feat(table): add cell wrapping using virt lines by MaxDillon · Pull Request #617 · MeanderingProgrammer/render-markdown.nvim

MaxDillon · 2026-03-03T22:51:39Z

Closes: #616

What it does

Adds max_table_width for controlling wrapping behavior for tables. Wrap table when exceeds width setting
Updates offset to carry virt_text metadata (necesary for reconstructing hl groups inside tables)
compute_layout in Renderer:setup() calculates layout (column widths, row heights)
Covers original buffer line with overlay to prevent text overflow sticking out to right of table
Adds virtual lines below to display wrapped text
Re-render on resize or wrap toggle

MeanderingProgrammer

This is a really interesting PR, it's going to take me a bit of time to parse through it.

I appreciate the effort! This could be a truly awesome feature.

MeanderingProgrammer · 2026-03-03T23:40:41Z

lua/render-markdown/render/markdown/table.lua

+    end
+
+    -- Table already fits; use existing renderer
+    if total_natural <= text_budget then


I haven't dug in much here but this seems to not handle the case where concealing results in the rows fitting within the available width. At least in the specific case of this repos README:

| Screenshot | Video | | ---------- | --------- | | ![Heading](https://github.com/user-attachments/assets/40655575-b091-4ab8-b830-38f8004d7746) | ![Heading](https://github.com/user-attachments/assets/03f629ea-f6da-4f05-a035-827fd944e8be) | | ![Table](https://github.com/user-attachments/assets/7d021918-e89c-4b7d-b33a-869390f9a826) | ![Table](https://github.com/user-attachments/assets/fdbcfbfa-5f9e-49b7-8c19-f7e837979a7a) | | ![Quote](https://github.com/user-attachments/assets/822ae62c-bc0f-40a7-b8bb-fb3a885a95f9) | ![Quote](https://github.com/user-attachments/assets/aa002ac7-b30f-4079-bba9-505160a4ad78) | | ![Callout](https://github.com/user-attachments/assets/e468a463-bc8d-420c-bb4c-da1263795092) | ![Callout](https://github.com/user-attachments/assets/d56cc5c7-43cd-4ce7-ad33-6164c2e23875) | | ![Latex](https://github.com/user-attachments/assets/68f27ff3-49c8-42b5-bb7a-3b89c1e98401) | ![Latex](https://github.com/user-attachments/assets/41e657a6-bcc2-464d-ab8c-a23bfcb80b0f) |

When I reduce the window size it renders as:

It gets caught on this line as it thinks the width is enough to hold the content. But that would need to based on the unconcealed width rather than the visible width.

I think I figured out what is going wrong here and it is a bug. Try this to fix it:

Go into any one of the link names (Heading, Table, Quote, Callout, or Latex) and write a bunch more text in there

100iLatex<esc>

This should cause that cell to have enough rendered text to enable wrapping, and that should resolve the issue.

When we are in wrapping mode, the solution mentioned in #616 works for wrapping even with concealed text, but when there is not enough text, I am falling back on the old wrapping strategy. In this scenario, where the un-rendered content is wrapping but the table hasn't entered into wrapping mode, we get this issue.

For now, test with at least one cell having enough rendered content to wrap, and I can either
A) Always use the wrapping renderer regardless of whether any text is getting wrapped

or

B) Include a check to enable wrap mode if any unrendered row content is wrapping

Should be fixed by fb51061

Just disabling early exit here. We will still default to nowrap if applicable, it will just use the full range of checks now. Try your old experiment with the README

MeanderingProgrammer · 2026-03-03T23:41:04Z

lua/render-markdown/settings.lua

+    -- | 0.1–1.0 | fraction of window width, e.g. 0.8 = 80%          |
+    -- | 2+      | absolute character width, e.g. 80 = 80 columns     |
+    -- | < 0     | window width minus N, e.g. -10 = width minus 10    |
+    max_table_width = 0,


One minor thing is if this does get merged in we'd definitely enable it by default, probably set the default to 1.0. Unless we feel it's unstable and want to keep stress testing it before switching it on for everyone.

