fix: fix ts_range_to_region for nodes that end with newlines

[TheLeoP]

• I have read CONTRIBUTING.md
• I have read CODE_OF_CONDUCT.md

Fixes the issue mentioned in #2010 (comment)

Instead of only checking this for the last line of the buffer (which previously would incorrectly exclude some regions), this PR modifies all the ranges with an end that's row-exclusive, col-0 to be row-inclusive, col-exclusive (just like for nodes that don't end with a newline).

For reference, from #2010 (comment)

I asked in the "Neovim treesitter" matrix channel and clason explained to me that, if a node contains a trailing newline, it's end_ will change from row-inclusive, col-exclusive to row-exclusive, col-0.

[echasnovski]

Thanks for the PR!

If it fixes the issue, then it looks good after suggested changes (propagated to both modules).

Also, does it have to be a check for col == 0? Or can it be only for the case when line > vim.api.nvim_buf_line_count(0) and clamp to the last line plus last column position? Probably, yes, accounting for all col == 0 is better. As per the spec, region should be end-inclusive and col starts from 1. The fact that search currently works for cases like { to = { line = 3, col = 0 } } looks more like a coincidence to me (there are no tests and documentation for it).

I briefly thought about fixing this more globally by adjusting how best region/span is searched with the following diff:

@@ -1894,7 +1894,8 @@ H.get_neighborhood = function(reference_region, n_neighbors)
   -- Compute '2d neighborhood' of (possibly empty) region
   local from_line, to_line = reference_region.from.line, (reference_region.to or reference_region.from).line
   local line_start = math.max(1, from_line - n_neighbors)
-  local line_end = math.min(vim.api.nvim_buf_line_count(0), to_line + n_neighbors)
+  local max_lines = vim.api.nvim_buf_line_count(0) + 1
+  local line_end = math.min(max_lines, to_line + n_neighbors)
   local neigh2d = vim.api.nvim_buf_get_lines(0, line_start - 1, line_end, false)
   -- Append 'newline' character to distinguish between lines in 1d case
   for k, v in pairs(neigh2d) do

However, as the case of col == 0 is not explicitly supported, fixing this in specific textobject indeed looks better. It might change in the future to allow col = 0 to mean "the whole previous line" (that's why I am posting it here now to not forget), but probably unlikely.

[echasnovski]

Had time, so decided to go ahead and make the changes outside of this PR. Hope this is okay.