In what scenario does this happen?

There's existing logic to account for concealed text from treesitter highlights.

If this is virtual text added by other plugins then that's something I don't want to attempt to support. There's too many potential edge cases to open up this can of worms, including what about all the other places plugins could add virtual text, not just table rows. I can provide some kind of hook to let users modify the width via a function parameter, but not some generic way to handle virtual text from any source.

Issues with the logic as written:

• hl_group == 'Conceal' doesn't cause the text to be concealed, it's just a highlight group
• #conceal_highlight[4].virt_text[1] is not measuring text width, there are always 2 entries for each section of virtual text, should iterate over them and add the width of the first element
• does not account for different conceal level values changing width

If this is virtual text added by other plugins then that's something I don't want to attempt to support.

Ah, apologies I did not know that. That was my main intent when writing the patch 😅

No worries at all, it's not obvious why I would prefer to avoid that behavior, and I appreciate you going through the effort of putting up a PR.

I did add a cell_offset configuration option that you can use in your config to make these adjustments: e5c3c50

Porting your logic to it would be something like:

require('render-markdown').setup({
    pipe_table = {
        cell_offset = function(ctx)
            local offset = 0
            local row, start_col, _, end_col = ctx.node:range()
            local highlights = vim.api.nvim_buf_get_extmarks(
                0,
                -1,
                { row, start_col },
                { row, end_col },
                { details = true, type = 'highlight' }
            )
            for _, conceal_highlight in ipairs(highlights) do
                if conceal_highlight[4].hl_group == 'Conceal' then
                    offset = offfset - (conceal_highlight[4].end_col - conceal_highlight[3])
                    if conceal_highlight[4].virt_text then
                        offset = offset + #conceal_highlight[4].virt_text[1]
                    end
                end
            end
            return offset
        end,
    },
})

Thanks!!