wansmer / treesj

Neovim plugin for splitting/joining blocks of code

Wansmer

treesj's Issues

Fix tests that fail

Fix tests that fail

Certain tests for sql and dart fail, but in these cases the nodes are formatted correctly when the plugin is run manually.
TODO: find and fix the problem

Problematic tests:

treesj/tests/langs/dart/join_spec.lua

Line 43 in 3203aa5

-- {

treesj/tests/langs/dart/split_spec.lua

Line 43 in 3203aa5

-- {

treesj/tests/langs/sql/split_spec.lua

Line 25 in 3203aa5

-- {

Rust: split match arms into blocks

I have migrated from splitjoin.vim to this plugin. This plugin works great, but my favorite feature of splitjoin.vim was how it could split expression in match arms into a block. This feature is a little broken in splitjoin.vim, and I always planned to remake it using tree sitter before I encounter this plugin.

I was planning to implement this myself and submit a PR, but from what I looked, this plugin only has support for splitting nodes with clear delimiters. So I became unsure to how to implement it.

// It would allow spliting a "match_arm.value"
match x {
    _ => 12
}
// Into this
match x {
    _ => {
        12
    }
}

This could also be extended to apply to Rust closures (|| 1 into || { 1 }) and JavaScript arrow function (() => 1 into () => { return 1; }), for example.

If you could provide any directions, I can try implementing it.

Python: don't drop trailing comma from single-element tuple on join

By default, TreeSJ drops the trailing comma when joining tuples, so:

(
  1,
)

becomes (1) instead of (1, ). This is a problem in Python if the tuple has only a single member, as the two forms mean different things: a parenthesized number vs. a tuple.

This can be easy to miss if the join happens as part of a nested collapse. Would be great if TreeSJ could automatically preserve the comma in this particular case when dealing with Python.

Join adds unnecessary commas and spaces, in Python

Given a Python file such as

bar(zzzz, {"thing": "blah"})

Call :TSJToggle twice and it becomes

bar(
    zzzz,
    { "thing": "blash"},
)

and then

bar(zzzz, { "thing": "blash", },)

Note the extra around { and } which weren't there before, plus the extra two ,s within , },)

Python node import_from_statement fails for single item

Splitting the following line results in invalid syntax.

from re import search

from re import (,
search,
)

I am aware there is little use in splitting the line but maybe it is easy to fix.

The expected result would be

from re import (
search,
)

Add an option to add trailing comma during split

For Go language a trailing comma is required when items place each on the own line, for example

func a(b int, c int, d int)

func a(
 b int,
 c int,
 d int, // <-comma here is required
)

The same applied for arrays and dictionaries.

Can't find an option for this, can you point on it if I miss?

Can only split/join/toggle once

Can only split/join/toggle once.

The split/join/toggle actions works fine the first time, but when trying to do a split/join/toggle again then nothing happens.

Im on NVIM v0.9.1

Ive tested with a clean config and only treesj and treesitter.

Ruby: split-join bracketed blocks into do-end blocks

Ruby uses different delimiters depending on whether a block is inline or multiline. For example,

# inline block (joined)
some_list.each { |x| puts x }

# multiline block (split)
some_list.each do |x|
  puts x
end

It does not look like the ruby treesj preset supports the block (or do_block) node type as of right now. I would love to see this added as a feature and would be happy to create a PR.

From what I can tell - and correct me if I'm wrong - there are two limitations that are blocking this kind feature:

1. There appears to be a limitation on non_bracketed_node where the "split" and "join" node representations must use the same syntax. In this case, support for individual bracketed nodes would need to be supported

   -- boolean: Non-bracket nodes (e.g., with then end instead of {}|<>|[])
   -- NOTE: Must be same for both modes
   non_bracket_node = false,

2. Another limitation (which goes hand-in-hand with the first) is that there are no preset options for defining node replacement. I.e. do replaces { and end replaces }

This plugin has some great improvements over splitjoin.vim, hoping to see it continuously improve 🙂

Happy to hear your thoughts!

Feat: Rust support?

Hey, thanks for making this, it's really useful.

Do you think it would be possible to add support for Rust?

