C-r and C-s in minibuffer should search completion

C-r and C-s in minibuffer should search completion

[Stefan Monnier]

I believe C-s and C-r in the minibuffer should not only search the
history but the completion table as well (at least when it's not
a function).

Could someone implement this feature?


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

> I believe C-s and C-r in the minibuffer should not only search the
> history but the completion table as well (at least when it's not
> a function).

There is one problem: what to search first - the history or the
completion table.  Either preference will cause inconveniences.
If the user wants to search the history (like implemented in shells),
we should not let isearch to go thru all completion matches before
starting to search the history, and vice versa.

But really there is no need to mix two separate search spaces.
There exist already a convenient key sequence to search the completion
table: `PgUp C-s'.  PgUp switches to the *Completions* buffer and
C-s starts searching among completion candidates.

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

Re: C-r and C-s in minibuffer should search completion

[Lennart Borgman (gmail)]

Juri Linkov wrote:
>> I believe C-s and C-r in the minibuffer should not only search the
>> history but the completion table as well (at least when it's not
>> a function).
> 
> There is one problem: what to search first - the history or the
> completion table.  Either preference will cause inconveniences.
> If the user wants to search the history (like implemented in shells),
> we should not let isearch to go thru all completion matches before
> starting to search the history, and vice versa.
> 
> But really there is no need to mix two separate search spaces.
> There exist already a convenient key sequence to search the completion
> table: `PgUp C-s'.  PgUp switches to the *Completions* buffer and
> C-s starts searching among completion candidates.

Why not let PgDn switch between searching history and completions? Or 
maybe cycling history/completions/history+completions?

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> But really there is no need to mix two separate search spaces.
>> There exist already a convenient key sequence to search the completion
>> table: `PgUp C-s'.  PgUp switches to the *Completions* buffer and
>> C-s starts searching among completion candidates.
>
> Why not let PgDn switch between searching history and completions? Or
> maybe cycling history/completions/history+completions?

This would be very clumsy user interface since it is not an intuitive
key for switching search modes, and also will require an indication
somewhere for the current mode.

However, I think PgDn is a good candidate for a new command that
just like PgUp displays a list of completions in a separate buffer,
will display a list of elements from the history list in a
separate buffer.  I think this would be a useful feature.
It will allow browsing all elements of the history list in one buffer
and using normal isearch to search them.

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

RE: C-r and C-s in minibuffer should search completion

[Drew Adams]

> I believe C-s and C-r in the minibuffer should not
> only search the history but the completion table
> as well (at least when it's not a function).
> 
> Could someone implement this feature?

1. What's wrong with keeping C-s and C-r for searching minibuffer text?
That's useful.

2. We have M-s and M-r for searching the history.

3. I proposed M-o for completion against the history:

  > > Could you propose a different key to complete
  > > history lists in *all* minibuffers?
  > 
  > Yes, I could. The key that Icicles uses for this is M-o.
  >
  > http://www.emacswiki.org/cgi-bin/wiki/
  > Icicles_-_History_Enhancements#toc4

(That link also describes other useful ways to access history items.)

Pick a different key from M-o, if you like, but a separate binding for
history completion is better than co-opting the isearch bindings.

RE: C-r and C-s in minibuffer should search completion

[Drew Adams]

> However, I think PgDn is a good candidate for a new command that
> just like PgUp displays a list of completions in a separate buffer,
> will display a list of elements from the history list in a
> separate buffer.  I think this would be a useful feature.
> It will allow browsing all elements of the history list in one buffer
> and using normal isearch to search them.

I proposed M-o, but which key you choose is not so important. 

However, I would argue to reserve the pair of keys `next' and `prior'
(sometimes labeled PgDn and PgUp) for something that, well, comes in pairs
and has some notion of, well, next and previous. That just makes sense.

There is no need to use a different buffer from *Completions*. When the user
hits M-o (or whatever), the current history list should simply become the
*Completions* set; that's all. Nothing special.

You don't need to use isearch to search the completion candidates (you can
complete against them, as usual), but you of course can (as always), just by
moving the cursor to buffer *Completions* and then using C-s.

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> I believe C-s and C-r in the minibuffer should not only search the
>> history but the completion table as well (at least when it's not
>> a function).

Sorry, I meant M-r/M-s.

> There is one problem: what to search first - the history or the
> completion table.  Either preference will cause inconveniences.

No problem: Make sure you start the search in between the two sets, so
that M-r will first search one and M-s will first search the other.
I'd guess M-r should first search through history, while M-s should
first search through the completion table.

BTW, once this is done, we can revert the recent change to the
`read-buffer' that adds all the buffers to the list of defaults.


        Stefan

RE: C-r and C-s in minibuffer should search completion

[Drew Adams]

> >> I believe C-s and C-r in the minibuffer should not only search the
> >> history but the completion table as well (at least when it's not
> >> a function).
> 
> Sorry, I meant M-r/M-s.

Ah, I wondered about that. Thanks for clarifying.

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>>> I believe C-s and C-r in the minibuffer should not only search the
>>> history but the completion table as well (at least when it's not
>>> a function).
>
> Sorry, I meant M-r/M-s.
>
>> There is one problem: what to search first - the history or the
>> completion table.  Either preference will cause inconveniences.
>
> No problem: Make sure you start the search in between the two sets, so
> that M-r will first search one and M-s will first search the other.
> I'd guess M-r should first search through history, while M-s should
> first search through the completion table.

I think it would be unnatural to use M-s to search a list of elements
not available for navigation in the minibuffer using M-p and M-n,
because a set of keys M-p/M-n/M-r/M-s should operate on the same
space for consistency.  Otherwise, it would be confusing when
M-s will insert in the minibuffer an element not from the list
where M-p and M-n navigate.

One solution is to add a new key to switch dimensions like proposed
by Drew and Lennart, i.e. after typing a special key to change the
current mode, M-n/M-p will navigate to the next/previous completion
candidate in the minibuffer, M-s/M-r will search the completion list
forward/backward in the minibuffer.

> BTW, once this is done, we can revert the recent change to the
> `read-buffer' that adds all the buffers to the list of defaults.

Why?  The buffer list in the list of defaults is useful not only for
searching, but also for browsing, i.e. switching to the penultimate
buffer is easy with `C-x b M-n M-n RET', to the third-to-last buffer -
`C-x b M-n M-n M-n RET', etc.  This is harmless improvement since
it doesn't change the existing behavior, just adds more buffers
sorted by recency beyond a single most recent buffer.

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

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> However, I think PgDn is a good candidate for a new command that
>> just like PgUp displays a list of completions in a separate buffer,
>> will display a list of elements from the history list in a
>> separate buffer.  I think this would be a useful feature.
>> It will allow browsing all elements of the history list in one buffer
>> and using normal isearch to search them.
>
> I proposed M-o, but which key you choose is not so important.
>
> However, I would argue to reserve the pair of keys `next' and `prior'
> (sometimes labeled PgDn and PgUp) for something that, well, comes in pairs
> and has some notion of, well, next and previous. That just makes sense.

I agree that PgUp and PgDn are not the best keys since when there is
no completion they just do the same as M-n and M-p, so they behave
differently when there is a completion list.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> BTW, once this is done, we can revert the recent change to the
>> `read-buffer' that adds all the buffers to the list of defaults.

> Why?

Because it's a special case and I'm not convinced it's justified.
I expect most people who like this will prefer iswitchb anyway.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>>> BTW, once this is done, we can revert the recent change to the
>>> `read-buffer' that adds all the buffers to the list of defaults.
>
>> Why?
>
> Because it's a special case and I'm not convinced it's justified.

The intention of adding a list of default values was to allow selecting
more than one default value extracted from the context of the command.
So after running a command the user can type M-n to see if the first
proposed default value is that the user likes.  If not, the user can
then type M-n again to see the next proposed value and accept it
or type M-n to see a few next useful values.

However, I agree that putting all buffers pushes this features to the
extreme because then the list of defaults is the same as the completion
list.  But there is one significant difference: the list of defaults is
sorted by recency, so switching between several work buffers (more than
two) is much easier with `C-x b M-n M-n M-n RET'.

OTOH, searching the completion list would be useful in itself
independent of searching the history list or the list of defaults
because often these three are disjoint sets, and the list of defaults
can be presented even when the completion list is empty.

> I expect most people who like this will prefer iswitchb anyway.

iswitchb radically changes the user interface of the minibuffer, so
I highly recommend providing most features of iswitchb to the users
who don't like the user interface of iswitchb.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>>>> BTW, once this is done, we can revert the recent change to the
>>>> `read-buffer' that adds all the buffers to the list of defaults.
>> 
>>> Why?
>> 
>> Because it's a special case and I'm not convinced it's justified.

> The intention of adding a list of default values was to allow selecting
> more than one default value extracted from the context of the command.

Yes, I said nothing about this.

> However, I agree that putting all buffers pushes this features to the
> extreme because then the list of defaults is the same as the completion
> list.  But there is one significant difference: the list of defaults is
> sorted by recency, so switching between several work buffers (more than
> two) is much easier with `C-x b M-n M-n M-n RET'.

I never said there's no difference.  I just think this special case is
not justified.  Most likely `C-x b M-p M-p M-p RET' would work just
as well.

> OTOH, searching the completion list would be useful in itself
> independent of searching the history list or the list of defaults
> because often these three are disjoint sets, and the list of defaults
> can be presented even when the completion list is empty.

If you like it, feel free to customize your Emacs to provide you this
kind of refinement with different commands to search the history, the
list of defaults, and the list of completions.  But I think the default
search function bound to M-r/M-s would benefit from searching all
of them.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> However, I agree that putting all buffers pushes this features to the
>> extreme because then the list of defaults is the same as the completion
>> list.  But there is one significant difference: the list of defaults is
>> sorted by recency, so switching between several work buffers (more than
>> two) is much easier with `C-x b M-n M-n M-n RET'.
>
> I never said there's no difference.  I just think this special case is
> not justified.  Most likely `C-x b M-p M-p M-p RET' would work just
> as well.

This assumes that the user prefers C-x b to switch to a new buffer.
But it is not the case since more often many other commands switch to
a buffer that don't leave a new buffer name in the history list.

