markdownlint takes absurdly long to parse a 27,000 line table due to work for each inline incorrectly processing the entire table

STR:

1. Install both Text Power Tools (TPT) and markdownlint.
2. Open a long MD document. I was viewing this one.
3. Ctrl+P, and choose any of the TPT commands. I use “filter lines including a string”

*Note: I see this bug even without opening an MD file. But a long MD file is what makes it easier to reproduce.

Expected a popup from TPT to ask you to input query

Observed It shows “activating extensions…” at status bar for a while, then nothing happens. No popup. If I click “activating extensions” during the time, it shows mardownlint is “unresponsive”.

Afterwards, markdownlint freezes (no longer recognize lighting errors) until restarting.

Tested on newest stable VS Code and Insiders

Version: 1.62.2 (system setup)
Commit: 3a6960b964327f0e3882ce18fcebd07ed191b316
Date: 2021-11-11T20:56:38.428Z
Electron: 13.5.2
Chrome: 91.0.4472.164
Node.js: 14.16.0
V8: 9.1.269.39-electron.0
OS: Windows_NT x64 6.1.7601
----------------
Version: 1.63.0-insider
Commit: 4e303fcc52269ab578affa6cb1eb455eacab445f
Date: 2021-11-22T05:28:15.683Z
Electron: 13.5.2
Chromium: 91.0.4472.164
Node.js: 14.16.0
V8: 9.1.269.39-electron.0
OS: Windows_NT x64 6.1.7601

See also: https://github.com/qcz/vscode-text-power-tools/issues/29

[DavidAnson]

Thanks! Your trace shows an unreasonable amount of time being spent in markdownlint’s helpers.emphasisMarkersInContent function which is used exclusively by MD037/no-space-in-emphasis. It doesn’t seem to be stuck, but rather spending all its time in the “search code spans” phase splitting code with newLineRe and running emphasisMarkersRe.

Interestingly, if I weren’t looking at your issue, I was going to be looking at this function anyway because it relates to a commit I did yesterday. I’ll see if I can reproduce the issue using the file you link to above.

[DavidAnson]

Oh no… The pattern of calling forEachInlineCodeSpan(tokenLines.join("\n"), ...) is very problematic at the moment. In this case, token corresponds to one of the cells of the table and tokenLines includes all 26695 lines of the table. Each line contains 2-4 tokens depending on the scenario, and the entire table is re-scanned for every token. Not good. This is fixable, but it’s an issue with the library and I’ll transfer the issue there.