Incorrect potential form search pattern

[rbleuse]

Hi Martin,

I noticed some issues when looking for a potential inflection :

Yellow : あえます -> あう, あえる (Potential: ....)
That's not correct, in that case that's the masu form of just あえる.

The rule that is applied is
d.add(EndsWithDeinflector("emasu", VerbInflection.Form.ABLE_TO_DO2, mapOf("u" to InflectableWordClass.VERB_GODAN, "eru" to InflectableWordClass.VERB_ICHIDAN)))

Which I think is redundant because
d.add(ShortVerbDeinflector("masu", VerbInflection.Form.POLITE, mapOf("ru" to InflectableWordClass.VERB_ICHIDAN)))
and
d.add(EruDeinflector())
will give the same results, with the correct form applied for あえる (Polite: ...)

I prepared a fix but I don't know if it will break something. At least unit tests are running fine.
Do you know some side effects if we remove that redundant rule ?

Red : あえる -> あえる, あう (Potential: ....)
For あう that's correct, it's the potential form. But あえる is not correct (あえる is not potential form of あえる).
For that one it's a bit tough because the rule EruDeinflector creates both, so VerbInflection.Form.ABLE_TO_DO2 is applied to both results

Do you know why this rule keeps the original deinflection (eru ending for ichidan) in addition to the potential deinflection (u ending for godan) ?

[rbleuse]

Do you know why this rule keeps the original deinflection (eru ending for ichidan) in addition to the potential deinflection (u ending for godan) ?

By reading comments and the code, I think I understood why, it's for the engine to stop after this deinflection (to keep it as a potential result) and to not deinflect taberu into tabu

[mvysny]

Hi Remi, yes that is true. Unfortunately, the current engine is rather simplistic and this is an unfortunate outcome of that simplicity. But perhaps we can find a workaround 🤔

Could you please open a bug report at https://gitlab.com/mvysny/jputils/-/issues to discuss the issue further?