Deinflection cycle testing

[toasted-nutbread]

This change is intended to test for cycles in the deinflection definitions.

Currently this is a draft as it is incomplete, just a proof of concept.

Related: #438, #471

• Rebase and test after fix deinflection bug #547.

[github-actions]

✔️ No visual differences introduced by this PR.

View Playwright Report (note: open the "playwright-report" artifact)

[Casheeew]

What I meant in #471 (comment) was;
At the risk of a contrived example, if there is a rule

{"kanaIn": "てもうた", "kanaOut": "てしまった", "rulesIn": [], "rulesOut": []},
{"kanaIn": "しまった", "kanaOut": "もうた", "rulesIn": [], "rulesOut": []}

then
なってもうた -> なってしまった -> なってもうた -- Cycle!
Which I think is not covered by this implementation yet (my understanding could be wrong)

[toasted-nutbread]

What I meant in #471 (comment) was; At the risk of a contrived example, if there is a rule

{"kanaIn": "てもうた", "kanaOut": "てしまった", "rulesIn": [], "rulesOut": []},
{"kanaIn": "しまった", "kanaOut": "もうた", "rulesIn": [], "rulesOut": []}

then なってもうた -> なってしまった -> なってもうた -- Cycle! Which I think is not covered by this implementation yet (my understanding could be wrong)

Adding that rule group and setting checkRules=true does yield a cycle:

Cycle detected:
  ?てもうた (start)
  ?てしまった (test, =>, てもうた=>てしまった)
  ?てもうた (test, =>, しまった=>もうた)
  ?てしまった (test, =>, てもうた=>てしまった)

[Casheeew]

Adding that rule group and setting checkRules=true does yield a cycle:

My misunderstanding then. 😄

[toasted-nutbread]

In general, the danger comes from situations where kanaOut.length >= kanaIn.length, which is why your first example rule triggered a cycle detection.

[djahandarie]

LGTM. Did you still want to Rebase and test after https://github.com/themoeway/yomitan/pull/547. or are we good to go?