Ideas on possibly improving matching

[murlakatamenka]

There is a new project released with similar functionality:

https://github.com/homerours/jumper

Jumper is a command-line program that helps you jumping to / fuzzy-finding the directories and files that you frequently visit. It uses FZF for fuzy-finding and is heavily inspired by z.

It differentiates itself from the plethora of similar tools on the following points:

• It is not restricted to folders. It allows to quickly navigate files, or anything you want (you can easily create and query a new custom database).
• Efficient ranking mechanism which combines the "frecency" of the match (as z does) and the accuracy of the match (as fzf or fzy do). More details here.
• It allows fuzzy-finding.

---

Some of its ideas can be probably used to improve zoxide, that is already well-known and popular.

• https://github.com/homerours/jumper/blob/4b5b42c52a7118c7a8c48ab0916ba45001f9478f/README.md
• https://github.com/homerours/jumper/blob/4b5b42c52a7118c7a8c48ab0916ba45001f9478f/doc/algorithm.md

---

Not really an issue, but hopefully of some value in general.

[anasouardini]

I'm a noob at this, but from my experience, I think a matching algorithm with above ~60% accuracy in this regard with the few criteria it depends on is impossible.

Getting a bunch of numbers to predict what a human is thinking is so hard, the tool that aims for that would at least need to track other user activities on the system which might make it a privacy concern.

I've used zoxide, z, jump, portal (my creation) and non of them predict my target path consistently, at first, it feels like the new tool is better until you have tons of paths in the DB, it's always going to be a hit-or-miss.

I might be wrong, but that's how I see it.