HunspellStemFilter does not work with zero affix

Hello!

Our hunspell aff file contains some rules for suffixes and prefixes that have in affix 0 value. For example: aff file

SET UTF-8
TRY esianrtolcdugmphbyfvkwzESIANRTOLCDUGMPHBYFVKWZ’

FLAG long

SFX AA Y 1
SFX AA er 0 er

dic file

1
worker/AA

So if the work word is analyzed then the worker word must be returned. But returns work - the rule does not apply.

The reason is the value of the affix 0, because changing the rule fixes the analysis:

SFX AA ker k er

I found in code a todo about zero-affix handling fix. But I don’t understand if this todo relates to my problem or it is another one problem.

If this the same problem could you tell me there is a plan to fix it soon?

[NightOwl888]

We are releasing 4.8.0, which is not the most stable 4.x version. Hunspell was buggy until at least 4.10, primarily it seems because there is no offficial spec for it so it took some exploration to make it work. So, we have 3 choices.

1. Strictly follow Lucene 4.8.0 including all of its bugs.
2. Patch Lucene 4.8.0 in some regards to make it less buggy.
3. Plan to release 4.10.4 before moving on to 8.x.

IMO, since we have a long road to get to 8.x we should leave stable software in our wake. So, that either means patching some parts of 4.8.0 to eliminate known bugs or doing a full upgrade of the whole project to 4.10.4 before moving on to 8.x. Clearly, of the two options the patch is simpler.

[rclabo]

I’ll do this one. It can be assigned to me.