Page 2 of 2

Re: [jiri][0.16.12] Unintuitive search term parsing

Posted: Thu Jan 11, 2018 5:09 pm
by jiri
I've spent some time today looking at the various string distance algorithms and didn't find a solution I'd like outright. I'll spend some more time playing around and testing various solutions, until I find one that's intuitive. Thank you for the feedback, this is a feature that I want to make as user friendly as possible, so this helps a lot with that :)

In the meantime, I've disabled the fuzzy search by default and restored the previous search functionality, but the fuzzy search is still available through a checkbox in the interface settings, in case you want to play with it.

Re: [jiri][0.16.12] Unintuitive search term parsing

Posted: Thu Jan 11, 2018 5:14 pm
by Bilka
jiri wrote:I've spent some time today looking at the various string distance algorithms and didn't find a solution I'd like outright. I'll spend some more time playing around and testing various solutions, until I find one that's intuitive. Thank you for the feedback, this is a feature that I want to make as user friendly as possible, so this helps a lot with that :)

In the meantime, I've disabled the fuzzy search by default and restored the previous search functionality, but the fuzzy search is still available through a checkbox in the interface settings, in case you want to play with it.
Thank you! I think the tooltip for it should be more descriptive though, it doesnt really help if you dont already know what it does.

Re: [jiri][0.16.12] Unintuitive search term parsing

Posted: Tue Jan 23, 2018 2:36 pm
by eradicator
Thanks for sneaking the fix into 0.16.17. I sure noticed even though it's not in the changelog. :P

And..i hate to be that guy who comes back and complains that the fix wasn't enough, buuuut... concatenating multiple search terms in non-fuzzy mode using an OR junction leads to the weird situation where typing more search terms make the result less precise. I.e. it's more like i'm searching for several different things at once and not doing one more precise search? (Hm. I noticed there's "" mode now while typing this, which kinda solves this.... so i'm left with just "thank you" :p).
Though there was some weird logistics GUI jumping going on when searching (related to the very narrow interface i used for taking the video), so i'll post my demo-gif anyway:
multisearch.gif
multisearch.gif (3.11 MiB) Viewed 2023 times

Re: [jiri][0.16.12] Unintuitive search term parsing

Posted: Thu Feb 01, 2018 2:53 am
by kovarex
jiri wrote:I've spent some time today looking at the various string distance algorithms and didn't find a solution I'd like outright. I'll spend some more time playing around and testing various solutions, until I find one that's intuitive. Thank you for the feedback, this is a feature that I want to make as user friendly as possible, so this helps a lot with that :)

In the meantime, I've disabled the fuzzy search by default and restored the previous search functionality, but the fuzzy search is still available through a checkbox in the interface settings, in case you want to play with it.
In that case, you should probably move this bug to the resolved for the next release category.

Re: [jiri][0.16.12] Unintuitive search term parsing

Posted: Thu Feb 01, 2018 7:42 am
by quyxkh
As I recall code I wrote . . . looonng ago . . . heuristics work really well on sorting matches: assess penalties per matching character that skip s (in decreasing penalty order) word starts, inside-word text, and a minimal penalty for leaving an unmatched suffix (to give complete matches a 0 penalty). WIth identifiers in programming languages you have to get cute about identifying word starts, case shifts and runs of same cases and all, but that won't come into play here.