E.G. Struct definitions/initialization:

struct SomeStruct {x: i32, y: i32 }
// ---
struct SomeStruct {
    x: i32,
    y: i32,
}

let x = SomeStruct { x: 123, y: 234 };
// ---
let x = SomeStruct {
    x: 123,
    y: 234,
};

Or function definitions/calls:

fn f(x: i32, y: i32) -> i32 {}
// ---
fn f(
    x: i32,
    y: i32,
) -> i32 {}

f(123, 234);
// ---
f(
    123,
    234,
);

[feat] Add Golang support

I think the Go language can be a good candidate for adding to preset languages.

No indent in LUA files

Hello,

When splitting the following code

local dict = { 'a', 'b', 'c' }

the result is:

local dict = {
'a',
'b',
'c',
}

My question: Why are the dict elements not properly indented like it's done in e.g. .js documents? Is this intended behaviour? How can I change this behaviour? I would expect the code to be like:

local dict = {
    'a',
    'b',
    'c',
}

[feat] add ruby support

Great work! Just wanted to create this ticket so I can track the feature. Currently I'm using splitjoin.vim but I'd love to switch to a lua based plugin that supports treesitter

Typescript: Split/Join some arrow functions results in invalid code

👋, thanks for building this plugin!

I found an issue when splitting or joining arrow functions in Typescript. The plugin seems to have trouble handing blocks without a return statement in the function body.

I have the following function

export async function pretendDelay() {
  const delay = Math.floor(Math.random() * 1000); // Between 0 and 1 seconds
  return new Promise((resolve) => {
    setTimeout(resolve, delay);
  });
}

Note that the setTimeout doesn't have a return.

• When I try to join the arrow function inside the Promise, the plugin tries to return the value of the block (changing the semantics), and places an extra semicolon in the end (adding a syntax error).
• If I try to revert back splitting the function, it now adds a return statement, and keeps the invalid semicolon.

splitjoin.mov

Fallback to `mini.splitjoin`

It would make sense to fall back onto mini.splitjoin if TreeSitter isn't configured for a buffer or split operation fails.

For example, trying to split a long string in Bash returns:

[TreeSJ]: Node "string_content" for lang "bash" is not configured

...while the same line can be perfectly split by mini.splitjoin.

[BUG] Incorrect behaviour for toggling Ruby blocks

Example:

CleanShot.2023-06-02.at.14.36.02.mp4

For construction

    it { is_expected.to validate_presence_of(:email) }

While calling require('treesj').join() inside block it will not convert do/end to { and }. Everything works ok if the cursor outside, or directly on do or end

Adds semicolon at the end of function declarations in Rust

Setup:

struct Test;

impl Test {
    fn test() { }
    fn test2() { }
}

What it actually does:

impl Test { fn test() { }; fn test2() { }; }

What is expected(and works):

impl Test { fn test() { } fn test2() { } }

The semicolons should not be added

Bug: Splitting an array adds a trailing comma, even in languages where that is not valid

Adding a trailing comma is actually a nice touch, but it's also applied in languages where a dangling comma is invalid, like js.

List the default keymaps in the readme

I can see a config called use_default_keymaps but there is no information about what the default keymaps are.

Error when Running Lua Config from README

Ran into an issue when using the Lua config example from the README. Got this error:

attempt to index global 'lu' (a nil value)

which points to this line:

  langs = lu._prepare_presets(langs.presets),

Ruby module/namespace joins

Ruby has some syntactic sugar with namespaces. In particular modules (and classes) can be nested within modules, like so:

module MyModule
  class MyClass
  end
end

# is the same as
class MyModule::MyClass
end

Some examples taken from splitjoin.vim would be:

# RESULT OF JOIN
class Foo::Bar::Baz < Quux
  def initialize
    # foo
  end
end

# RESULT OF SPLIT
module Foo
  module Bar
    class Baz < Quux
      def initialize
        # foo
      end
    end
  end
end

# RESULT OF JOIN
class Foo::Bar::Baz < Quux
end

# RESULT OF SPLIT
module Foo
  module Bar
    class Baz < Quux
    end
  end
end


# RESULT OF JOIN
module A::B
end

# RESULT OF SPLIT
module A
  module B
  end
end


# EXAMPLE WITH MIXED 
module Foo::Bar
  class Baz::Qux
  end
end

# RESULT OF JOIN
class Foo::Bar::Baz::Qux
end

This one may be tricky to implement given that they may join into a class or a module (depending on whether the nested object is one or the other)

module 'treesj' not found

Hey, how do you do guys?