>> OTOH, searching the completion list would be useful in itself
>> independent of searching the history list or the list of defaults
>> because often these three are disjoint sets, and the list of defaults
>> can be presented even when the completion list is empty.
>
> If you like it, feel free to customize your Emacs to provide you this
> kind of refinement with different commands to search the history, the
> list of defaults, and the list of completions.  But I think the default
> search function bound to M-r/M-s would benefit from searching all
> of them.

Hmm, now I tend to agree with you that M-s could search the list of
completions as well.  Since the list of defaults usually is very short,
we could search it first and after that search the list of completions.
When the list of defaults is not very short then often there is no
completions at all, so this should work quite well.

The only problem with C-x b is that the buffer list is sorted
alphabetically in the list of completions, but it would be much more
useful to search them in the order of recency.

I see they are sorted in the function `minibuffer-completion-help' as

      /* Sort and remove duplicates.  */
      Lisp_Object tmp = completions = Fsort (completions, Qstring_lessp);

Maybe we should create a new function `minibuffer-completion-help-unsorted'
for C-x b and possibly other commands that might need to keep the original
order of elements too?

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> The only problem with C-x b is that the buffer list is sorted
> alphabetically in the list of completions, but it would be much more
> useful to search them in the order of recency.

> I see they are sorted in the function `minibuffer-completion-help' as

>       /* Sort and remove duplicates.  */
>       Lisp_Object tmp = completions = Fsort (completions, Qstring_lessp);

I'm not sure I understand the relevance: M-r and M-s would not use
minibuffer-completion-help anyway.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

> I believe C-s and C-r in the minibuffer should not only search the
> history but the completion table as well (at least when it's not
> a function).
>
> Could someone implement this feature?

I now have an idea how to implement this feature.

Some time ago I submitted a patch that introduces a new variable
`minibuffer-default-set-function' that extends a list of defaults
when minibuffer position hits the bottom of the list of defaults.

http://lists.gnu.org/archive/html/emacs-devel/2007-12/msg00179.html

This patch was approved by Richard but not installed because
it was undecided where to move mailcap.el.

The patch uses this variable for `shell-command' to append a list of
available mailcap commands to the list of defaults only when the user
types M-n below the default value.

The same could be done for the list of completions.  The patch below
implements a new function `minibuffer-default-add-completions' that
appends a list of all completions to the end of the list of defaults.
It would be good to use this function by default since it is useful
to search the completion table in all minibuffers that have it.

Index: lisp/simple.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/simple.el,v
retrieving revision 1.911
diff -c -r1.911 simple.el
*** lisp/simple.el	23 Mar 2008 22:52:20 -0000	1.911
--- lisp/simple.el	25 Mar 2008 21:43:34 -0000
***************
*** 1304,1313 ****
--- 1305,1339 ----
  
  (defvar minibuffer-temporary-goal-position nil)
  
+ (defvar minibuffer-default-set-function 'minibuffer-default-add-completions
+   "*Function run by `goto-history-element' before using `minibuffer-default'.
+ This is useful to dynamically set the value of `minibuffer-default'
+ before `goto-history-element' reads it when moves into the future.")
+ 
+ (make-variable-buffer-local 'minibuffer-default-set-function)
+ 
+ (defun minibuffer-default-add-completions ()
+   "Return a list of all completions without the default value.
+ This function is used to add all elements of the completion table to
+ the end of the list of defaults after the default value."
+   (interactive)
+   (let ((def minibuffer-default)
+ 	(all (all-completions ""
+ 			      minibuffer-completion-table
+ 			      minibuffer-completion-predicate
+ 			      t)))
+     (if (listp def)
+ 	(append def all)
+       (setq all (cons def (delete def all))))))
+ 
  (defun goto-history-element (nabs)
    "Puts element of the minibuffer history in the minibuffer.
  The argument NABS specifies the absolute history position."
    (interactive "p")
+   (when (and (functionp minibuffer-default-set-function)
+ 	     (< nabs (if (stringp minibuffer-default) -1 0)))
+     (setq minibuffer-default (funcall minibuffer-default-set-function)
+ 	  minibuffer-default-set-function nil))
    (let ((minimum (if minibuffer-default
  		     (- (if (listp minibuffer-default)
  			    (length minibuffer-default)

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> I believe C-s and C-r in the minibuffer should not only search the
>> history but the completion table as well (at least when it's not
>> a function).
>> 
>> Could someone implement this feature?

> I now have an idea how to implement this feature.

> Some time ago I submitted a patch that introduces a new variable
> `minibuffer-default-set-function' that extends a list of defaults
> when minibuffer position hits the bottom of the list of defaults.

How would that help?  M-r/M-s doesn't seem to search minibuffer-default.
But I guess it does help with C-s, which is what I mistyped.
Actually, now that I take a closer look.  Why do we have both C-s/C-r
and M-r/M-s?  They seem redundant.

> + (defun minibuffer-default-add-completions ()
> +   "Return a list of all completions without the default value.
> + This function is used to add all elements of the completion table to
> + the end of the list of defaults after the default value."
> +   (interactive)
> +   (let ((def minibuffer-default)
> + 	(all (all-completions ""
> + 			      minibuffer-completion-table
> + 			      minibuffer-completion-predicate
> + 			      t)))
> +     (if (listp def)
> + 	(append def all)
> +       (setq all (cons def (delete def all))))))

This is not good when the completion table is a function that takes
a long time, because then you have to wait for the completion table to
be built before you can get the actual real DEF.

Better would be to call the function only after we consumed
minibuffer-default.  Or maybe only do it when the completion-table is
not a function.


        Stefan

RE: C-r and C-s in minibuffer should search completion

[Drew Adams]

> Actually, now that I take a closer look.  Why do we have both C-s/C-r
> and M-r/M-s?  They seem redundant.

They are not redundant now, and I hope we don't make them redundant or combine
them. C-s searches the text in the minibuffer. M-s searches the input history.
M-s is a regexp search; C-s is not (but C-M-s is).

I'm sure you know that, so I probably misunderstand your proposal.

> This is not good when the completion table is a function that takes
> a long time, because then you have to wait for the completion table to
> be built before you can get the actual real DEF.
> 
> Better would be to call the function only after we consumed
> minibuffer-default. Or maybe only do it when the completion-table is
> not a function.

I really do not see any gain by confounding history-list search with
completion-table search (and now with minibuffer-text search also?). I think the
proposed amalgam is a mistake. What problem is it trying to solve?

[What is useful, however, is adding a key to _complete_ against the current
history - in all minibuffer maps. I suggested M-o for that.]

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> I now have an idea how to implement this feature.
>
>> Some time ago I submitted a patch that introduces a new variable
>> `minibuffer-default-set-function' that extends a list of defaults
>> when minibuffer position hits the bottom of the list of defaults.
>
> How would that help?  M-r/M-s doesn't seem to search minibuffer-default.

It would be trivial to add in `previous-matching-history-element' the
same extending a list of defaults with a list of completions when M-s
hits the bottom of the list.

> But I guess it does help with C-s, which is what I mistyped.
> Actually, now that I take a closer look.  Why do we have both C-s/C-r
> and M-r/M-s?  They seem redundant.

As everywhere in normal modes there are two search types: incremental
and non-incremental, in the minibuffer C-s/C-r provides incremental
search, and M-r/M-s - non-incremental.

In normal modes non-incremental search is provided by `C-s RET'
but in the minibuffer it displays a different prompt ("Search: ")
than M-s "Next element matching (regexp): ".

Also M-s is a backward-compatible keybinding for users that already
use it for many years.

> This is not good when the completion table is a function that takes
> a long time, because then you have to wait for the completion table to
> be built before you can get the actual real DEF.

Just as when the user needs to see a full list of completions, the user
can bear a long time to built the completion table, when the user needs
to search a full list of completions, it is exactly the same situation: the
user have to wait when the user wants to search the completion list
(though I don't know where in Emacs building the completion table takes
too long time, even C-h f takes less than 1 sec on modern hardware, and
woman takes 2 sec to build her very large completion table but does before
displaying the minibuffer.)

> Better would be to call the function only after we consumed
> minibuffer-default.  Or maybe only do it when the completion-table is
> not a function.

It now calls the function after moved below minibuffer-default so it seems
this is good, and also it seems good to build the completion-table even
from a function because I think it is not a problem to wait a few seconds
if there is a need to search the completion-table built from a function.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> C-s searches the text in the minibuffer.

It used to do it in Emacs-22, but it's not doing just that any more now.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> How would that help?  M-r/M-s doesn't seem to search minibuffer-default.

> It would be trivial to add in `previous-matching-history-element' the
> same extending a list of defaults with a list of completions when M-s
> hits the bottom of the list.

Oh, I see, so I didn't miss anything.  I guess that's OK, except that
part is still missing ;-)
And since C-s works better anyway (who wants non-incremental search
when you can have isearch instead).

>> This is not good when the completion table is a function that takes
>> a long time, because then you have to wait for the completion table to
>> be built before you can get the actual real DEF.

> Just as when the user needs to see a full list of completions, the user
> can bear a long time to build the completion table, when the user needs
> to search a full list of completions, it is exactly the same situation:

The difference is that in this case the user may not want to search
anything at all, she just wants her M-n to bring up the default which
was already provided and doesn't require any computation.

> (though I don't know where in Emacs building the completion table takes
> too long time, even C-h f takes less than 1 sec on modern hardware, and
> woman takes 2 sec to build her very large completion table but does before
> displaying the minibuffer.)

Believe me there are such cases.  Info-lookup is one, M-. is another,
Tramp is yet another, tags-completion in PCL-CVS, .... the list goes on
and on.

>> Better would be to call the function only after we consumed
>> minibuffer-default.  Or maybe only do it when the completion-table is
>> not a function.

> It now calls the function after moved below minibuffer-default so it seems
> this is good,

There seem to be some typo or something missing from your sentence, but
IIUC you have implemented what I suggested so that M-n only computes the
completion table after having consumed the normal defaults?  Can you
send the new code?


        Stefan

RE: C-r and C-s in minibuffer should search completion

[Drew Adams]

> > C-s searches the text in the minibuffer.
> 
> It used to do it in Emacs-22, but it's not doing just that 
> any more now.

I haven't seen the details, but my first impression is, "Pity".

Cursory reading (apologies for not digging in more earnestly) hints that
completion is willy nilly becoming a chimera. Some things belong together and
work well together; other things don't.

