes harm, because people waste their time troubleshooting problems in a tool that is already known to be broken. I wasted my time just two weeks ago when I was writing this step-by-step instruction. It happened because I found some misleading information about unmunch on Stack Overflow or somewhere else. Now my own guide can be found by someone else and that unfortunate person may waste their time too. The unfortunate cycle continues... Only Hunspell maintainers can put an end to this by finally deleting unmunch rather than pretending that it might have some utility.

A correct .dic unpacker tool can produce its output in any order. Trying to accurately emulate the same order of output as unmunch would be a fool's errand. The sorted order is also unfortunately locale dependent, but it's better than nothing. Hunaftool does sorting assuming the "C" locale and this is equivalent to:

LC_ALL=C sort -u words.txt > sorted_words.txt

And such sorted list should be identical for any correct unpacker.

[ssvb]

You can check hunspell/hunspell#968 to find an upstream bugreport.

By the way, the @fin-w's proof-of-concept decoder script linked from that bugreport isn't quite correct either. For example, it fails the following testcase:

$ cat test.aff
SET UTF-8
TRY ayvburger
PFX X Y 1
PFX X aay bu aayr
SFX Y Y 1
SFX Y aaa yv/Z aaa
SFX Z Y 1
SFX Z v rger v

$ cat test.dic
1
aaaaa/XY

$ echo "burger" | hunspell -d test
Hunspell 1.7.1
+ aaaaa

$ hunaftool -i=dic -o=txt test.aff test.dic
aaaaa
aayrger
aayv
burger

$ wordforms -g test.aff test.dic test.txt
$ cat test.txt
aaaaa
aayrger
aayv

[fin-w]

Thanks for the heads-up, it's good to see someone else working on this problem! I used hunaftool on a real dictionary (not a test) and it generates fewer entries than my current tool (what that shell script turned into) https://github.com/fin-w/wordforms-rust so it might be worth you looking at the discrepancies too, in case you're missing some entries (all the entries wordforms-rust generates can be spelled by Hunspell). I'll look into improving wordforms-rust to correctly deal with the case you mention here 👍

[375gnu]

I've run both tools with the dictionary I'm working on. It has the following prefix:

PFX u Y 2
PFX u   ў     у          .          # replace lowercase ў with lowercase у, used with common words
PFX u   Ў     У          .          # replace uppercase Ў with uppercase У, used with personal names

hunaftool generates correct new words:

$ grep '^[УуЎў]яўны$' /tmp/hunaftool
уяўны
ўяўны

While wordforms-rust generates an additional word starting with the uppercase У, so the prefix is applied incorrectly:

$ grep '^[УуЎў]яўны$' /tmp/wordforms
уяўны
Уяўны
ўяўны

[ssvb]

@fin-w If you are talking about https://github.com/fin-w/LibreOffice-Geiriadur-Cymraeg-Welsh-Dictionary then the support for the FORBIDDENWORD option was missing. I have added it in the https://github.com/ssvb/hunaftool/tree/forbiddenword branch and will probably merge it soon. Both .dic decoding and encoding should be working with it. Even though encoding doesn't take full advantage of FORBIDDENWORD.

As @375gnu mentioned, https://github.com/fin-w/wordforms-rust seems to produce redundant upcased words. Even though they are spellchecked as correct by Hunspell, they shouldn't be necessary in the list.

And yes, it's definitely good to have multiple implementations to compare. I'm getting bogged down by trying to ensure full compatibility with Hunspell.

[ssvb]

I hope to make a "public release" soon with proper documentation and all the necessary sanity checks to warn about unsupported options, etc.