bug#53126: 29.0.50; [PATCH] Lazy highlight/count when reading query-replace string, etc.

[Juri Linkov <juri@linkov.net>]

>> Strange, I expected that the lazy-count feature applied to query-replace
>> should do a different thing - when querying about replacing the current match,
>> it should show the number of current match and the number of total matches.
>> The latter number should be updated after every replacement.
>
> That would be handy too.  I haven't checked how the `perform-replace'
> loop works, but wouldn't this require a hook into the lazy count
> procedure --- precisely what my patch introduces through the new
> variable `isearch-lazy-count-display-function'?

It seems the "Query replacing" prompt in perform-replace could just
display the values of isearch-lazy-count-current and
isearch-lazy-count-total.  Or for asynchronous lazy-count a new hook
needs to be run at the end of isearch-lazy-highlight-buffer-update
in addition to the call of isearch-message.

>> But your patches are intended for a different feature - highlighting
>> of matches in the buffer while entering an input string in the minibuffer.
>>
>> I wonder how many users need this feature, when it's easy to construct
>> a query-replace string using highlighting/counting in isearch-mode,
>> then type M-%
>> (isearch-query-replace) that invokes query-replace with the
>> query-replace string.
>
> Sure, this alternative method works.  But somehow it's not the way I
> usually start a replace, and I think there's nothing wrong with that
> preference :-)

But this method is more limiting - no keys to pull text from the buffer
like 'C-w' (isearch-yank-word-or-char) does in isearch-mode, no way
to navigate to the first match like it's possible in isearch-mode, etc.

>> Perhaps better to ask on emacs-devel if anyone needs this feature.
>
> We can ask.  But note that the bulk of the patch has a fairly general
> purpose, namely to read a regexp with live preview of the matches.
> There's probably a couple extra uses for this.

Maybe then this feature could be added to read-regexp or even
to read-from-minibuffer?  And activated by adding a setup function
to minibuffer-setup-hook like other minibuffer's features do, such as
icomplete-minibuffer-setup, minibuffer-depth-setup, rfn-eshadow-setup-minibuffer.

Then maybe a new feature could be named e.g. "lazy-minibuffer"?