How about summarizing the changes made since Emacs 22 as well as the proposed
behavior and its rationale?

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

> There seem to be some typo or something missing from your sentence, but
> IIUC you have implemented what I suggested so that M-n only computes the
> completion table after having consumed the normal defaults?  Can you
> send the new code?

Yep, this was already implemented in my previous patch.  If there exist
a default value, then the first M-n just inserts this default value to
the minibuffer without any computation, and only next M-n (that would
otherwise signal an error) computes the completion table.

Below is a new patch with a small improvement: it introduces a new
variable `minibuffer-default-add-p' useful to add more elements to the
list of defaults several times in chunks when needed (it can keep the
current state of added elements to the list of defaults):

Index: lisp/simple.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/simple.el,v
retrieving revision 1.912
diff -c -r1.912 simple.el
*** lisp/simple.el	26 Mar 2008 03:40:40 -0000	1.912
--- lisp/simple.el	27 Mar 2008 00:42:03 -0000
***************
*** 1304,1313 ****
--- 1305,1352 ----
  
  (defvar minibuffer-temporary-goal-position nil)
  
+ (defvar minibuffer-default-add-function 'minibuffer-default-add-completions
+   "Function run by `goto-history-element' before consuming `minibuffer-default'.
+ This is useful to dynamically set the value of `minibuffer-default'
+ before `goto-history-element' reads it when it moves into the future.")
+ 
+ (make-variable-buffer-local 'minibuffer-default-add-function)
+ 
+ (defvar minibuffer-default-add-p t
+   "When non-nil, add more elements to the end of the list of default values.
+ Any non-nil value causes `goto-history-element' to add more elements
+ to a list of defaults by calling a function defined by
+ `minibuffer-default-add-function'.  When this function
+ doesn't need to add more elements after its call, it should
+ set this variable to nil.")
+ 
+ (make-variable-buffer-local 'minibuffer-default-add-p)
+ 
+ (defun minibuffer-default-add-completions ()
+   "Return a list of all completions without the default value.
+ This function is used to add all elements of the completion table to
+ the end of the list of defaults after the default value."
+   (interactive)
+   (let ((def minibuffer-default)
+ 	(all (all-completions ""
+ 			      minibuffer-completion-table
+ 			      minibuffer-completion-predicate
+ 			      t)))
+     (setq minibuffer-default-add-p nil)
+     (if (listp def)
+ 	(append def all)
+       (cons def (delete def all)))))
+ 
  (defun goto-history-element (nabs)
    "Puts element of the minibuffer history in the minibuffer.
  The argument NABS specifies the absolute history position."
    (interactive "p")
+   (when (and minibuffer-default-add-p
+ 	     (functionp minibuffer-default-add-function)
+ 	     (< nabs (- (if (listp minibuffer-default)
+ 			    (length minibuffer-default)
+ 			  1))))
+     (setq minibuffer-default (funcall minibuffer-default-add-function)))
    (let ((minimum (if minibuffer-default
  		     (- (if (listp minibuffer-default)
  			    (length minibuffer-default)

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> There seem to be some typo or something missing from your sentence, but
>> IIUC you have implemented what I suggested so that M-n only computes the
>> completion table after having consumed the normal defaults?  Can you
>> send the new code?

> Yep, this was already implemented in my previous patch.

I must have missed it.

> If there exist a default value, then the first M-n just inserts this
> default value to the minibuffer without any computation, and only next
> M-n (that would otherwise signal an error) computes the
> completion table.

Yes, that's good.

> Below is a new patch with a small improvement: it introduces a new
> variable `minibuffer-default-add-p' useful to add more elements to the
> list of defaults several times in chunks when needed (it can keep the
> current state of added elements to the list of defaults):

OK, then please remove (make-variable-buffer-local
'minibuffer-default-add-function).  And I suggest to change
minibuffer-default-add-p into minibuffer-default-add-done (and to
reverse its default to nil).

Also the docstring of minibuffer-default-add-function needs to mention
that the function should set minibuffer-default-add-done.

This said, I'm wondering when the chunked computation makes sense.
Do you have an example in mind?


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> Below is a new patch with a small improvement: it introduces a new
>> variable `minibuffer-default-add-p' useful to add more elements to the
>> list of defaults several times in chunks when needed (it can keep the
>> current state of added elements to the list of defaults):
>
> OK, then please remove (make-variable-buffer-local
> 'minibuffer-default-add-function).  And I suggest to change
> minibuffer-default-add-p into minibuffer-default-add-done (and to
> reverse its default to nil).

In the patch below I renamed minibuffer-default-add-p into
minibuffer-default-add-done, changed its default to nil, and
removed make-variable-buffer-local for minibuffer-default-add-function.

Also I moved (setq minibuffer-default-add-done t) to goto-history-element
just before the call to minibuffer-default-add-function, so there is no
need to set this variable to t in minibuffer-default-add-function
explicitly in normal cases.  When necessary, minibuffer-default-add-function
can nullify minibuffer-default-add-done to be able call this function more
than once.

> Also the docstring of minibuffer-default-add-function needs to mention
> that the function should set minibuffer-default-add-done.

minibuffer-default-add-function now doesn't need to set
minibuffer-default-add-done any more, so I updated the
docstrings accordingly.

> This said, I'm wondering when the chunked computation makes sense.
> Do you have an example in mind?

I don't have a concrete example.  When the chunked computation makes
sense, a function defined by minibuffer-default-add-function can use
a new specific global variable that will count the number of
function calls.  But this is not necessary now, and the main thing
I added is the possibility to call this function more than once,
when this functions sets minibuffer-default-add-done to nil.

This patch also changes the error message "End of history; no next item"
to "End of defaults; no next item" for the end of the list of defaults.

Index: lisp/simple.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/simple.el,v
retrieving revision 1.912
diff -c -r1.912 simple.el
*** lisp/simple.el	26 Mar 2008 03:40:40 -0000	1.912
--- lisp/simple.el	27 Mar 2008 23:35:05 -0000
***************
*** 1304,1313 ****
--- 1305,1353 ----
  
  (defvar minibuffer-temporary-goal-position nil)
  
+ (defvar minibuffer-default-add-function 'minibuffer-default-add-completions
+   "Function run by `goto-history-element' before consuming `minibuffer-default'.
+ This is useful to dynamically add more elements to the list `minibuffer-default'
+ when `goto-history-element' reaches the end of this list.
+ Before calling this function `goto-history-element' sets the variable
+ `minibuffer-default-add-done' to t, so it will call this function only
+ once.  In special cases, when this function needs to be called more
+ than once, it can set `minibuffer-default-add-done' to nil explicitly,
+ overriding the setting of this variable to t in `goto-history-element'.")
+ 
+ (defvar minibuffer-default-add-done nil
+   "When nil, add more elements to the end of the list of default values.
+ The value nil causes `goto-history-element' to add more elements to
+ the list of defaults when it reaches the end of this list.  It does
+ this by calling a function defined by `minibuffer-default-add-function'.")
+ 
+ (make-variable-buffer-local 'minibuffer-default-add-done)
+ 
+ (defun minibuffer-default-add-completions ()
+   "Return a list of all completions without the default value.
+ This function is used to add all elements of the completion table to
+ the end of the list of defaults just after the default value."
+   (interactive)
+   (let ((def minibuffer-default)
+ 	(all (all-completions ""
+ 			      minibuffer-completion-table
+ 			      minibuffer-completion-predicate
+ 			      t)))
+     (if (listp def)
+ 	(append def all)
+       (cons def (delete def all)))))
+ 
  (defun goto-history-element (nabs)
    "Puts element of the minibuffer history in the minibuffer.
  The argument NABS specifies the absolute history position."
    (interactive "p")
+   (when (and (not minibuffer-default-add-done)
+ 	     (functionp minibuffer-default-add-function)
+ 	     (< nabs (- (if (listp minibuffer-default)
+ 			    (length minibuffer-default)
+ 			  1))))
+     (setq minibuffer-default-add-done t
+ 	  minibuffer-default (funcall minibuffer-default-add-function)))
    (let ((minimum (if minibuffer-default
  		     (- (if (listp minibuffer-default)
  			    (length minibuffer-default)
***************
*** 1320,1326 ****
  	      (minibuffer-contents-no-properties)))
      (if (< nabs minimum)
  	(if minibuffer-default
! 	    (error "End of history; no next item")
  	  (error "End of history; no default available")))
      (if (> nabs (length (symbol-value minibuffer-history-variable)))
  	(error "Beginning of history; no preceding item"))
--- 1360,1366 ----
  	      (minibuffer-contents-no-properties)))
      (if (< nabs minimum)
  	(if minibuffer-default
! 	    (error "End of defaults; no next item")
  	  (error "End of history; no default available")))
      (if (> nabs (length (symbol-value minibuffer-history-variable)))
  	(error "Beginning of history; no preceding item"))

Index: lisp/bindings.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/bindings.el,v
retrieving revision 1.202
diff -c -r1.202 bindings.el
*** lisp/bindings.el	26 Mar 2008 03:32:46 -0000	1.202
--- lisp/bindings.el	27 Mar 2008 23:35:36 -0000
***************
*** 650,656 ****
  	"^No later matching history item$"
  	"^No earlier matching history item$"
  	"^End of history; no default available$"
! 	"^End of history; no next item$"
  	"^Beginning of history; no preceding item$"
  	"^No recursive edit is in progress$"
  	"^Changes to be undone are outside visible portion of buffer$"
--- 650,656 ----
  	"^No later matching history item$"
  	"^No earlier matching history item$"
  	"^End of history; no default available$"
! 	"^End of defaults; no next item$"
  	"^Beginning of history; no preceding item$"
  	"^No recursive edit is in progress$"
  	"^Changes to be undone are outside visible portion of buffer$"

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

Re: C-r and C-s in minibuffer should search completion

[Xavier Maillard]

   But really there is no need to mix two separate search spaces.
   There exist already a convenient key sequence to search the completion
   table: `PgUp C-s'.  PgUp switches to the *Completions* buffer and
   C-s starts searching among completion candidates.

Gotcha ! I now understand why sometimes there is this
*Completion* buffer that pops up when I do something wrong on my
tiny keyboard :) Thank you for the tip !

