bug#55110: 29.0.50; Regression in query-replace prompt

bug#55110: 29.0.50; Regression in query-replace prompt

[Juri Linkov]

X-Debbugs-Cc: Augusto Stoffel <arstoffel@gmail.com>

I noticed a recent regression after minibuffer-lazy-highlight changes:

0. emacs -Q
1. perform one query-replace to add some FROM and TO to its history:
   M-% a RET b RET
2. check that isearch still works in the minibuffer history:
   M-% C-r a
3. activate the region, e.g. C-SPC and move point
4. try to isearch in the history again:
   M-% C-r a

It fails with: [Failing I-search backward: a]

Maybe this is caused by minibuffer-lazy-highlight-setup
that sets filter to replace--region-filter in the minibuffer
instead of the original buffer?

bug#55110: 29.0.50; Regression in query-replace prompt

[Augusto Stoffel]

Thanks for the heads up, I'll take a look at this some time soon.

On Mon, 25 Apr 2022 at 18:37, Juri Linkov <juri@linkov.net> wrote:

> Maybe this is caused by minibuffer-lazy-highlight-setup
> that sets filter to replace--region-filter in the minibuffer
> instead of the original buffer?

Most likely, yes.  `replace--region-filter' is modified globally, so a
similar problem should happen if you temporarily leave the minibuffer
and do Isearch in any other buffer.

If that's the case, I think we would have two options:

1) Add a quick fix for the minibuffer Isearch only.

2) A more complicated change that solves the issue generally by saving
the region filter in the fashion of isearch-lazy-highlight-regexp et
alii.

WDYT?

bug#55110: 29.0.50; Regression in query-replace prompt

[Juri Linkov]

>> Maybe this is caused by minibuffer-lazy-highlight-setup
>> that sets filter to replace--region-filter in the minibuffer
>> instead of the original buffer?
>
> Most likely, yes.  `replace--region-filter' is modified globally, so a
> similar problem should happen if you temporarily leave the minibuffer
> and do Isearch in any other buffer.
>
> If that's the case, I think we would have two options:
>
> 1) Add a quick fix for the minibuffer Isearch only.
>
> 2) A more complicated change that solves the issue generally by saving
> the region filter in the fashion of isearch-lazy-highlight-regexp et
> alii.
>
> WDYT?

Recently we fixed a similar problem in `perform-replace'
by creating a dynamically bound value in `let':

  (let ((opos (point-marker))
        ;; Restore original isearch filter to allow
        ;; using isearch in a recursive edit even
        ;; when perform-replace was started from
        ;; `xref--query-replace-1' that let-binds
        ;; `isearch-filter-predicate' (bug#53758).
        (isearch-filter-predicate #'isearch-filter-visible))

So maybe a buffer-local value of `isearch-filter-predicate'
in the minibuffer would help.

Also I recommend to make all hooks in `minibuffer-lazy-highlight-setup'
local by adding the argument LOCAL to add-hook/remove-hook.

bug#55110: 29.0.50; Regression in query-replace prompt

[Augusto Stoffel]

[-- Attachment #1: Type: text/plain, Size: 1944 bytes --]

On Wed, 27 Apr 2022 at 10:44, Juri Linkov <juri@linkov.net> wrote:

>>> Maybe this is caused by minibuffer-lazy-highlight-setup
>>> that sets filter to replace--region-filter in the minibuffer
>>> instead of the original buffer?
>>
>> Most likely, yes.  `replace--region-filter' is modified globally, so a
>> similar problem should happen if you temporarily leave the minibuffer
>> and do Isearch in any other buffer.
>>
>> If that's the case, I think we would have two options:
>>
>> 1) Add a quick fix for the minibuffer Isearch only.
>>
>> 2) A more complicated change that solves the issue generally by saving
>> the region filter in the fashion of isearch-lazy-highlight-regexp et
>> alii.
>>
>> WDYT?
>
> Recently we fixed a similar problem in `perform-replace'
> by creating a dynamically bound value in `let':
>
>   (let ((opos (point-marker))
>         ;; Restore original isearch filter to allow
>         ;; using isearch in a recursive edit even
>         ;; when perform-replace was started from
>         ;; `xref--query-replace-1' that let-binds
>         ;; `isearch-filter-predicate' (bug#53758).
>         (isearch-filter-predicate #'isearch-filter-visible))
>
> So maybe a buffer-local value of `isearch-filter-predicate'
> in the minibuffer would help.

