On help pages, "<" not concealed when followed by " - "

[dstein64]

Problem

On help pages with code, "<" is not concealed when followed by " - ".

This started at the following Neovim commit: 095bd8d.

Steps to reproduce

nvim --clean

Create a file with the following content:

text >
 let x = 0
< - text

Then run:

:set ft=help

Expected behavior

Expected (no "<"):

text
 let x = 0
 - text

Actual (includes "<"):

text
 let x = 0
< - text

(the problem is gone if "- text" is changed to "-text" or "text")

Nvim version (nvim -v)

v0.11.0

Vim (not Nvim) behaves the same?

No

Operating system/version

Kubuntu 24.04

Terminal name/version

Konsole 23.08.5

$TERM environment variable

xterm-256color

Installation

appimage

[clason]

Why are you expecting this to work?

[dstein64]

"Why are you expecting this to work?"

The example matches the format described under :help help-codeblock. There's nothing in the documentation that suggests that "<" shouldn't be followed by " - ".

[clason]

So you're reporting an issue with the vimdoc parser. (Which is wontfix. We are more strict with syntax.)

[dstein64]

"Which is wontfix. We are more strict with syntax."

In that case, helphelp.txt (which has the aforementioned help-codeblock) wouldn't be the only help document requiring updates.

The help pages ft_sql.txt and insert.txt are currently depending on "<" ending a code block even when followed by " - " (although for insert.txt, "<" is followed by "-" without a leading space, which is presumably unintentional).

• https://github.com/neovim/neovim/blob/2c960e8f047cf78b34e581990846fce08b62e586/runtime/doc/ft_sql.txt#L680
  • There are other examples throughout that file.
• https://github.com/neovim/neovim/blob/2c960e8f047cf78b34e581990846fce08b62e586/runtime/doc/insert.txt#L1583

[clason]

Again, these are upstream docs which we don't tend to touch without very good reason.

[dstein64]

This is related to (or could be the same issue as) #118.

[clason]

Yes, that is related. The parser relies on whitespace for tokenization (since it doesn't use an external scanner), and lists are difficult to detect if you can't rely on isolation.

Instead, the policy so far has been "just write better docs smh".

[justinmk]

Maybe it's feasible for the grammar to let \s*- terminate a codeblock.