	Xavier
-- 
http://www.gnu.org
http://www.april.org
http://www.lolica.org

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> In the patch below I renamed minibuffer-default-add-p into
> minibuffer-default-add-done, changed its default to nil, and
> removed make-variable-buffer-local for minibuffer-default-add-function.

Thanks, please install it.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>    But really there is no need to mix two separate search spaces.
>    There exist already a convenient key sequence to search the completion
>    table: `PgUp C-s'.  PgUp switches to the *Completions* buffer and
>    C-s starts searching among completion candidates.
>
> Gotcha ! I now understand why sometimes there is this
> *Completion* buffer that pops up when I do something wrong on my
> tiny keyboard :) Thank you for the tip !

No wonder that you sometimes accidentally get the *Completions* buffer.
Normally <prior> is bound to previous-history-element, but in some
minibuffers it is bound to switch-to-completions.  Maybe we should
avoid this inconsistency by using another key for switch-to-completions?

What about binding switch-to-completions to <M-down> because typing
simply <down> starts moving to the next item in a list of defaults
and completions. And also to bind a new command switch-to-history
(that will display a list of history items in a separate window)
to <M-up> because simply <up> moves to the previous history item.

Note that these keys are already used in many other programs to open
a pop-up list of history/completion items, so they will do what a typical
user would expect.

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

RE: C-r and C-s in minibuffer should search completion

[Drew Adams]

> Normally <prior> is bound to previous-history-element, but in some
> minibuffers it is bound to switch-to-completions.

Yes, that is poor design, IMO. Not the fact that it has a different meaning in
different minibuffer maps (completion vs no completion), but the fact that the
`prior' key is naturally associated with the `next' key, and Emacs does not
exploit that association - it breaks it.

> Maybe we should avoid this inconsistency by using
> another key for switch-to-completions?

FWIW, Icicles uses `C-insert' for this. And hitting it again takes you back to
the minibuffer (and copies the completion at point in *Completions* to the
minibuffer` for editing). That key won't work for emacs -nw, but `M-insert'
presumably would (in the form of `ESC insert'). 

(Icicles uses `insert' itself for something else. And Emacs likewise should
perhaps reserve `insert' for something that needs to be quick. Switching buffers
need not be so quick as to deserve a single key.)

> What about binding switch-to-completions to <M-down> because typing
> simply <down> starts moving to the next item in a list of defaults
> and completions. And also to bind a new command switch-to-history
> (that will display a list of history items in a separate window)
> to <M-up> because simply <up> moves to the previous history item.

Your association is a bit far-fetched, IMO - you are stretching things to fit
your proposal, rather than proposing something that seems natural.

There is little connection between the idea of having `down' and `up' move to
the next and previous items in a series and the idea of switching
windows/buffers.

Other things being equal: Keys that form natural pairs, such as `down'/`up',
`next'/`prior', and `home'/`end' should be reserved for things that perform some
sort of succession (i.e., next or previous) behavior. Or for some sort of
bracketing (e.g. grouping) behavior, such as start-something...end-something. Or
for some sort of complementary behavior (e.g. `insert'/`delete'). The same is
true for other such natural pairs of keys: (), [], {}, <>.

Succession is a natural choice for pairs, such as `next'/`prior', whose names
suggest succession. Opposite behavior can be a natural choice for pairs, such as
`insert'/`delete', whose names suggest that.

Finally, why use a separate window for history items? Why not simply use
*Completions* and let users complete against history items? IOW, use the normal
completion mechanism, with everything that it provides.

That is what I suggested when I mentioned `M-o', which is the key that Icicles
uses for this. You can use `M-o' from any minibuffer, not just during
completion. It uses a recursive minibuffer to let you choose from the current
input history (using completion) - your choice is inserted in the minibuffer,
where you can edit it for the original minibuffer reading.

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> Maybe we should avoid this inconsistency by using
>> another key for switch-to-completions?
>
> FWIW, Icicles uses `C-insert' for this. And hitting it again takes you back to
> the minibuffer (and copies the completion at point in *Completions* to the
> minibuffer` for editing). That key won't work for emacs -nw, but `M-insert'
> presumably would (in the form of `ESC insert').

I agree that `M-insert' makes sense for inserting the completion from
the *Completions* buffer to the minibuffer.

> There is little connection between the idea of having `down' and `up' move to
> the next and previous items in a series and the idea of switching
> windows/buffers.

<M-down> and <M-up> are used by many other programs that open a pop-up
list of history/completion items.  So they are natural key for most users
and especially for newbies.

I also suggest using <M-down> for IDE-like autocompletion that opens
a pop-up window with completion in a program buffer.  This is usually
bound to C-SPC in IDE like Eclipse, but C-SPC is already taken for
set-mark-command in Emacs.

> Finally, why use a separate window for history items? Why not simply use
> *Completions* and let users complete against history items? IOW, use the normal
> completion mechanism, with everything that it provides.

As there exists already a key that open the *Completions* buffer,
it makes sense to do the same for the history list.

Below is a patch that implements these ideas.

> That is what I suggested when I mentioned `M-o', which is the key that Icicles
> uses for this. You can use `M-o' from any minibuffer, not just during
> completion. It uses a recursive minibuffer to let you choose from the current
> input history (using completion) - your choice is inserted in the minibuffer,
> where you can edit it for the original minibuffer reading.

You earlier objected to using the same keys that have different
bindings in the minibuffer and outside it (cf. using `TAB' in the
minibuffer).  Since by default M-o is a prefix key for font-lock-fontify
function, I think we should find another key.  One candidate I propose
is C-M-TAB.

Index: lisp/simple.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/simple.el,v
retrieving revision 1.913
diff -c -r1.913 simple.el
*** lisp/simple.el	29 Mar 2008 22:56:17 -0000	1.913
--- lisp/simple.el	30 Mar 2008 00:38:48 -0000
***************
*** 5520,5525 ****
--- 5574,5604 ----
        (goto-char (point-min))
        (search-forward "\n\n" nil t)
        (forward-line 1))))
+ 
+ (defun switch-to-history-completions ()
+   "Select the completion list window for elements of the minibuffer history."
+   (interactive)
+   (let ((minibuffer-completion-table
+          (symbol-value minibuffer-history-variable))
+ 	(minibuffer-completion-predicate nil))
+     (switch-to-completions)))
+ 
+ (define-key minibuffer-local-completion-map [(meta down)] 'switch-to-completions)
+ (define-key minibuffer-local-completion-map [(meta up)] 'switch-to-history-completions)
+ 
+ (defun minibuffer-history-complete ()
+   "Complete the minibuffer input against the history list.
+ Locally bind the current value of `minibuffer-history-variable'
+ to the completion table `minibuffer-completion-table' before
+ calling `minibuffer-complete'."
+   (interactive)
+   (let ((minibuffer-completion-table
+          (symbol-value minibuffer-history-variable))
+         (minibuffer-completion-predicate nil))
+     (minibuffer-complete)))
+ 
+ (define-key minibuffer-local-map [(meta control tab)] 'minibuffer-history-complete)
+ 
  \f
  ;;; Support keyboard commands to turn on various modifiers.

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

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> In the patch below I renamed minibuffer-default-add-p into
>> minibuffer-default-add-done, changed its default to nil, and
>> removed make-variable-buffer-local for minibuffer-default-add-function.
>
> Thanks, please install it.

Done.

I also reverted the recent change in `call-interactively' that adds
buffers to the list of defaults.

However, there is one problem now: `C-x b M-n M-n' puts in the minibuffer
the name of the current buffer because the current buffer is in the front
of the buffer list acquired from the completion table in read-buffer.

But it makes no sense to switch to the current buffer, so the case 'B'
in `call-interactively' specially removes the current buffer from the
default value by using `Fother_buffer' specially for the command
`switch-to-buffer'.

I see no other way of removing the current buffer from the list for M-n
than using the following patch for `minibuffer-default-add-completions'.
It seems there is no other suitable place to remove the current buffer
in the call chain: `C-x b' -> `call-interactively' -> `read-buffer' ->
`completing-read' with `internal-complete-buffer' ->
`minibuffer-default-add-completions'.  This change should also work
for all commands that read a buffer name because when they need the
current buffer, they usually put the current buffer to the default value.
So it would be ok to remove it from the list.

Index: lisp/simple.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/simple.el,v
retrieving revision 1.913
diff -c -r1.913 simple.el
*** lisp/simple.el	29 Mar 2008 22:56:17 -0000	1.913
--- lisp/simple.el	30 Mar 2008 00:43:14 -0000
***************
*** 1332,1337 ****
--- 1333,1343 ----
  			      minibuffer-completion-table
  			      minibuffer-completion-predicate
  			      t)))
+ 
+     ;; Remove the current buffer from the list of buffer completions
+     (when (eq minibuffer-completion-table 'internal-complete-buffer)
+       (setq all (cdr all)))
+ 
      (if (listp def)
  	(append def all)
        (cons def (delete def all)))))

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

minibuffer-default-add-shell-commands (was: C-r and C-s in minibuffer should search completion)

[Juri Linkov]

Since minibuffer-default-add-function is now in CVS, it is possible
to install a feature approved by Richard in
http://lists.gnu.org/archive/html/emacs-devel/2007-12/msg00179.html

It adds commands from mailcap to the list of defaults of M-!.  I renamed
`dired-read-shell-command-default' to `mailcap-file-default-commands' as
Richard suggested, and moved it to mailcap.el (a question of moving the
file mailcap.el out of the gnus subdirectory should be decided by Gnus
maintainers).

Index: lisp/simple.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/simple.el,v
retrieving revision 1.913
diff -c -r1.913 simple.el
*** lisp/simple.el	29 Mar 2008 22:56:17 -0000	1.913
--- lisp/simple.el	30 Mar 2008 00:44:01 -0000
***************
*** 1960,1965 ****
--- 1990,2014 ----
  is run interactively.  A value of nil means that output to stderr and
  stdout will be intermixed in the output stream.")
  
