Use M-s prefix and `search-map` bindings when using ctrlf about ctrlf

[andyleejordan]

Heh:

C-w runs the command isearch-yank-word-or-char

Hit this one too. I think it would be ideal if ctrlf were as close as a drop-in upgrade to isearch as possible.

[raxod502]

Yep. Great suggestions, thanks.

[jnboehm]

I'm also in full support of this, the lack of C-w tripped me up.

[publicimageltd]

Thanks a lot for a nice package, I like it very much! I just wanted to add 1 for C-w, which I use often, too. :-)

[raxod502]

Hmmm... actually, not sure I can totally support C-w as a top-level binding. Maybe we can put it under M-s or something, and we can certainly make the keymap customizable so you could add it yourself, but one of the goals of CTRLF is that the search input behaves normally. So you'd expect C-w to kill the region. I suppose I wouldn't be opposed to installing by default a special command that will delegate to isearch-yank-word-or-char behavior if and only if there is no active region.

[publicimageltd]

Good idea, making it depend on the active region. I tried a simple conditional which checks for active mark, but simply calling isearch-yank-word-or-char caused an error. I couldn't find out what caused it. There's some implicit dependence in the code I couldn't figure out. I also tried to explicitly set the text buffer current before calling isearch...., no luck. So I came up with this solution, which is at least a first workaround:

  (defun ctrlf-c-w (begin end)
    "Kill region if active, else yank symbol at point."
    (interactive "r")
    (if mark-active
	(kill-region begin end)
      (ctrlf-forward-symbol-at-point)))

This works. If you have any ideas how to delegate it to isearch, I'd add this. But as I said, simply calling isearch-yank-word-or-char, either with the text buffer explicitly set as current buffer or not, does not work.

[raxod502]

Well, looking at the source for isearch-yank-internal, I'd say the problem is Isearch keeps point in the actual buffer being searched, while CTRLF keeps it in the minibuffer. Thus buffer-substring-no-properties has entirely the wrong effect. Furthermore, isearch-yank-word-or-char operates by setting a bunch of Isearch variables and calling some Isearch functions, not by just inserting text into the minibuffer. The whole thing would need to be redone for CTRLF.

Protip: there is a function use-region-p you can use instead of checking the mark-active variable.