Yes, that indeed solves the problem.  See attached patch.

> Also I recommend to make all hooks in `minibuffer-lazy-highlight-setup'
> local by adding the argument LOCAL to add-hook/remove-hook.

Indeed, the minibuffer lazy highlight feature is currently incompatible
with recursive minibuffers.  The patch fixes that as well.

There's a caveat, though: isearch in a recursive minibuffer is again
affected by the presence of an inappropriate filter function.  Fixing
that in a robust way might require a bigger refactoring of the lazy
highlight feature, I think.  Another option might be to make
`replace--region-filter' also check for the value of `(current-buffer)'.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-Make-minibuffer-lazy-highlight-setup-buffer-local-wh.patch --]
[-- Type: text/x-patch, Size: 1599 bytes --]

From 37de4e53fce6adbf5e52f3184c163574e3129def Mon Sep 17 00:00:00 2001
From: Augusto Stoffel <arstoffel@gmail.com>
Date: Sat, 14 May 2022 17:21:27 +0200
Subject: [PATCH] Make minibuffer lazy highlight setup buffer-local where
 appropriate

* lisp/isearch.el (minibuffer-lazy-highlight-setup): Modify hooks
buffer-locally, so that recursive minibuffers are not affected by the
special behavior of lazy-highlight.  Also make
'isearch-filter-predicate' buffer-local, so that isearch in the
minibuffer is not affected by the region filter (bug#55110).
---
 lisp/isearch.el | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/lisp/isearch.el b/lisp/isearch.el
index 96168f94bd..3e1dab4d15 100644
--- a/lisp/isearch.el
+++ b/lisp/isearch.el
@@ -4441,12 +4441,13 @@ minibuffer-lazy-highlight-setup
                                 (format minibuffer-lazy-count-format
                                         isearch-lazy-count-total)))))
       (lambda ()
-        (add-hook 'minibuffer-exit-hook unwind)
-        (add-hook 'after-change-functions after-change)
+        (add-hook 'minibuffer-exit-hook unwind nil t)
+        (add-hook 'after-change-functions after-change nil t)
         (when minibuffer-lazy-count-format
           (setq overlay (make-overlay (point-min) (point-min) (current-buffer) t))
           (add-hook 'lazy-count-update-hook display-count))
         (when filter
+          (make-local-variable 'isearch-filter-predicate)
           (add-function :after-while isearch-filter-predicate filter))
         (funcall after-change nil nil nil)))))
 
-- 
2.36.1

bug#55110: 29.0.50; Regression in query-replace prompt

[Juri Linkov]

>> So maybe a buffer-local value of `isearch-filter-predicate'
>> in the minibuffer would help.
>
> Yes, that indeed solves the problem.  See attached patch.

Thanks, pushed.

>> Also I recommend to make all hooks in `minibuffer-lazy-highlight-setup'
>> local by adding the argument LOCAL to add-hook/remove-hook.
>
> Indeed, the minibuffer lazy highlight feature is currently incompatible
> with recursive minibuffers.  The patch fixes that as well.
>
> There's a caveat, though: isearch in a recursive minibuffer is again
> affected by the presence of an inappropriate filter function.  Fixing
> that in a robust way might require a bigger refactoring of the lazy
> highlight feature, I think.  Another option might be to make
> `replace--region-filter' also check for the value of `(current-buffer)'.

I see no problem to fix this because every minibuffer uses own buffer,
so they have separate buffer-local variables.