+ (declare-function mailcap-file-default-commands "mailcap" (files))
+ 
+ (defun minibuffer-default-add-shell-commands ()
+   "Return a list of all commands associted with the current file.
+ This function is used to add all related commands retieved by `mailcap'
+ to the end of the list of defaults just after the default value."
+   (interactive)
+   (let* ((filename (if (listp minibuffer-default)
+ 		       (car minibuffer-default)
+ 		     minibuffer-default))
+ 	 (commands (and filename (require 'mailcap nil t)
+ 			(mailcap-file-default-commands (list filename)))))
+     (setq commands (mapcar (lambda (command)
+ 			     (concat command " " filename))
+ 			   commands))
+     (if (listp minibuffer-default)
+ 	(append minibuffer-default commands)
+       (cons minibuffer-default commands))))
+ 
  (defun minibuffer-complete-shell-command ()
    "Dynamically complete shell command at point."
    (interactive)
***************
*** 2034,2042 ****
  In an interactive call, the variable `shell-command-default-error-buffer'
  specifies the value of ERROR-BUFFER."
  
!   (interactive (list (read-shell-command "Shell command: ")
! 		     current-prefix-arg
! 		     shell-command-default-error-buffer))
    ;; Look for a handler in case default-directory is a remote file name.
    (let ((handler
  	 (find-file-name-handler (directory-file-name default-directory)
--- 2083,2096 ----
  In an interactive call, the variable `shell-command-default-error-buffer'
  specifies the value of ERROR-BUFFER."
  
!   (interactive
!    (let ((minibuffer-default-add-function
! 	  'minibuffer-default-add-shell-commands))
!      (list (read-shell-command "Shell command: " nil nil
! 			       (and buffer-file-name
! 				    (file-relative-name buffer-file-name)))
! 	   current-prefix-arg
! 	   shell-command-default-error-buffer)))
    ;; Look for a handler in case default-directory is a remote file name.
    (let ((handler
  	 (find-file-name-handler (directory-file-name default-directory)

Index: lisp/dired-aux.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/dired-aux.el,v
retrieving revision 1.164
diff -c -r1.164 dired-aux.el
*** lisp/dired-aux.el	26 Mar 2008 03:34:06 -0000	1.164
--- lisp/dired-aux.el	30 Mar 2008 00:44:16 -0000
***************
*** 464,530 ****
  \f
  ;;; Shell commands
  
! (declare-function mailcap-parse-mailcaps "mailcap" (&optional path force))
! (declare-function mailcap-parse-mimetypes "mailcap" (&optional path force))
! (declare-function mailcap-extension-to-mime "mailcap" (extn))
! (declare-function mailcap-mime-info "mailcap"
!                   (string &optional request no-decode))
! 
! (defun dired-read-shell-command-default (files)
!   "Return a list of default commands for `dired-read-shell-command'."
!   (require 'mailcap)
!   (mailcap-parse-mailcaps)
!   (mailcap-parse-mimetypes)
!   (let* ((all-mime-type
! 	  ;; All unique MIME types from file extensions
! 	  (delete-dups (mapcar (lambda (file)
! 				 (mailcap-extension-to-mime
! 				  (file-name-extension file t)))
! 			       files)))
! 	 (all-mime-info
! 	  ;; All MIME info lists
! 	  (delete-dups (mapcar (lambda (mime-type)
! 				 (mailcap-mime-info mime-type 'all))
! 			       all-mime-type)))
! 	 (common-mime-info
! 	  ;; Intersection of mime-infos from different mime-types;
! 	  ;; or just the first MIME info for a single MIME type
! 	  (if (cdr all-mime-info)
! 	      (delq nil (mapcar (lambda (mi1)
! 				  (unless (memq nil (mapcar
! 						     (lambda (mi2)
! 						       (member mi1 mi2))
! 						     (cdr all-mime-info)))
! 				    mi1))
! 				(car all-mime-info)))
! 	    (car all-mime-info)))
! 	 (commands
! 	  ;; Command strings from `viewer' field of the MIME info
! 	  (delq nil (mapcar (lambda (mime-info)
! 			      (let ((command (cdr (assoc 'viewer mime-info))))
! 				(if (stringp command)
! 				    (replace-regexp-in-string
! 				     ;; Replace mailcap's `%s' placeholder
! 				     ;; with dired's `?' placeholder
! 				     "%s" "?"
! 				     (replace-regexp-in-string
! 				      ;; Remove the final filename placeholder
! 				      "\s*\\('\\)?%s\\1?\s*\\'" "" command nil t)
! 				     nil t))))
! 			    common-mime-info))))
!     commands))
  
  (defun dired-read-shell-command (prompt arg files)
  ;;  "Read a dired shell command prompting with PROMPT (using read-string).
  ;;ARG is the prefix arg and may be used to indicate in the prompt which
  ;;  files are affected.
  ;;This is an extra function so that you can redefine it, e.g., to use gmhist."
!   (dired-mark-pop-up
!    nil 'shell files
!    (function read-string)
!    (format prompt (dired-mark-prompt arg files))
!    nil 'shell-command-history
!    (dired-read-shell-command-default files)))
  
  ;; The in-background argument is only needed in Emacs 18 where
  ;; shell-command doesn't understand an appended ampersand `&'.
--- 464,494 ----
  \f
  ;;; Shell commands
  
! (declare-function mailcap-file-default-commands "mailcap" (files))
! 
! (defun minibuffer-default-add-dired-shell-commands ()
!   "Return a list of all commands associted with current dired files.
! This function is used to add all related commands retieved by `mailcap'
! to the end of the list of defaults just after the default value."
!   (interactive)
!   (let ((commands (and (boundp 'files) (require 'mailcap nil t)
! 		       (mailcap-file-default-commands files))))
!     (if (listp minibuffer-default)
! 	(append minibuffer-default commands)
!       (cons minibuffer-default commands))))
  
  (defun dired-read-shell-command (prompt arg files)
  ;;  "Read a dired shell command prompting with PROMPT (using read-string).
  ;;ARG is the prefix arg and may be used to indicate in the prompt which
  ;;  files are affected.
  ;;This is an extra function so that you can redefine it, e.g., to use gmhist."
!   (let ((minibuffer-default-add-function
! 	 'minibuffer-default-add-dired-shell-commands))
!     (dired-mark-pop-up
!      nil 'shell files
!      (function read-string)
!      (format prompt (dired-mark-prompt arg files))
!      nil 'shell-command-history)))
  
  ;; The in-background argument is only needed in Emacs 18 where
  ;; shell-command doesn't understand an appended ampersand `&'.

Index: lisp/gnus/mailcap.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/gnus/mailcap.el,v
retrieving revision 1.21
diff -c -r1.21 mailcap.el
*** lisp/gnus/mailcap.el	8 Jan 2008 20:45:19 -0000	1.21
--- lisp/gnus/mailcap.el	30 Mar 2008 00:44:07 -0000
***************
*** 1007,1012 ****
--- 1007,1059 ----
  	       (cdr l))))
        mailcap-mime-data)))))
  
+ ;;;
+ ;;; Useful functions
+ ;;;
+ 
+ (defun mailcap-file-default-commands (files)
+   "Return a list of default commands for FILES."
+   (mailcap-parse-mailcaps)
+   (mailcap-parse-mimetypes)
+   (let* ((all-mime-type
+ 	  ;; All unique MIME types from file extensions
+ 	  (delete-dups (mapcar (lambda (file)
+ 				 (mailcap-extension-to-mime
+ 				  (file-name-extension file t)))
+ 			       files)))
+ 	 (all-mime-info
+ 	  ;; All MIME info lists
+ 	  (delete-dups (mapcar (lambda (mime-type)
+ 				 (mailcap-mime-info mime-type 'all))
+ 			       all-mime-type)))
+ 	 (common-mime-info
+ 	  ;; Intersection of mime-infos from different mime-types;
+ 	  ;; or just the first MIME info for a single MIME type
+ 	  (if (cdr all-mime-info)
+ 	      (delq nil (mapcar (lambda (mi1)
+ 				  (unless (memq nil (mapcar
+ 						     (lambda (mi2)
+ 						       (member mi1 mi2))
+ 						     (cdr all-mime-info)))
+ 				    mi1))
+ 				(car all-mime-info)))
+ 	    (car all-mime-info)))
+ 	 (commands
+ 	  ;; Command strings from `viewer' field of the MIME info
+ 	  (delq nil (mapcar (lambda (mime-info)
+ 			      (let ((command (cdr (assoc 'viewer mime-info))))
+ 				(if (stringp command)
+ 				    (replace-regexp-in-string
+ 				     ;; Replace mailcap's `%s' placeholder
+ 				     ;; with dired's `?' placeholder
+ 				     "%s" "?"
+ 				     (replace-regexp-in-string
+ 				      ;; Remove the final filename placeholder
+ 				      "\s*\\('\\)?%s\\1?\s*\\'" "" command nil t)
+ 				     nil t))))
+ 			    common-mime-info))))
+     commands))
+ 
  (provide 'mailcap)
  
  ;;; arch-tag: 1fd4f9c9-c305-4d2e-9747-3a4d45baa0bd

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

RE: C-r and C-s in minibuffer should search completion

[Drew Adams]

> >> Maybe we should avoid this inconsistency by using
> >> another key for switch-to-completions?
> >
> > FWIW, Icicles uses `C-insert' for this. And hitting it 
> > again takes you back to the minibuffer (and copies the
> > completion at point in *Completions* to the
> > minibuffer` for editing). That key won't work for emacs 
> > -nw, but `M-insert' presumably would (in the form of
> > `ESC insert').
> 
> I agree that `M-insert' makes sense for inserting the completion from
> the *Completions* buffer to the minibuffer.

It makes sense to use the same key to go both directions (minibuffer <-->
*Completions*), as I mentioned. Why not?

> > There is little connection between the idea of having 
> > `down' and `up' move to the next and previous items in
> > a series and the idea of switching windows/buffers.
> 
> <M-down> and <M-up> are used by many other programs that open a pop-up
> list of history/completion items.  So they are natural key 
> for most users and especially for newbies.

As I said, there is little connection with the meaning of "next" and "previous".
That's the point.

Slavishly copying other programs is not a sufficient argument, especially when
it comes to key bindings. It can be a secondary argument, but it shouldn't be
the only or the primary argument. Otherwise, Emacs would quickly resemble the
lowest common denominator, or worse. Other programs are for the most part
inferior to Emacs wrt key bindings.

Emacs generally chooses key bindings in a reasoned way. I gave arguments why
down/up should be reserved for successor/predecessor operations, if possible.
You've given no argument for using them for something that is unrelated to that,
beyond saying that some other programs use them to open a popup list.