I'm trying to use the package in the following way:

use({
  'Wansmer/treesj',
  requires = { 'nvim-treesitter' },
  commit = '7be9ebd',
  config = function()
    require('treesj').setup({
      notify = true,
      dot_repeat = true,
      use_default_keymaps = true,
      check_syntax_error = true,
      max_join_length = 120,
      cursor_behavior = 'hold',
    })
  end,
})

But every time I start Neovim, I encounter this error:

packer.nvim: Error running setup for treesj: [string "..."]:0: module 'treesj' not found:
^Ino field package.preload['treesj']
^Ino file './treesj.lua'
^Ino file '/__w/neovim/neovim/.deps/usr/share/luajit-2.1.0-beta3/treesj.lua'
^Ino file '/usr/local/share/lua/5.1/treesj.lua'
^Ino file '/usr/local/share/lua/5.1/treesj/init.lua'
^Ino file '/__w/neovim/neovim/.deps/usr/share/lua/5.1/treesj.lua'
^Ino file '/__w/neovim/neovim/.deps/usr/share/lua/5.1/treesj/init.lua'
^Ino file '/home/adrianmiranda/.cache/nvim/packer_hererocks/2.1.0-beta3/share/lua/5.1/treesj.lua'
^Ino file '/home/adrianmiranda/.cache/nvim/packer_hererocks/2.1.0-beta3/share/lua/5.1/treesj/init.lua'
^Ino file '/home/adrianmiranda/.cache/nvim/packer_hererocks/2.1.0-beta3/lib/luarocks/rocks-5.1/treesj.lua'
^Ino file '/home/adrianmiranda/.cache/nvim/packer_hererocks/2.1.0-beta3/lib/luarocks/rocks-5.1/treesj/init.lua'
^Ino file './treesj.so'
^Ino file '/usr/local/lib/lua/5.1/treesj.so'
^Ino file '/__w/neovim/neovim/.deps/usr/lib/lua/5.1/treesj.so'
^Ino file '/usr/local/lib/lua/5.1/loadall.so'
^Ino file '/home/adrianmiranda/.cache/nvim/packer_hererocks/2.1.0-beta3/lib/lua/5.1/treesj.so'

Is there anything wrong with my configuration?

if I want use treesj in javascript need TSInstall javascript ?

if I want use treesj in javascript need TSInstall javascript ?

Allow ignoring the max join length for single calls

Hi, I'm trying to add a forced join keymap to my config, but apparently passing a preset like this to the toggle and join methods is ignored: { max_join_length = 999 }.

I would love if I could disable the max join length feature on demand. Ideally this would be done via a separate boolean setting or setting max_join_length = 0.

Split classnames on new lines

When working with React, it would be amazing if it split each class on a new line. At the moment it only splits the className property itself.

Thank you for this great plugin!

`noexpandtab` is not respected when splitting

to reproduce

-- before
if 1 == 1 then print("beep") end

run this three commands (listchars only changed for visibiliy)

:set noexpandtab
:lua vim.opt.listchars = "tab:> ,space:."
:TSJSplit

-- after
if 1 == 1 then
    print("beep") -- has four spaces, not a tab character
end

[TreeSJ]: Cannot join "table_constructor" containing node from one of this: { "commend" }

Please help. I am not sure what is wrong in my configuration.

My config is

require("lazy").setup({
{
    'Wansmer/treesj',
    keys = { '<space>m', '<space>j', '<space>s' },
    dependencies = { 'nvim-treesitter/nvim-treesitter' },
    config = function()
      require('treesj').setup({ --[[ your config ]] })
    end,
  },
})

The above gets lazy loaded till I press one of `m', 'j' or 's'.

After that when I press :TSJToggle, it is giving me the below error

[TreeSJ]: Cannot join "table_constructor" containing node from one of this: { "commend" }

Paranthesized expression support for Rust

fn main() {
    assert!(true);
           ^
}

When the cursor is at ^, performing require("treesj").split() will cause the following error:

"parenthesized_expression" for lang "rust" is not configured

Can we have built-in support for this?

Cant disable the preset keybindings.

