> ~everyone wants lower data storage and wants some kind of compression.
> Can this be made to automatically retrain when analyzing (makes sense?)?
> And create a new dictionary only if it changes compared to the last one.
It's an interesting idea. However I doubt it could be automated in
one-size-fits-all manner. One users would like to do re-learning during
analyzing, others during vacuum, once a month or say using triggers and
some sort of heuristics.
Despite that I see no reason not to accept pull requests with
implementations of different re-learning automation strategies. It's
just not a priority for me personally.
--
Best regards,
Aleksander Alekseev