Handling cases when cursor is not at the end of line

[simonmandlik]

Is your feature request related to a problem? Please describe.
When working on #35, I struggled a lot when dealing with cases when the cursor is not at the very end of the line.

Pseudocode examples:

function f(<cursor>)

leads to

function f(
    <cursor>
end)

or for example

if f(<cursor>)

leads to

if f(
    <cursor>
)

I guess this behavior applies to multiple languages, not only julia. For example, try the following test for lua:

test "lua, function", <<~END
-function foo(█)
+function foo(
+  )
END

Maybe I'm missing something, but writing this in TS queries seems quite difficult — I have to always capture the last ")", which leads to many corner cases and queries become large.

Describe the solution you'd like

I have only written some queries and I don't know anything about internals of this plugin. What is the preferred way to deal with that?

What if the following heuristic was used: Whenever on the current line there still is a non-whitespace character after the cursor, do not add an "end"? Is there any case where this would be undesirable?

Thanks!

[RRethy]

What if the following heuristic was used: Whenever on the current line there still is a non-whitespace character after the cursor, do not add an "end"? Is there any case where this would be undesirable?

Yes it would be undesirable. I don't want to not add an end when it's needed. We have the heuristic that if the text that follows the cursor is alphanumeric, then put it on the line of the cursor, if it's non-alphanumeric then put it on the line of the end. It's especially useful to handling functions defined as arguments.

The code for this is at https://github.com/RRethy/nvim-treesitter-endwise/blob/master/lua/nvim-treesitter/endwise.lua#L81.

This test shows what happens, https://github.com/RRethy/nvim-treesitter-endwise/blob/master/tests/endwise/lua.rb#L185-L189.

[RRethy]

For what it's worth, there is no perfect solution for this since we don't get a full syntax tree, we have to interpret a broken syntax tree, these are tradeoffs I made a long time ago and honestly aren't going to change.

[simonmandlik]

We have the heuristic that if the text that follows the cursor is alphanumeric, then put it on the line of the cursor, if it's non-alphanumeric then put it on the line of the end. It's especially useful to handling functions defined as arguments.

Yes, I have noticed. It works great in cases like

function f(a, <cursor>b)

but not when it is in front of "," or ")".

Sow how about adding these some punctuation mark characters to the regexp here https://github.com/RRethy/nvim-treesitter-endwise/blob/master/lua/nvim-treesitter/endwise.lua#L81?