I am new to reporting an issue, so please excuse me if I don't do it right.

I made some adjustments to the keybindings for my usage, but now the default won't turn off, I have tryed changing the default settings, but they persist.

Here is how I am trying to modify them:

---

From my (lazy) ~/.config/nvim/lua/pluginslist.lua

    {
        'Wansmer/treesj',
        -- keys = { '<space>m', '<space>j', '<space>s' },
        dependencies = { 'nvim-treesitter/nvim-treesitter' },
        config = function()
            require('treesj').setup({--[[ your config ]]})
        end,
    },

---

From my ~/.config/nvim/after/plugin/treesj.lua

-- Load TreesJ modules
local tsj = require('treesj')

-- local langs = {--[[ configuration for languages ]]}

tsj.setup({
  ---@type boolean Use default keymaps (<space>m - toggle, <space>j - join, <space>s - split)
  use_default_keymaps = false,
  ---@type boolean Node with syntax error will not be formatted
  check_syntax_error = true,
  ---If line after join will be longer than max value,
  ---@type number If line after join will be longer than max value, node will not be formatted
  max_join_length = 120,
  ---Cursor behavior:
  ---hold - cursor follows the node/place on which it was called
  ---start - cursor jumps to the first symbol of the node being formatted
  ---end - cursor jumps to the last symbol of the node being formatted
  ---@type 'hold'|'start'|'end'
  cursor_behavior = 'hold',
  ---@type boolean Notify about possible problems or not
  notify = true,
  ---@type table Presets for languages
  -- langs = lu._prepare_presets(langs.presets),
  ---@type boolean Use `dot` for repeat action
  dot_repeat = true,
  ---@type nil|function Callback for treesj error handler. func (err_text, level, ...other_text)
  on_error = nil,
})

-- Load which-key for keybinding docs
local wk = require("which-key")

-- Register TreesJ keybindings
wk.register({
    ["<leader>tst"] = { function() require('treesj').toggle() end, "Toggle node under cursor" },
    ["<leader>tss"] = { function() require('treesj').split() end, "Split node under cursor" },
    ["<leader>tsj"] = { function() require('treesj').join() end, "Join node under cursor" },
})

kotlin: split on parameters in a function_declaration

Thank you for the great plugin! :)

I have been able to configure language support for some kotlin language constructs but I have trouble with function declarations:
Its treesitter ast does not group function parameters in a common parent node like other languages do.

For example the function
fun test(a: String, b: String){ }

has the tree

function_declaration [0, 0] - [1, 1]
  simple_identifier [0, 4] - [0, 8]
  parameter [0, 9] - [0, 18]
    simple_identifier [0, 9] - [0, 10]
    user_type [0, 12] - [0, 18]
      type_identifier [0, 12] - [0, 18]
  parameter [0, 20] - [0, 29]
    simple_identifier [0, 20] - [0, 21]
    user_type [0, 23] - [0, 29]
      type_identifier [0, 23] - [0, 29]
  function_body [0, 30] - [1, 1]

Is it still possible to configure treesj for this case?

Lag on space keypress even when default keybindings are not in use

I use the space key to enter the command mode with nmap <space> : in my vim config

So I've disabled treesj's default mappings with use_default_keymaps = false, in favor of my custom mappings but there is still a noticeable delay of half a second to a second while using the spacebar.

Bug: Joining arrays in YAML adds invalid `-`

given this array (with | as cursor position)

no-empty:
  - wa|rn
  - allowEmptyCatch: true

joining will result in this:

  no-empty: [ warn, - { allowEmptyCatch: true } ]

which is invalid yaml, as it should be this:

  no-empty: [ warn, { allowEmptyCatch: true } ]

[Feature Request] Also handle joining disjoint nodes like comment lines and concatenated strings?

I moved from Intellij (IdeaVim) to NeoVim, but for some reason, I wasn't able to find a plugin that behaves like smartjoin in IdeaVim, which added features to the original join(J) like making single line methods in Kotlin(fun foo() = //...), combining multi-line concat strings("ABC" + <newline> "Next Line" -> "ABCNext Line"), folding multiple ifs, and folding lambdas into a single line({ doStuff() }). I see that this project seems to already have the groundwork down for something similar, so I thought it wouldn't be that hard to make this. Since the project is about joining and splitting, the reverse of the actions above doesn't seem too hard either. Though I do see that it's a bit different cause you are looking into neighboring nodes instead of only up the tree...