There is no reason to have two different keys for switching to *Completions*,
which is what you are proposing, AFAICT. If there is already a key that switches
to *Completions*, then why can't users also use it to switch to *Completions*
when it contains the history list? 

Why do they need another key for that? That's like saying that although we
already have `C-f' for forward-char, we should add `C-<whatever>' for
forward-char whenever the buffer contains lines of poetry.

> I also suggest using <M-down> for IDE-like autocompletion that opens
> a pop-up window with completion in a program buffer.  This is usually
> bound to C-SPC in IDE like Eclipse, but C-SPC is already taken for
> set-mark-command in Emacs.
> 
> > Finally, why use a separate window for history items? Why 
> > not simply use *Completions* and let users complete against
> > history items? IOW, use the normal completion mechanism,
> > with everything that it provides.
> 
> As there exists already a key that open the *Completions* buffer,
> it makes sense to do the same for the history list.

Makes sense to do what "the same"? Add another key, just for switching to
*Completions* whenever it contains a history list? Why not really "do the same
for the history list" - treat it exactly the same as every other list in
*Completions*? When you say "same" it seems you mean "different".

Unless I'm missing something, all you've done is fill *Completions* with a set
of completions that come from the history list. That's useful, but there is
nothing special about it. Completing previous inputs is no different from
completing command names or file names or buffer names or anything else. Why add
key bindings unnecessarily (Occam)?

You need one minibuffer key binding to let users complete against the history
list; that's all.

> Below is a patch that implements these ideas.
> 
> > That is what I suggested when I mentioned `M-o', which is 
> > the key that Icicles uses for this. You can use `M-o' from
> > any minibuffer, not just during completion. It uses a
> > recursive minibuffer to let you choose from the current
> > input history (using completion) - your choice is inserted 
> > in the minibuffer, where you can edit it for the original
> > minibuffer reading.
> 
> You earlier objected to using the same keys that have different
> bindings in the minibuffer and outside it (cf. using `TAB' in the
> minibuffer).

Because the indentation binding of TAB is useful in the minibuffer. I made a
specific argument about TAB. You cannot generalize from my opposition to
changing the meaning of TAB in the minibuffer to suppose that I oppose the reuse
of any keys by giving them different minibuffer bindings.

And TAB is a particular case in another sense: it is apparently in demand for
everything under the sun. I argued to reserve it for things that are related to
its existing meanings.

> Since by default M-o is a prefix key for font-lock-fontify
> function, I think we should find another key.

I chose `M-o' precisely because its global binding is not useful in the
minibuffer. (Do you think it is?) This is not about blanket rules; you have to
consider the details - TAB is not `M-o'.

But I don't care whether you use `M-o' or not. It was just a suggestion. I do
care that you not use TAB for that, with or without modifiers.

> One candidate I propose is C-M-TAB.

When will you guys leave TAB alone? Save M-TAB, S-TAB, C-M-TAB, C-S-TAB, and so
on for operations that are more closely related to TAB's existing behaviors (and
there are already enough of those!). Yes, you could make an argument that this
behavior is slightly related, but tomorrow we will find 23 other possible uses
for TAB that are all more closely related.

I also suggested that completion against the current history use a recursive
minibuffer, and that the choice be inserted into the parent minibuffer for
editing. That is useful, especially for minibuffer input that does not involve
completion. You don't want to be limited to completing the current input against
a previous input. You might want to insert more than one previous input (using
completion), and edit the combination.

I don't agree with any of the proposed changes, but I have no illusions that my
disagreement will make any difference. On n'arrete pas le "progres".

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> Done.

> I also reverted the recent change in `call-interactively' that adds
> buffers to the list of defaults.

Thanks.

> However, there is one problem now: `C-x b M-n M-n' puts in the minibuffer
> the name of the current buffer because the current buffer is in the front
> of the buffer list acquired from the completion table in read-buffer.

This is a very minor problem.

> But it makes no sense to switch to the current buffer, so the case 'B'

Then switch-to-buffer should use a completion predicate that rules out
the current-buffer.  And the history functions should pay attention to it.

> ***************
> *** 1332,1337 ****
> --- 1333,1343 ----
>   			      minibuffer-completion-table
>   			      minibuffer-completion-predicate
>   			      t)))
> + 
> +     ;; Remove the current buffer from the list of buffer completions
> +     (when (eq minibuffer-completion-table 'internal-complete-buffer)
> +       (setq all (cdr all)))
> + 
>       (if (listp def)
>   	(append def all)
>         (cons def (delete def all)))))

No, this is a hack that will surely bite us down the road.  The minor
problem it's trying to fix does not justify breaking abstraction in
this way.


        Stefan

Re: minibuffer-default-add-shell-commands

[Stefan Monnier]

> !   (interactive
> !    (let ((minibuffer-default-add-function
> ! 	  'minibuffer-default-add-shell-commands))
> !      (list (read-shell-command "Shell command: " nil nil
> ! 			       (and buffer-file-name
> ! 				    (file-relative-name buffer-file-name)))
> ! 	   current-prefix-arg
> ! 	   shell-command-default-error-buffer)))

That means that whenever if I ever leave a M-! prompt open and go do
something else, all my completions will now use
minibuffer-default-add-shell-commands?  That's not right.
Let-binding is not right for minibuffer-default-add-function.


        Stefan

Re: minibuffer-default-add-shell-commands

[Reiner Steib]

On Sun, Mar 30 2008, Juri Linkov wrote:

> I renamed `dired-read-shell-command-default' to
> `mailcap-file-default-commands' as Richard suggested, and moved it
> to mailcap.el (a question of moving the file mailcap.el out of the
> gnus subdirectory should be decided by Gnus maintainers).

As long as no incompatibilities (Emacs 21+ and XEmacs 21.4+) are
introduced and syncing is preserved, I won't object.

[I removed Gnus dependencies from `mailcap.el' in December, see also
<http://thread.gmane.org/gmane.emacs.devel/80123/focus=84745>.]

Bye, Reiner.
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/

Re: minibuffer-default-add-shell-commands

[Juri Linkov]

[-- Attachment #1: Type: text/plain, Size: 376 bytes --]

> That means that whenever if I ever leave a M-! prompt open and go do
> something else, all my completions will now use
> minibuffer-default-add-shell-commands?  That's not right.
> Let-binding is not right for minibuffer-default-add-function.

This can be fixed by using `minibuffer-with-setup-hook'.
A diff below shows differences between my previous patch and a new one:


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Type: text/x-patch, Size: 1735 bytes --]

diff --git a/lisp/simple.el b/lisp/simple.el
index 06f1be9..516e475 100644
--- a/lisp/simple.el
+++ b/lisp/simple.el
@@ -2084,13 +2084,16 @@ (defun shell-command
 specifies the value of ERROR-BUFFER."
 
   (interactive
-   (let ((minibuffer-default-add-function
-	  'minibuffer-default-add-shell-commands))
-     (list (read-shell-command "Shell command: " nil nil
-			       (and buffer-file-name
-				    (file-relative-name buffer-file-name)))
-	   current-prefix-arg
-	   shell-command-default-error-buffer)))
+   (list
+    (minibuffer-with-setup-hook
+	(lambda ()
+	  (set (make-local-variable 'minibuffer-default-add-function)
+	       'minibuffer-default-add-shell-commands))
+      (read-shell-command "Shell command: " nil nil
+			  (and buffer-file-name
+			       (file-relative-name buffer-file-name))))
+    current-prefix-arg
+    shell-command-default-error-buffer))
   ;; Look for a handler in case default-directory is a remote file name.
   (let ((handler
 	 (find-file-name-handler (directory-file-name default-directory)
diff --git a/lisp/dired-aux.el b/lisp/dired-aux.el
index b8e2872..faf515d 100644
--- a/lisp/dired-aux.el
+++ b/lisp/dired-aux.el
@@ -482,8 +482,10 @@ (defun dired-read-shell-command
 ;;ARG is the prefix arg and may be used to indicate in the prompt which
 ;;  files are affected.
 ;;This is an extra function so that you can redefine it, e.g., to use gmhist."
-  (let ((minibuffer-default-add-function
-	 'minibuffer-default-add-dired-shell-commands))
+  (minibuffer-with-setup-hook
+      (lambda ()
+	(set (make-local-variable 'minibuffer-default-add-function)
+	     'minibuffer-default-add-dired-shell-commands))
     (dired-mark-pop-up
      nil 'shell files
      (function read-string)

[-- Attachment #3: Type: text/plain, Size: 45 bytes --]


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

Re: minibuffer-default-add-shell-commands

[Juri Linkov]

>> I renamed `dired-read-shell-command-default' to
>> `mailcap-file-default-commands' as Richard suggested, and moved it
>> to mailcap.el (a question of moving the file mailcap.el out of the
>> gnus subdirectory should be decided by Gnus maintainers).
>
> As long as no incompatibilities (Emacs 21+ and XEmacs 21.4+) are
> introduced and syncing is preserved, I won't object.

Do you see any incompatibilities?

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

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> However, there is one problem now: `C-x b M-n M-n' puts in the minibuffer
>> the name of the current buffer because the current buffer is in the front
>> of the buffer list acquired from the completion table in read-buffer.
>
> This is a very minor problem.

This is a usability problem for users who will prefer using `C-x b M-n M-n'
to switch buffers by recency.  And the current buffer will be annoying
in this list.

>> But it makes no sense to switch to the current buffer, so the case 'B'
>
> Then switch-to-buffer should use a completion predicate that rules out
> the current-buffer.  And the history functions should pay attention to it.

The completion list still should contain the current buffer for other
commands that read a buffer name and whose semantics may require
the presence of the current buffer in the completion list.

