Make MCFLY_FUZZY a tuneable int #203
Conversation
|
What happens for folks who have it currently set to |
Any string value resolves to 0, disabling fuzzy searching until an int factor >0 is supplied. It is a little disruptive, but I'm also thinking of people who might otherwise not realize there are improvements -- and, hopefully, getting a few more interested parties from the issue to help test and improve the weighting before this lands. |
|
I think it'd be better to be backward compatible and make 'true' map to '1' (or whatever numerical score you think has the best experience). |
|
Also, please run rustfmt. Thanks! |
|
I had the idea of including match start position in the weighting, so earlier as well as shorter matches are preferred. That makes a big difference already, but a |
| if let Ok(fuzzy) = i16::from_str(&fuzzy) { | ||
| settings.fuzzy = fuzzy; | ||
| } else { | ||
| settings.fuzzy = 2; |
There was a problem hiding this comment.
I think this needs to be something like (in pseudocode) if fuzzy.to_lowercase() != "false" { settings.fuzzy = 2; }
There was a problem hiding this comment.
oops, you're right -- fixed.
|
Thanks! |
|
Released in 0.5.10. |
Experimenting based on discussion in #183
No extra weight:
No elaboration needed, it's not great.
Shorter match weighted x2:
Immediate improvement, with actual ssh commands at the top and most other sorting fairly reasonable. The
./target/relea[se/mcfly search]match is ahead ofvi [sql.zsh], but I've definitely run the former several times in this directory (and the latter never).Shorter match weighted x10:
There's a point of diminishing returns here and 10 may be past it;
./target/relea[se/mcfly search]is pushed way down, and even higher values (x20, x50) don't look much different.