For reference: IdeaJoin(smartjoin) examples

`dot_repeat` is not working for me

here's my config for lazy.nvim:

local opts = { noremap = true, silent = true, expr = false }

return {
  "Wansmer/treesj",
  dependencies = { "nvim-treesitter/nvim-treesitter" },
  opts = { use_default_keymaps = false, max_join_length = 150, dot_repeat = true },
  cmd = { "TSJToggle", "TSJJoin", "TSJSplit" },
  keys = {
    { "gM", "<cmd>TSJToggle<cr>", opts },
    { "gJ", "<cmd>TSJJoin<cr>",   opts },
    { "gS", "<cmd>TSJSplit<cr>",  opts },
  },
}

Allow fallback when no node is found

I've looked through the readme, and it looks like right now it's only possible to add a fallback that will be used for buffers where treesitter is not present. However, I'd like to also have a fallback in cases where I do have treesitter, but no node is found.

The main use case for this would be literal strings that can't be parsed by tresitter, but still contain code that can be split/joined.

I'd be up for implementing this as well, what do you think about this feature?

[feat] add C/C++ support

hey! amazing work on this plugin, works very well for me in some of the languages I have tested it on!

I wanted to request support for C/C++ as well, as it would be very useful and the fact that there are very few options (if any) for these languages for this level of functionality. I am aware that it might be a complicated endeavor (as suggested by discussions in splitjoin.vim plugin), but its good to start a new discussion.

I think it would be a great addition to this plugin, and will attract many more devs and appreciation!

Splitting not properly formatted for longhand PHP array syntax

First of all, thank you for this awesome plugin!

PHP supports creating arrays with both array() and `[]. For example, this is a common array structure in PHP:

array( 'one', 'two' )

When splitting, this is what I would expect to happen:

array(
  'one',
  'two'
)

However, for some reason the array notation gets split up and this happens instead:

array
  (
  'one',
  'two'
)

Any idea how I might be able to fix this in my config? Thanks!

feat(python): split python import list

Split code like:

from re import search, match, sub...

Thanks a lot for the plugin.

FR: Support for functions and if-statements

Thanks for this plugin, I really like the idea of auto-detection :)

My feature suggestions would be to extend the plugin functionality to also work for functions and if-statements, as I often find myself splitting/joining them in these situations.

Below are two examples in lua.

function foo () return "bar" end

function foo ()
  return "bar"
end

if foo then return end

if foo then
  return
end

repeat last action using `.` support

Thanks for this awesome plugin.

I think it would be great if the repeat last action using dot(.) can be implemented.

Join adds unnecessary space for empty Lua table

When I use <space>j in Lua to join something like this:

local myvar = {
    a = 'something',
    b = {}
}

It becomes:

local myvar = { a = 'something', b = { } }

Notice the extra space in the empty dict. I would like if the space was not there, like this:

local myvar = { a = 'something', b = {} }

Join/Split a chain of method calls

Would it be possible split/join a chain of method calls?

Ruby example

# from:
all_names = some_object.associations.map{ |o| o.name }.join(',').downcase

# to:
all_names = some_object
  .associations
  .map{ |o| o.name }
  .join(',')
  .downcase

Note: I just installed this plugin and it works amazing, and so fast! Thanks for all the work <3

[bug] empty function calls get split as well

local function test()
    print('hello')
end

test()
    ^^ call TSJSplit with cursor on/within parens

--- becomes

test(
)

Using default config, in Lua file.
can a check be added so that splits dont happen when no elements are present?

Please help to configure Pug

Hi, thank you for this awesome tool.
I'm struggling to configure it for pug.
I've looked into the html implementation but did not find anything useful. I always get the message:

Cannot "join" node longer than 120 symbols. Check your settings to change it.