Since removing the current buffer from the list of defaults is necessary
only for three commands `C-x b', `C-x 4 b' and `C-x 5 b', I propose just
to check for their names before removing the current buffer from the list:

Index: lisp/simple.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/simple.el,v
retrieving revision 1.913
diff -c -r1.913 simple.el
*** lisp/simple.el	29 Mar 2008 22:56:17 -0000	1.913
--- lisp/simple.el	30 Mar 2008 18:32:11 -0000
***************
*** 1304,1309 ****
--- 1305,1318 ----
  
  (defvar minibuffer-temporary-goal-position nil)
  
+ (defvar minibuffer-original-command nil
+   "Name of the command that invoked this minibuffer.")
+ 
+ (add-hook 'minibuffer-setup-hook 'minibuffer-initialize)
+ 
+ (defun minibuffer-initialize ()
+   (set (make-local-variable 'minibuffer-original-command) this-command))
+ 
  (defvar minibuffer-default-add-function 'minibuffer-default-add-completions
    "Function run by `goto-history-element' before consuming `minibuffer-default'.
  This is useful to dynamically add more elements to the list `minibuffer-default'
***************
*** 1332,1337 ****
--- 1341,1354 ----
  			      minibuffer-completion-table
  			      minibuffer-completion-predicate
  			      t)))
+ 
+     ;; Remove the current buffer from the list of buffer completions
+     (when (member minibuffer-original-command
+ 		  '(switch-to-buffer
+ 		    switch-to-buffer-other-window
+ 		    switch-to-buffer-other-frame))
+       (setq all (cdr all)))
+ 
      (if (listp def)
  	(append def all)
        (cons def (delete def all)))))

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

Re: minibuffer-default-add-shell-commands

[Reiner Steib]

On Sun, Mar 30 2008, Juri Linkov wrote:

>>> (a question of moving the file mailcap.el out of the gnus
>>> subdirectory should be decided by Gnus maintainers).
>>
>> As long as no incompatibilities (Emacs 21+ and XEmacs 21.4+) are
>> introduced and syncing is preserved, I won't object.
>
> Do you see any incompatibilities?

Not yet.  But if people start to modify the code in a way that it only
works for Emacs 23 or remove compatibility code, synchronization and
maintenance would get more difficult.

Bye, Reiner.
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>>> However, there is one problem now: `C-x b M-n M-n' puts in the minibuffer
>>> the name of the current buffer because the current buffer is in the front
>>> of the buffer list acquired from the completion table in read-buffer.
>> 
>> This is a very minor problem.

> This is a usability problem for users who will prefer using `C-x b M-n M-n'
> to switch buffers by recency.  And the current buffer will be annoying
> in this list.

Again, a very minor problem.

>>> But it makes no sense to switch to the current buffer, so the case 'B'
>> 
>> Then switch-to-buffer should use a completion predicate that rules out
>> the current-buffer.  And the history functions should pay attention to it.

> The completion list still should contain the current buffer for other
> commands that read a buffer name and whose semantics may require
> the presence of the current buffer in the completion list.

Is that supposed to disagree with what I wrote?

> Since removing the current buffer from the list of defaults is necessary
> only for three commands `C-x b', `C-x 4 b' and `C-x 5 b', I propose just
> to check for their names before removing the current buffer from the list:

No, this is the wrong way to go about it.  The special code should be in
C-x b, C-x 4 b, and C-x 5 b.  Checking the caller inside the minibuffer
is *bad*.  I already suggested a much better solution: use
a minibuffer-completion-predicate.  That's what it's for.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

> The special code should be in C-x b, C-x 4 b, and C-x 5 b.
> Checking the caller inside the minibuffer is *bad*.  I already
> suggested a much better solution: use a minibuffer-completion-predicate.
> That's what it's for.

C-x b switch-to-buffer has an interactive spec in buffer.c
"BSwitch to buffer: ".  Do you mean replacing it with an
expression that uses `minibuffer-with-setup-hook' to bind
`minibuffer-completion-predicate' to a new predicate that
rules out the current-buffer?

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> The special code should be in C-x b, C-x 4 b, and C-x 5 b.
>> Checking the caller inside the minibuffer is *bad*.  I already
>> suggested a much better solution: use a minibuffer-completion-predicate.
>> That's what it's for.

> C-x b switch-to-buffer has an interactive spec in buffer.c
> "BSwitch to buffer: ".  Do you mean replacing it with an
> expression that uses `minibuffer-with-setup-hook' to bind
> `minibuffer-completion-predicate' to a new predicate that
> rules out the current-buffer?

Actually, looking at the code some more, I'd suggest to add
a `predicate' arg to `read-buffer' and then use it in switch-to-buffer
(which would clearly need to be changed not to use a string-spec but
a lisp-spec for its interactive spec).


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

> Actually, looking at the code some more, I'd suggest to add
> a `predicate' arg to `read-buffer' and then use it in switch-to-buffer
> (which would clearly need to be changed not to use a string-spec but
> a lisp-spec for its interactive spec).

Adding a new argument to `read-buffer' will cause the argument list
mismatch between `read-buffer' and `read-buffer-function', and
calling `read-buffer-function' with an additional argument will
break existing code.

But I now have an idea how to accomplish this task without adding a
new argument to `read-buffer'.  The patch below adds a new function
`read-buffer-to-switch' that binds `minibuffer-default-add-function' to
`minibuffer-default-add-buffers' that omits the current buffer from the
list of defaults before calling `read-buffer':

Index: src/buffer.c
===================================================================
RCS file: /sources/emacs/emacs/src/buffer.c,v
retrieving revision 1.557
diff -c -r1.557 buffer.c
*** src/buffer.c	2 Apr 2008 20:15:11 -0000	1.557
--- src/buffer.c	3 Apr 2008 22:56:56 -0000
***************
*** 1767,1773 ****
    return buf;
  }
  
! DEFUN ("switch-to-buffer", Fswitch_to_buffer, Sswitch_to_buffer, 1, 2, "BSwitch to buffer: ",
         doc: /* Select buffer BUFFER in the current window.
  If BUFFER does not identify an existing buffer,
  then this function creates a buffer with that name.
--- 1767,1774 ----
    return buf;
  }
  
! DEFUN ("switch-to-buffer", Fswitch_to_buffer, Sswitch_to_buffer, 1, 2,
!        "(list (read-buffer-to-switch \"Switch to buffer: \"))",
         doc: /* Select buffer BUFFER in the current window.
  If BUFFER does not identify an existing buffer,
  then this function creates a buffer with that name.

Index: lisp/files.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/files.el,v
retrieving revision 1.967
diff -c -r1.967 files.el
*** lisp/files.el	3 Apr 2008 20:06:45 -0000	1.967
--- lisp/files.el	3 Apr 2008 22:57:00 -0000
***************
*** 1033,1038 ****
--- 1033,1064 ----
      (rename-file encoded new-encoded ok-if-already-exists)
      newname))
  \f
+ (defun minibuffer-default-add-buffers ()
+   "Return a list of all buffer completions without the current buffer.
+ This function is used to add all elements of the buffer name completion
+ table to the end of the list of defaults just after the default value."
+   (interactive)
+   (let ((def minibuffer-default)
+ 	(all (all-completions ""
+ 			      minibuffer-completion-table
+ 			      minibuffer-completion-predicate
+ 			      t)))
+     ;; Remove the current buffer from the list of buffer completions
+     (setq all (cdr all))
+     (if (listp def)
+ 	(append def all)
+       (cons def (delete def all)))))
+ 
+ (defun read-buffer-to-switch (prompt)
+   "Read the name of a buffer to switch to and return as a string.
+ It is intended for `switch-to-buffer' family of commands since they
+ need to omit the name of current buffer from the list of defaults."
+   (minibuffer-with-setup-hook
+       (lambda ()
+ 	(set (make-local-variable 'minibuffer-default-add-function)
+ 	     'minibuffer-default-add-buffers))
+     (read-buffer prompt (other-buffer (current-buffer)))))
+ 
  (defun switch-to-buffer-other-window (buffer &optional norecord)
    "Select buffer BUFFER in another window.
  If BUFFER does not identify an existing buffer, then this function
***************
*** 1047,1053 ****
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive "BSwitch to buffer in other window: ")
    (let ((pop-up-windows t)
  	;; Don't let these interfere.
  	same-window-buffer-names same-window-regexps)
--- 1073,1080 ----
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive
!    (list (read-buffer-to-switch "Switch to buffer in other window: ")))
    (let ((pop-up-windows t)
  	;; Don't let these interfere.
  	same-window-buffer-names same-window-regexps)
***************
*** 1061,1067 ****
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive "BSwitch to buffer in other frame: ")
    (let ((pop-up-frames t)
  	same-window-buffer-names same-window-regexps)
      (prog1
--- 1088,1095 ----
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive
!    (list (read-buffer-to-switch "Switch to buffer in other frame: ")))
    (let ((pop-up-frames t)
  	same-window-buffer-names same-window-regexps)
      (prog1

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> Actually, looking at the code some more, I'd suggest to add
>> a `predicate' arg to `read-buffer' and then use it in switch-to-buffer
>> (which would clearly need to be changed not to use a string-spec but
>> a lisp-spec for its interactive spec).

> Adding a new argument to `read-buffer' will cause the argument list
> mismatch between `read-buffer' and `read-buffer-function', and
> calling `read-buffer-function' with an additional argument will
> break existing code.

Then lets fix the existing code.

> But I now have an idea how to accomplish this task without adding a
> new argument to `read-buffer'.  The patch below adds a new function
> `read-buffer-to-switch' that binds `minibuffer-default-add-function' to
> `minibuffer-default-add-buffers' that omits the current buffer from the
> list of defaults before calling `read-buffer':

That's ugly.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> Adding a new argument to `read-buffer' will cause the argument list
>> mismatch between `read-buffer' and `read-buffer-function', and
>> calling `read-buffer-function' with an additional argument will
>> break existing code.
>
> Then lets fix the existing code.

This means changing several places in Emacs and still breaking
compatibility with external packages, and all this hassle just
for fixing a minor problem.  I think this fix is disproportional
to the problem.

OTOH, my latest patch implements exactly what is needed: using your
suggestion to place code in the interactive spec of switch-buffer-*,
it also sets a special function minibuffer-default-add-function
that removes the current buffer from the list of defaults of
switch-buffer-* commands.  Thus, this is a very localized fix.
I spent enough time trying hard to find a better solution,
but I see no better.  I propose to install this fix and
move to more important things.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> OTOH, my latest patch implements exactly what is needed: using your
> suggestion to place code in the interactive spec of switch-buffer-*,
> it also sets a special function minibuffer-default-add-function
> that removes the current buffer from the list of defaults of
> switch-buffer-* commands.

Yes, that's an acceptable solution, but really if there are only
2 buffers, C-x b TAB should select the only other buffer, so the better
solution is to fix the completion table.

> Thus, this is a very localized fix.  I spent enough time trying hard
> to find a better solution, but I see no better.  I propose to install
> this fix and move to more important things.

