Wrong hat positions in dired mode b/c of hidden text

[rntz]

I think this is because my dired-mode setup hides some text (the file modes, owner, mtime, etc), but it still shows up in the buffer contents and gets sent to cursorless, so the hat positions are messed up.

[dyamito]

I think we need to be using buffer-substring-no-properties instead of write-region here:

emacs-cursorless/cursorless-state.el

Line 73 in 5e0f8e7

(write-region nil nil (cursorless-temporary-file-path) nil 'ignore-message))

That way it will only write out the visible text. One downside is I think that would bring everything into RAM, which might be wasteful for bigger files, shells, etc.

[rntz]

buffer-substring-no-properties doesn't do what we want either. eg. if I go to a dired buffer I see:

  /home/rntz/t/example:
  total used in directory 25 available 40.6 GiB
  * example.txt

but the actual text that is produced by either write-region or by buffer-substring-no-properties or by copying is:

   /home/rntz/t/example:
  total used in directory 25 available 40.6 GiB
  -rw-rw-r-- rntz rntz     5 09-03 23:24 example.txt

This is because of dired-details: it hides the modes, owner, group, mtime, etc; replacing them by *. It does this using an overlay which makes the hidden text invisible and inserts the string "* " before it.

Clearly we don't want to send the hidden text to VSCode. Oddly enough we also might not want to send the replacement text ("* ") to VSCode, because it's not editable and the cursor can't actually select it. Either way involves some annoying arithmetic to figure out how to sync hat/cursor positions b/w emacs & vscode.

[dyamito]

Ugh, you're right. I ran into this same issue with magit, both are using the invisibility spec stuff - https://www.gnu.org/software/emacs/manual/html_node/elisp/Invisible-Text.html.

Oddly enough we also might not want to send the replacement text ("* ") to VSCode, because it's not editable and the cursor can't actually select it.

I think we can punt on this for a while, this is a whole can of worms that we're going to run into with shells too - e.g. allow editing the command line but not any of the output from previous commands.