https://bugs.kde.org/show_bug.cgi?id=340283
--- Comment #39 from Natalie Clarius <natalie_clar...@yahoo.de> --- The more I'm working with runner code the more I think there is potential to make things better besides (instead of or in addition to) introducing yet another source of truth to ranking. Apart from my MRs to make the assignment of match types more balaned between runners, here is a new suggestion: KRunner already manipulates the relevance value, boosting frequently launched results. Move that logic into Milou, where all the rest of the sorting happens anyway, so that then instead of the relevance value (which is only used as a secondary sorting parameter and can never beat a higher match type) it can be made to affect the global ranking. And do something similar for categories, i.e. with each result launch keep a record of its category and use it to boost the rank of the category in the final sorting similar to the multiplier proposed for manual sorting, so that Milou will eventually learn user's preferences on its own. How feasible do you think this is, Alex? -- You are receiving this mail because: You are watching all bug changes.
