interactive global search

[dofinn]

The reason I use vim is for ripgrep and fzf. This makes for extremly fast navigation.

Request:

• Ability to index files in active directory and search via FZF
  

• Ability to grep through fies in active directory and pipe results in fzf

• ripgrep file_config | fzf <- this opens up in same preview window that FZF provides.
  
  

There are many blogs etc writing about this method in vim. EG: https://sidneyliebrand.medium.com/how-fzf-and-ripgrep-improved-my-workflow-61c7ca212861

[kirawi]

Should this be a core feature of Helix or be implemented as a plugin?

[pickfire]

I personally think this should be core feature. We already have like skim like thing, implementing ripgrep is just near.

[dofinn]

is a dependecy on ripgrep something you want to make mandatory? I would imagine having these functions pluggable
grep=plugin(ripgrep)
find=plugin(fzf)

[pickfire]

I don't think we will pull in ripgrep. We may be implementing our own which is slightly slower compared to ripgrep but work well for our usecase.

[vv9k]

We could also just import the grep_searcher crate that ripgrep uses internally. It's published to crates.io

https://docs.rs/grep-searcher/0.1.7/grep_searcher/

And walkdir for recursive file system traversal https://docs.rs/walkdir/2.3.2/walkdir/

[dofinn]

so excuse me for being an idiot. The file picker is already aweesome! cancel the FZF request

[archseer]

I do think a preview window + some way to actually search through files would be useful though :)

I think there should be a general way to pipe data into the built-in picker.

[dofinn]

I agree, after using it last night i did miss the preview window + the actual fuzzy finding where i could use multiple terms seperated by whitespace rather then the exact path/to/target/file

[archseer]

The finding is fuzzy (we use skim's algorithm, we just don't highlight the matched characters yet.

[nrdxp]

Does Helix use skim as a library?

[sudormrfbin]

Does Helix use skim as a library?

We're using the fuzzy-matcher crate which is used internally by skim.

[heliostatic]

Another vote for searching via the grep_searcher crate in the preview UI used for the picker, per #196 (comment)

[pppKin]

Hi! I too think that being able to perform a search in current workspace would be immensely helpful. So I started messing around and implemented something myself . It's not pretty I know, but it's something for us to talk about as least.
My implementation bind space + / to display a prompt (like the search prompt), and on Enter, start the search, gathers the result and pop up a FilePicker. A few necessary changes are made, such as regex_prompt now accepts an extra callback that would be invoked once Enter is pressed.
There may be some best practices I missed or some corner cases that I don't know about, but hopefully this can be improved by the community. So before I rush to create a PR, I just want to know if you guys have any thought on what this should look like and how it should be properly implemented. Thank you!

[archseer]

That's pretty cool! I think instead of grep you will want to import grep_matcher and grep_regex so we avoid some of the extra deps (we don't need grep_printer for example). grep is just a facade that re-exports all these crates: https://github.com/BurntSushi/ripgrep/blob/master/crates/grep/src/lib.rs

I'm also wondering if we might want to do this without the prompt and just do this interactively in the picker as the user types. It's a bit wasteful though so maybe there could be a key to refresh (ctrl-r or something like it)

The binary detection part is really cool and we likely want to use that in the file preview as well to avoid displaying binary files (on space + f).

[pppKin]

import grep_matcher and grep_regex so we avoid some of the extra deps (we don't need grep_printer for example)

You're right, I will be cleaning it up shortly :)

Edited: Updated. But I seem to have messed up my forked branch to include a commit that was not mine(due to some incorrect git rebase I think), after resolving that I'll be creating a PR so that it's easier for everyone to review:)

[pppKin]

just do this interactively in the picker as the user types

@archseer
Interactive picker might be a more intuitive option, but I wonder what would be a proper way to fuzzily select a file in search results in said picker? Perhaps add another input field and another key binding to switch between them?

[archseer]

That would be an option, but from what I've seen in other plugins (telescope.nvim) there's simply no fuzzy file select, you just scroll the list (we could probably add ctrl-d/u page up/down so the list is easier to scroll faster). When I'm searching for something I usually don't know the filename so I'll traverse the list to find the correct entry.

I do see a case for being able to limit the set of files we search for beforehand though, for example if I'm searching for results inside _test.go files. In that case I think we could space + f, _test.go to fuzzy find files, then space + \ should search through that subset of files.

[EpocSquadron]

@archseer I think this is implemented, any reason to keep this open?

[pickfire]

Interactive prompt is not available yet so it will be kept open.

[amstaniec]

I came up with nushell script command interactive global search (external to helix)
using rg and fzf, so also works on MS Windows
It shows only with single line of preview, but still quite usable.
For nushell users, put this in of your config.nu:

def rghx [] {
  let li = ((rg -e '.' --no-heading -n --color always | fzf --ansi) | lines |  split row ':')
  hx ([$li.0, ':', $li.1] | str join)  
}

run it in the repo root folder, and start typing to search, enter to open in helix

can be easily translated to different shell script

[davxy]

This is something I use in bash

#!/bin/bash

# Search using rg and fzf with preview
out=$(rg . --line-number --column --no-heading --glob '!.git' | \
      fzf +i --exact --delimiter : --preview 'batcat --style=full --color=always --highlight-line {2} {1}' --preview-window 'up,~4,+{2}+4/2')

# Remove cruft leaving something like: 'file:line:column'
ref=$(echo $out | sed -E 's/([a-zA-Z0-9/-_]*):([0-9]*):([0-9]*):.*/\1:\2:\3/')
hx $ref

Would be super useful to open the found file in a running instance of helix...
Is there a hack to do this?