…t is too long

MeanderingProgrammer · 2026-03-04T02:16:11Z

lua/render-markdown/core/manager.lua

        end,
    })
+    -- terminal / GUI resize — re-render all visible attached windows
+    vim.api.nvim_create_autocmd('VimResized', {


Is this not covered by the WinResized autocmd?

MeanderingProgrammer · 2026-03-05T00:53:07Z

lua/render-markdown/render/markdown/table.lua

+        end
+    end
+
+    local total_natural = 0


MeanderingProgrammer · 2026-03-05T01:45:19Z

lua/render-markdown/render/markdown/table.lua

+    })
+
+    -- Lines 1..height-1: overlay buffer wrap continuations, then virt_lines.
+    if height > 1 then


remove, if height <= 1 then the for vl = 1, height - 1 do will be skipped all on its own since the value will be < 1.

MeanderingProgrammer · 2026-03-05T01:52:10Z

lua/render-markdown/render/markdown/table.lua

+    -- uses the capped widths (padding is included in delim col width).
+    if self.layout.wrap then
+        for i, w in ipairs(self.layout.col_widths) do
+            self.data.delim.cols[i].width = w + 2 * self.config.padding


This implementation (along with the way the line is built) will always add additional padding to cells that do not need it. The point of padding is to make sure that each cell has space around it, i.e. |A| -> | A |.

However if the cell already has padding like | A | it can be kept as is. With this logic it gets changed to | A | (with 2 extra spaces).

MeanderingProgrammer · 2026-03-05T23:26:13Z

lua/render-markdown/render/markdown/table.lua

+    end
+
+    local filler = self.config.filler
+    local function build_line(visual_line)


This currently doesn't use the alignment signal, not strictly necessary from the start, but we probably wouldn't enable it by default without that. We would probably only try to align cells that are not wrapped.

MeanderingProgrammer · 2026-03-05T23:28:54Z

lua/render-markdown/render/markdown/table.lua

+---@param end_col integer
+---@return render.md.mark.Line
+function Render:cell_segments(row, start_col, end_col)
+    local raw_full = vim.api.nvim_buf_get_text(


I've updated the parsed columns so now they store the node associated with them: bd482f9

So you can pass that along to this method and do col.node.text. You'll need to rebase against main, will probably have a few merge conflicts to resolve.

MeanderingProgrammer · 2026-03-05T23:31:23Z

lua/render-markdown/render/markdown/table.lua

+        local buf_line = vim.api.nvim_buf_get_lines(
+            self.context.buf, row.node.start_row, row.node.start_row + 1, false
+        )[1] or ''
+        local raw_screen_lines = math.max(1, math.ceil(str.width(buf_line) / win_width))


local _, line = row.node:line('first', 0) local raw_screen_lines = math.ceil(str.width(line) / win_width)

MeanderingProgrammer · 2026-03-05T23:33:30Z

lua/render-markdown/render/markdown/table.lua

+    local buf_line = vim.api.nvim_buf_get_lines(
+        self.context.buf, row.node.start_row, row.node.start_row + 1, false
+    )[1] or ''


local _, buf_line = row.node:line('first', 0) buf_line = buf_line or ''

MeanderingProgrammer · 2026-03-05T23:36:27Z

lua/render-markdown/render/markdown/table.lua


+---@private
+---@return render.md.table.Layout
+function Render:compute_layout()


Currently does not handle tables with leading spaces, i.e.

| Col | Col | Col | | --- | --- | --- | | Row | Row | Row |

feat(table): add cell wrapping using virt lines

f065c2b

MeanderingProgrammer reviewed Mar 3, 2026

View reviewed changes

fix(table): prevent early return; enable wrapping when unrendered tex…

fb51061

…t is too long

MaxDillon force-pushed the feat/table-cell-wrapping branch from fa67526 to fb51061 Compare March 4, 2026 00:48

MeanderingProgrammer requested changes Mar 5, 2026

View reviewed changes

Conversation

MaxDillon commented Mar 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What it does

Uh oh!

MeanderingProgrammer left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

MaxDillon commented Mar 3, 2026 •

edited

Loading