feat: Add local search in buffer

[oxcrow]

Hi everyone,

I added a feature to do local search in the currently opened document buffer to improve user experience.

This feature is similar to global_search but is only limited to one document.

The command is activated by space+l and seems to work well.

Kindly review and allow this feature to be merged.

Thank you
~oxcrow

Reasoning

• The global_search method is quite nice since it works in a fuzzy finding method and provides a table of matches that we can easily preview to find what we want.
• A similar feature is needed for searching through the currently opened buffer (i.e only one document). We could use global_search however it produces too many false positive results from other files, which makes it difficult to use.
• The standard local search method using / does not feel very good to use.
• Other editors such as Emacs/Neovim provide plugins for local search such as Swiper, which feel better to use https://github.com/abo-abo/swiper
• This feature is implemented to provide a similar feature as Swiper.

Implementation

• I took the code of global_search and modified it little bit.
• Specifically I modified the code to only search through the current document.
• The code works but it should be improved by someone more experienced (if necessary).

[nik-rev]

Showing the path to the file in the global search picker where there can be lots of files makes sense. But here it doesn't really make sense, because you're searching in the same file - so all picker items will show the same path

Maybe instead of showing the path, show the line number and the string contents of that line in the 2nd column

[oxcrow]

@nik-rev Thank you for your feedback.

I was able to implement and push the changes you requested.

Here is the result.

On another note, I also found a bug that needs to be fixed. The local_search method seems to also search in all opened buffers. As shown in this image the search result contains results from two files README.md and test.txt.

I will try to fix it.

[nik-rev]

Nice! So you have a "line number" column. What about having a 2nd column for "text on that line"? Since there is quite a lot of whitespace

[oxcrow]

I agree that having the "text on the current line" being displayed with the search result will be helpful.

Since I'm new to hacking on helix I do not know how to extract the data from the buffer right now.

Once I know how to extract the data I can display it by modifying the contents of columns.

I will try to look into this and implement this.

Any help regarding which file/struct/variable contains the buffer text data so I can extract (clone) it will be helpful.

Thank you.

[nik-rev]

I agree that having the "text on the current line" being displayed with the search result will be helpful.

Since I'm new to hacking on helix I do not know how to extract the data from the buffer right now.

Once I know how to extract the data I can display it by modifying the contents of columns.

I will try to look into this and implement this.

Any help regarding which file/struct/variable contains the buffer text data so I can extract (clone) it will be helpful.

Thank you.

You have access to a cx: &mut Context. doc!(cx.editor) should give you the currently focused Document which has methods like Document::text() which returns a Rope data structure representing the text in the current file

[oxcrow]

Issue: I have not been able to extract data from cx.editor.text().

Since columns expects a fn and not a closure, I was not able to borrow/access cx.editor from inside columns.

I was able to pad the line number with whitespaces to align the results, and add a dummy / placeholder text (that needs to be fixed).

Now how can I access the text data within columns?

I'm thinking of storing the text data inside FileResult when we run the search.

This will most likely work.

#[derive(Debug)]
struct FileResult {                                            
    path: PathBuf, 
    line_num: usize,
    line_content: String, // << Just store the text data here ?
}          

[oxcrow]

Success: The line content can now be displayed.

The formatting looks awful so I will try to fix it by only showing the first N non-whitespace characters of the lines.

[oxcrow]

Success: The results are now formatted correctly.

[Axlefublr]

omg I've been wanting this for a while!!

[oxcrow]

Success: The bug has been fixed.

Now only the current open document buffer is searched.

The other document buffers are ignored.

Thus even though multiple buffers are open searching "x" only returns results from the current document buffer.

@nik-rev Is there anything else that needs to be added/modified?

Most things seem right to me. One thing that might need to be fixed, is how we format the results on smaller monitors. Since I allowed the max result length to be 80, the results might wrap around on smaller screens.

I added a variable for this in the code called local_search_result_line_length, which if necessary, could be exposed to the user so they can configure it.

However I do not know how.

Thanks
~oxcrow

[nik-rev]

Success: The bug has been fixed.

Now only the current open document buffer is searched.

The other document buffers are ignored.

Thus even though multiple buffers are open searching "x" only returns results from the current document buffer.

@nik-rev Is there anything else that needs to be added/modified?

Most things seem right to me. One thing that might need to be fixed, is how we format the results on smaller monitors. Since I allowed the max result length to be 80, the results might wrap around on smaller screens.

I added a variable for this in the code called local_search_result_line_length, which if necessary, could be exposed to the user so they can configure it.

However I do not know how.

Thanks ~oxcrow

Looks good to me so far. I'll use it in my own fork of Helix from now on to give it a proper test drive

@nik-rev Is there anything else that needs to be added/modified?

FYI I'm not a maintainer, so you'll need someone who is to review this too

[oxcrow]

@nik-rev The current local searcher uses grep but I'm not too happy with it.

It would be netter to use a fuzzy searcher so that it would be easier for users to find complex things easily.

Do you know how may I replace the grep searcher with a fuzzy searcher?

The fuzzy file finder seems to work in the way that I want to replicate here. So maybe I will start by looking in there first.

[ivanrg99]

I would love to see this merged.

@oxcrow you could take a look at nucleo, made by the helix team and already in the dependencies.

I can see it being used in the Picker

I will revisit this when I have some more free time. Would love to help in any way possible

[ivanrg99]

I got it working, but it now loads all the lines directly into the picker, instead of starting the picker empty. Would that be ok?

If that is the case, should I create a PR on your fork @oxcrow ? I'm just starting to contribute back to open source so I don't know the approach