Re: query-replace-interactive

[Juri Linkov <juri@jurta.org>]

Richard Stallman <rms@gnu.org> writes:
> I misunderstood the idea of "(default a->b)".  Now that I see what
> it really does, I agree with you that it has lots of problems.
>
> Here's what I THOUGHT the idea was.
>
> The idea is that M-x query-replace would read just one minibuffer argument.
> That argument would have the form FROM->TO.  After the arg is read,
> the interactive spec would split it apart at the -> to get the FROM
> and TO strings, and they would become the arguments in the call
> to query-replace.
>
> Further, the default for this one argument would be the previous
> argument to query-replace.  The default would be used if you type RET
> and available to edit with M-n.
>
> This is radical, perhaps too radical, but it is clean and coherent.

The original Stefan's idea was to display "(default a->b)" in the
prompt for the from-string and to interpret empty input of the
from-string as a request for repeating the last replacement.

I extended this idea to what you described above, i.e. splitting the
from-string argument in the form "FROM -> TO".  This would be more
consistent with standard Emacs handling of default values, but it has
one drawback: the separator should not appear in the text to replace.
But if the string " -> " is rare enough to appear in replacement strings
we could use it as a default separator and to allow to quote it (e.g.
" \-\> ") if someone really wants to use it in the from-string.

-- 
Juri Linkov
http://www.jurta.org/emacs/