[GNC-dev] Transaction matcher, code to detect many-to-one matches?

[Geert Janssens]

Op donderdag 7 mei 2020 17:19:24 CEST schreef jean laroche:
> I fully agree that it's non trivial. It's a combinatorial problem in
> theory, to find the match that gives the highest overall score.
> But this is academic musing! In practice, we're not faced with
> complicated types of scenario and the matcher only has to give a "good"
> solution, not necessarily the best.
> 
> But in any case, it should definitely NOT offer to match all OFX to the
> same account transaction!
> That's just a bug.
> Also, it should not try to match an OFX transaction to an already
> cleared, or reconciled transaction.
> That's also a bug.

Why is that a bug ? Particularly in the case of cleared splits I don't think 
it is. I can mark splits as cleared in various ways - during an import or 
simply by clicking in the reconcile column in an account register. There's no 
strong verification on that process so I may simply have cleared a split by 
accident (fat fingers). I agree with John's remark earlier that a cleared or 
reconciled status could lower the probability of being a good match. Outright 
ignoring them would be to strong IMO.

Geert