Searching tasks by content

[apraga]

Hi,

I stumbled upon your project on Reddit and quite like several of your design. One thing I cannot wrap my head around is how to search task.

1. find seems to fuzzy match but it returns too many results when searching for a single word. Basically, I want taskwarrior task /MYWORD/.
2. Also, I would like to search the task description by default
3. Finally, I would like to ave a short output, again like taskwarrior. (current output is too verbose !). Your documentation suggest tasklite ndjson but I don't want all the task metadata.

Just my 2 cents but thanks for your work already !

PS: a search like this and dependencies management and I will be sold :) Cross-platform on Windows is a must for me.

[apraga]

Ilustrating point 1. : tasklite find Christine returns 167 (!) result, with for example

449hp  {description: All That Breathes,
       end: 20250620T210916Z, entry: 20250620T210916Z, id: 0,
       modified: 20250620T210916Z, project: films,
       rating: 4.000000, status:
       completed, urgency: 1.87123, uuid:
       68443df3-0148-45b7-adc6-ab67df4a10ba}

by matching c h, s or ti. That is not specific enough as there are only 2 exact matches in the description. I'm better served with nushell than can open the database direclty and feed it into a table

 open ~/.local/share/tasklite/main.db | get tasks | where body =~ Christine

[ad-si]

Fully agreed! I was cleaning up my db a few days ago and I got annoyed by it as well 😅. Expect an update soon!

[ad-si]

I just pushed an update in b9fcbe1 that improves scoring of results for fuzzy search:

• Score continuous matches higher
• Weight body score higher
• Increase necessary minimum score
• Always include the body in result output

Unfortunately it got to slow for big databases and I'll need to improve the performance in a follow up commit.

Also, I like your idea of adding support for Taskwarrior style regexes.
Here is a follow up ticket: #110

[apraga]

After a few quick tests, it’s more intuitive that way ! There is a slight delay when searching.
Two notes:

• the output is too verbose as i only care about the description. When there are many results, it’s not very readable. Maybe it would run faster without less text output ?
• I have tasks named MR-82446. find MR does not find anything butfind MR- works.

[ad-si]

• But it only shows things when they contain matches. So what other output do you have that contains matches? No, the problem for the speed is just the scoring system, not the output, but I already have some ideas on how to make it faster.
• Mh, that's probably because I reduced the minimum score and 2 letters probably aren't enough to reach that score. Maybe it shouldn't use fuzzy matching if the search is only 2 letters as you probably only care actually about exactly that match at this length.

[apraga]

• Great if you can improve the speed ! My suggestion was actually: instead of dumping the full tasks description in JSON, maybe have the same output as open ? My use case is to get the main task information (description, ID, project) in a condensed way. But maybe I’m too used to takswarrior.
• I don’t know if there is a good threshold for exact match. I always have used fuzzy matching in an interactive way (file pickers like fzf or in editor like neovim telescope or emacs). My use case may be generalized to other identifiers. For example, you may want to get all PRs for a project by looking for PR or #.