Re: [EXPERIMENTAL PATCH] Extending Isearch-repeat-forward/backward to support a prefix argument following suggesion by Juri Linkov

[Gideon Stupp <gideon.stupp@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2707 bytes --]

On Mon, Jan 30, 2012 at 2:38 AM, Juri Linkov <juri@jurta.org> wrote:

> > I am attaching a short patch which implements prefix arg support for
> > 'isearch-repeat-forward' and 'isearch-repeat-backward'. It does not
> contain
> > any visual hints or even hooks for visual support because it seems
> > worthwhile to test this functionality for a while before adding anything
> > "fancy" as you say.  Negative arguments are supported in what seems to
> me a
> > reasonable way, though clearly not ideal for regex searches as point out.
> >
> > Please let me know if you think any more changes are required.
>
> It's not yet clear what is the most expected behavior.
>
> With your patch, typing `C-s a C-3 C-s C- C-3 C-s' returns the isearch
> point
> to the original position - this is good, thanks.
>
> But its behavior is not deterministic: it depends on the current search
> direction before calling these commands with a numeric argument.
> So e.g. `C-s a C-3 C-s C-r C- C-3 C-s' (where `C-r' just switches
> the search direction) doesn't return it to the original position.
> Perhaps you need to check the current direction defined in the variable
> `isearch-forward' and take it into account.
>
> This is bug; I did not think of this scenario. Thanks for pointing it out.


> Another question is why e.g. `C-s a C-3 C-s C-3 C-r' doesn't handle
> switching of search direction and does not return to the original position?
> In terms of implementation, the question is: why in
> `isearch-repeat-forward' and `isearch-repeat-backward' you don't add code
> that switches the current direction to the `(> arg 0)' code branch?
>

I did not add code that switches the current direction in the (> arg 0)
case for compatibility reasons; I wanted C-3 C-s to behave exactly like C-s
C-s C-s.
But I do see your point. So how about the following behavior (for brevity I
will describe only isearch-repeat-forward):  (1) if you use
isearch-repeat-forward then at the end of the operation you must be in a
search-forward state; (2) the numerical value decides how many searches are
done and (3) the sign defines the direction. More specifically

isearch-repeat-forward   -|  the usual "interactive" mode.

C-0 isearch-repeat-forward -|  if in backward search switch to forward
search but don't do any actual search, otherwise don't do anything.
C-<number> isearch-repeat-forward -| switch to forward search (if
necessary) and search <number> times forward.
C-u isearch-repeat-forward              -| like C-4

C-- C-<number> isearch-repreat-forward -| go back <number> times of search
but make sure to finish in forward-search state.
C-- isearch-repeat-forward  -|  like C- C-1 isearch-repeat-forward.

Thanks, Gideon.

[-- Attachment #2: Type: text/html, Size: 3414 bytes --]