[predmijat]

I'm not sure about the running instance, but you can open it in new one by editing FZF_DEFAULT_OPTS variable and adding something like this:

export FZF_DEFAULT_OPTS='... ,ctrl-o:execute(hx $(echo {} | cut -d ":" -f1))

I have this in my .zprofile since I'm using zsh, check where you shell sets FZF_DEFAULT_OPTS.

[obxhdx]

@davxy, and anyone else interested: I hacked something together for this, you can read about it on my post on r/HelixEditor.

[pinpox]

@davxy, and anyone else interested: I hacked something together for this, you can read about it on my post on r/HelixEditor.

This is cool, but limited to wezterm, right? Is there an option to do live-grep'ing project-wide inside helix itself or are there plans to add it?

[obxhdx]

This is cool, but limited to wezterm, right? Is there an option to do live-grep'ing project-wide inside helix itself or are there plans to add it?

It’s like I said on the Reddit post, it’s a opinionated setup but I’m certain it can be adjusted. I’m pretty sure Kitty and Tmux have similar mechanisms for sending keys/text to another pane, for example.

Adding it natively in Helix is the whole point of this issue. I just wanted to share my workaround for those who (like me) are waiting for the native feature.

[matu3ba]

Potential inspiration for a solution (does not have the matches in file preview due to being additionally complexity):
https://github.com/jake-stewart/jfind.nvim
and
https://github.com/jake-stewart/jfind

[gcanat]

Potential inspiration for a solution (does not have the matches in file preview due to being additionally complexity): https://github.com/jake-stewart/jfind.nvim and https://github.com/jake-stewart/jfind

More relevant would be to use this: #7814

[davxy]

Similar to #196 (comment) but using zellij+fzf

https://github.com/davxy/script-nest/blob/main/bin/helix-ext

The scripts allows to do a couple of IMO useful things (see the script help).

Modifications to helix config:

[keys.normal.'space'.'space']
# Extensions using custom script
# Interactive search in current working directory
s = ":pipe-to helix-ext search --current"
# Interactive search in current working directory (include path name in the search)
S = ":pipe-to helix-ext search --current-inc-path"
# Interactive search in current git directory
w = ":pipe-to helix-ext search --workspace"
# Interactive search in current git directory (include path name in the search)
W = ":pipe-to helix-ext search --workspace-inc-path"
# Interactive search in current buffer
b = ":pipe-to helix-ext search --buffer"
# Git status
m = ":pipe-to helix-ext git --status"
# Git blame
B = ":pipe-to helix-ext git --blame"
# Git browse github repo using default browser
O = ":pipe-to helix-ext git --browse"

[chtenb]

I think this feature is related to this issue: #2177
That thread also contains several ideas and workarounds to incorporate tools like broot that provide file-name and file-content search similar to fzf and ripgrep. Instead of re-inventing the wheel one could also consider supporting interaction with external tools through some kind of interface.

[pascalkuthe]

That would involve building in a terminal emulator insixe helix which is much much more work while also being a lot less integrated with the rest of our UI.

@the-mikedavis already has a working implementation of this based on #8021 and #7265. I have deamoed it and find it quite a bitee nicer tgab other options (and especially since it integrates well with the rest of the editor, which is the reason the picker is builtin to begin with). The columary picker really is a very nice approach.

So this issue just needs time for the implementation to be finished and land in master

[the-mikedavis]

Ah I meant to post a demo here after assigning myself: https://asciinema.org/a/607253

Global search becomes a DynamicPicker like workspace symbol search (<space>S). You input a regex that runs across the workspace / cwd and new results are streamed in (like on master after #7814) so the picker doesn't block on become unresponsive, even when you're searching a large directory (my ~/src is ~120GB).

It needs some more tweaking but I plan to make a draft PR for it after #8021 is merged.

[TornaxO7]

As far as I know, Helix uses nucleo when opening the fuzzy file finder. Maybe we could use that for grepping text as well?

[hakan-demirli]

@davxy, and anyone else interested: I hacked something together for this, you can read about it on my post on r/HelixEditor.

Until this issue is fixed here is what I use. I ported the wezterm specific method mentioned in the blog post to tmux since it is terminal agnostic.

Kooha-2023-12-05-15-02-36.online-video-cutter.com.mp4

Add these three files to your path:

• https://github.com/hakan-demirli/dotfiles/blob/main/.local/bin/live-grep
• https://github.com/hakan-demirli/dotfiles/blob/main/.local/bin/live-grep-helix
• https://github.com/hakan-demirli/dotfiles/blob/main/.local/bin/live-grep-in-split-tmux

Add it to your config:

[keys.normal.space]
"/" = ":pipe-to live-grep-in-split-tmux"

[cyruseuros]

@the-mikedavis - don't mean to rush things, just pinging in case this got buried (always does for me). Is there a draft of the code to build/play around with now that #8021 has been merged?

[gabydd]

It's implemented in #9647