Here is my current pug config:

local langs = require('treesj.langs')
local u = require('treesj.langs.utils')
langs["pug"] = {
    attributes = u.set_default_preset({
        both = {
            space_separator = 1,
            separator = ' '
        },
    }),
}

Error ：search first

Why does it search first, is there something wrong with my configuration?

    branch main
    commit 950d06e

   Error  20:41:07 msg_show.lua_error Error executing Lua callback: /Users/cj/.local/share/nvim/lazy/treesj/lua/treesj/init.lua:25: Vim:E486: Pattern not found: rstnersitnrsetnei
stack traceback:
	[C]: in function 'normal'
	/Users/cj/.local/share/nvim/lazy/treesj/lua/treesj/init.lua:25: in function 'set_opfunc_and_format'
	/Users/cj/.local/share/nvim/lazy/treesj/lua/treesj/init.lua:30: in function 'format'
	/Users/cj/.local/share/nvim/lazy/treesj/lua/treesj/init.lua:37: in function </Users/cj/.local/share/nvim/lazy/treesj/lua/treesj/init.lua:36>

J.mov

Node "use_list" for lang "rust" is not configured

basically title says it all

Splitting attributes in TSX slightly weird

I'm trying to split attributes in TSX with:

local tsj = require('treesj')
local u = require('treesj.langs.utils')
local typescript = require('treesj.langs.typescript')

tsj.setup({
  use_default_keymaps = false,
  max_join_length = 9999,
  langs = {
    tsx = u.merge_preset(typescript, {
      jsx_opening_element = u.set_default_preset({
        both = {
          omit = { 'identifier' },
        },
      }),
      jsx_self_closing_element = u.set_default_preset({
        both = {
          omit = { 'identifier' },
          no_format_with = {},
        },
      }),
    })
  },
})
    

This is the results I've got:

<PageCanvas
  class="px-2 py-1"
  canvasClass={clsx("page-thumbnail-canvas mx-auto shadow", props.active && 'ring-4')}
  page={props.page}
  zoom={() => 0.28}
  onUpdate={props.onUpdate}
  /
>

Notice the character / and > at the end also got split? How to fix this?
Thank you.

[Feature Request]: Toggle between <a></a> and <a /> (if tag is empty) in HTML/JSX etc.

Does it make sense to have this feature? Emmet has it, it's called https://docs.emmet.io/actions/split-join-tag/

[bug] cursor out of window error

|| Error executing Lua callback: ...e/nvim/site/pack/packer/opt/treesj/lua/treesj/format.lua:67: Cursor position outside buffer
|| stack traceback:
|| 	[C]: in function 'nvim_win_set_cursor'
|| 	...e/nvim/site/pack/packer/opt/treesj/lua/treesj/format.lua:67: in function '_format'
|| 	...are/nvim/site/pack/packer/opt/treesj/lua/treesj/init.lua:20: in function <...are/nvim/site/pack/packer/opt/treesj/lua/treesj/init.lua:19>

I am woking on adding support for C++, and encountered the above trace with the config:

treesj.setup({
    langs = {
        cpp = {
            compound_statement = treesj_utils.set_preset_for_list({
                both = {
                    separator = ';',
                    last_separator = true,
                    no_format_with = {}
                }
            })
        },
    },
})

in the code segment:

  if (true) {
     ^^^^^ cursor anywhere before `{`
    int a = 10;
    int b = 15;
  }

using the TSJSplit. The entire document is messed up after this crash as well.
TSJJoin in the same case, gives the error: Cannot join node longer .....

otherwise, it works as expected:

  if (true) {
              ^ cursor at or after `{`
    int a = 10;
    int b = 15;
  }

---

  if (true) { int a = 10; int b = 15; }

For reference, the entire block { .... } is the TS node compound_statement.

NOTE:
I know this use case does not make sense because why would you put different statements on the same line, but its just something I encountered while playing around. This is a bug though, and might be encountered elsewhere, because this config is 'correct'.

Don't use backslash in Python

Hi! How configure plugin for never using backslash in Python(imports, etc)?
For example:

from airflow.providers.apache.spark.operators.spark_submit import \
    SparkSubmitOperator

Prefer:

from airflow.providers.apache.spark.operators.spark_submit import (
    SparkSubmitOperator
)

FR: when on comment, just run `gqq`

Minor idea:

Right now, just gives out a warning and does nothing when on a comment. How about having treesj simply split the comment by textwidth instead? Essentially use gqq as behavior when trying to splitjoin a comment, since it's a straightforward and common split-join operation one does on comments.

Could add this feature as opt-in, since not everyone would like it, I guess