Sure,


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

> Yes, that's an acceptable solution, but really if there are only
> 2 buffers, C-x b TAB should select the only other buffer, so the better
> solution is to fix the completion table.

On a par with your efforts to move minibuffer stuff to minibuffer.el,
I've ported the existing function `internal-complete-buffer' to Lisp
and created a new function `internal-complete-buffer-sans-current' to
remove the current buffer from the completion table for `C-x b' family
of commands.  I also considered using `complete-with-action' internally
in `internal-complete-buffer' since it uses the same code pattern, but
`complete-with-action' doesn't call `all-completions' with the non-nil
arg `hide-spaces'.  Below is a complete patch that causes `C-x b TAB'
to omit the current buffer from the completion table:

Index: src/minibuf.c
===================================================================
RCS file: /sources/emacs/emacs/src/minibuf.c,v
retrieving revision 1.345
diff -c -r1.345 minibuf.c
*** src/minibuf.c	9 Apr 2008 19:33:51 -0000	1.345
--- src/minibuf.c	15 Apr 2008 22:27:45 -0000
***************
*** 1944,1967 ****
      return Qt;
  }
  
- DEFUN ("internal-complete-buffer", Finternal_complete_buffer, Sinternal_complete_buffer, 3, 3, 0,
-        doc: /* Perform completion on buffer names.
- If the argument FLAG is nil, invoke `try-completion', if it's t, invoke
- `all-completions', otherwise invoke `test-completion'.
- 
- The arguments STRING and PREDICATE are as in `try-completion',
- `all-completions', and `test-completion'. */)
-      (string, predicate, flag)
-      Lisp_Object string, predicate, flag;
- {
-   if (NILP (flag))
-     return Ftry_completion (string, Vbuffer_alist, predicate);
-   else if (EQ (flag, Qt))
-     return Fall_completions (string, Vbuffer_alist, predicate, Qt);
-   else				/* assume `lambda' */
-     return Ftest_completion (string, Vbuffer_alist, predicate);
- }
- 
  /* Like assoc but assumes KEY is a string, and ignores case if appropriate.  */
  
  DEFUN ("assoc-string", Fassoc_string, Sassoc_string, 2, 3, 0,
--- 1944,1949 ----
***************
*** 2220,2226 ****
    defsubr (&Sread_string);
    defsubr (&Sread_command);
    defsubr (&Sread_variable);
-   defsubr (&Sinternal_complete_buffer);
    defsubr (&Sread_buffer);
    defsubr (&Sread_no_blanks_input);
    defsubr (&Sminibuffer_depth);
--- 2202,2207 ----

Index: lisp/minibuffer.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/minibuffer.el,v
retrieving revision 1.13
diff -c -r1.13 minibuffer.el
*** lisp/minibuffer.el	14 Apr 2008 22:04:38 -0000	1.13
--- lisp/minibuffer.el	15 Apr 2008 22:27:03 -0000
***************
*** 730,734 ****
--- 730,758 ----
                              'completion--file-name-table)
    "Internal subroutine for `read-file-name'.  Do not call this.")
  
+ (defun internal-complete-buffer (string predicate flag)
+   "Perform completion on buffer names.
+ If the argument FLAG is nil, invoke `try-completion', if it's t, invoke
+ `all-completions', otherwise invoke `test-completion'.
+ 
+ The arguments STRING and PREDICATE are as in `try-completion',
+ `all-completions', and `test-completion'."
+   (let ((table (mapcar 'buffer-name (buffer-list))))
+     (cond
+      ((null flag) (try-completion string table predicate))
+      ((eq flag t) (all-completions string table predicate t))
+      (t (test-completion string table predicate)))))
+ 
+ (defun internal-complete-buffer-sans-current (string predicate flag)
+   "Perform completion on buffer names excluding the current buffer.
+ Like `internal-complete-buffer', but removes the current buffer from
+ the completion list."
+   (let ((table (mapcar 'buffer-name (delq (other-buffer (current-buffer) t)
+ 					  (buffer-list)))))
+     (cond
+      ((null flag) (try-completion string table predicate))
+      ((eq flag t) (all-completions string table predicate t))
+      (t (test-completion string table predicate)))))
+ 
  (provide 'minibuffer)
  ;;; minibuffer.el ends here

Index: lisp/files.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/files.el,v
retrieving revision 1.970
diff -c -r1.970 files.el
*** lisp/files.el	12 Apr 2008 03:14:56 -0000	1.970
--- lisp/files.el	15 Apr 2008 22:27:11 -0000
***************
*** 1033,1038 ****
--- 1033,1048 ----
      (rename-file encoded new-encoded ok-if-already-exists)
      newname))
  \f
+ (defun read-buffer-to-switch (prompt)
+   "Read the name of a buffer to switch to and return as a string.
+ It is intended for `switch-to-buffer' family of commands since they
+ need to omit the name of current buffer from the list of defaults."
+   (minibuffer-with-setup-hook
+       (lambda ()
+ 	(set (make-local-variable 'minibuffer-completion-table)
+ 	     'internal-complete-buffer-sans-current))
+     (read-buffer prompt (other-buffer (current-buffer)))))
+ 
  (defun switch-to-buffer-other-window (buffer &optional norecord)
    "Select buffer BUFFER in another window.
  If BUFFER does not identify an existing buffer, then this function
***************
*** 1047,1053 ****
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive "BSwitch to buffer in other window: ")
    (let ((pop-up-windows t)
  	;; Don't let these interfere.
  	same-window-buffer-names same-window-regexps)
--- 1057,1064 ----
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive
!    (list (read-buffer-to-switch "Switch to buffer in other window: ")))
    (let ((pop-up-windows t)
  	;; Don't let these interfere.
  	same-window-buffer-names same-window-regexps)
***************
*** 1061,1067 ****
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive "BSwitch to buffer in other frame: ")
    (let ((pop-up-frames t)
  	same-window-buffer-names same-window-regexps)
      (prog1
--- 1072,1079 ----
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive
!    (list (read-buffer-to-switch "Switch to buffer in other frame: ")))
    (let ((pop-up-frames t)
  	same-window-buffer-names same-window-regexps)
      (prog1

Index: src/buffer.c
===================================================================
RCS file: /sources/emacs/emacs/src/buffer.c,v
retrieving revision 1.557
diff -c -r1.557 buffer.c
*** src/buffer.c	2 Apr 2008 20:15:11 -0000	1.557
--- src/buffer.c	15 Apr 2008 22:27:01 -0000
***************
*** 1767,1773 ****
    return buf;
  }
  
! DEFUN ("switch-to-buffer", Fswitch_to_buffer, Sswitch_to_buffer, 1, 2, "BSwitch to buffer: ",
         doc: /* Select buffer BUFFER in the current window.
  If BUFFER does not identify an existing buffer,
  then this function creates a buffer with that name.
--- 1767,1774 ----
    return buf;
  }
  
! DEFUN ("switch-to-buffer", Fswitch_to_buffer, Sswitch_to_buffer, 1, 2,
!        "(list (read-buffer-to-switch \"Switch to buffer: \"))",
         doc: /* Select buffer BUFFER in the current window.
  If BUFFER does not identify an existing buffer,
  then this function creates a buffer with that name.



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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> + (defun internal-complete-buffer (string predicate flag)
> +   (let ((table (mapcar 'buffer-name (buffer-list))))

Why not just use (mapcar 'buffer-name (buffer-list)) rather than using
a function as the completion table?

If it's because of the hide-spaces argument to all-completions, then
please write a comment about it,


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> + (defun internal-complete-buffer (string predicate flag)
>> +   (let ((table (mapcar 'buffer-name (buffer-list))))
>
> Why not just use (mapcar 'buffer-name (buffer-list)) rather than using
> a function as the completion table?
>
> If it's because of the hide-spaces argument to all-completions, then
> please write a comment about it,

The reason of adding the C version of internal-complete-buffer was the
dynamic nature of the buffer list.  This is described in detail in
http://thread.gmane.org/gmane.emacs.devel/58453

  Date: Wed, 16 Aug 2006 13:55:32 -0700 (PDT)
  From: "Stuart D. Herring" <herring@lanl.gov>
  To: emacs-devel@gnu.org

  Fread_buffer directly uses Vbuffer_alist in its call to Fcompleting_read.
  This breaks if, while the minibuffer is active, the user switches buffers
  around.  The cons which was the head of the list when Fread_buffer was
  called will in general no longer be at the head, and so part of the list
  will vanish for completion purposes.  (Of course, it's slightly worse when
  REQUIRE-MATCH is set.)  Similar problems result from creating or killing
  buffers during the call.  I also have a vague suspicion that this
  procedure exposes the alist to modification by user code, but I can't
  think of how at the moment.

  Copying the alist for the call to Fcompleting_read (which is trivial to
  implement) would solve the reordering problems but not the ones involving
  creation/destruction; the real solution is to write a `complete-buffer'
  completion function that would re-consult the buffer list each time
  completion was needed.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> The reason of adding the C version of internal-complete-buffer was the
> dynamic nature of the buffer list.  This is described in detail in
> http://thread.gmane.org/gmane.emacs.devel/58453

Oh, that's right.  So I suggest to leave the internal-complete-buffer in
its C form (it works well and makes sense there since it accesses the
internal Vbuffer_list variable).  We can still provide a new
internal-complete-buffer-sans-current.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> The reason of adding the C version of internal-complete-buffer was the
>> dynamic nature of the buffer list.  This is described in detail in
>> http://thread.gmane.org/gmane.emacs.devel/58453
>
> Oh, that's right.  So I suggest to leave the internal-complete-buffer in
> its C form (it works well and makes sense there since it accesses the
> internal Vbuffer_list variable).  We can still provide a new
> internal-complete-buffer-sans-current.

If there is no equivalent of Vbuffer_alist in Lisp, then
internal-complete-buffer-sans-current will produce wrong results.

A comment in buffer.c says it was a Lisp variable initially:

  /* Alist of all buffer names vs the buffers. */
  /* This used to be a variable, but is no longer,
   to prevent lossage due to user rplac'ing this alist or its elements.  */
  Lisp_Object Vbuffer_alist;

Maybe then we should expose Vbuffer_alist to Lisp via a new function like
`(buffer-alist)' to prevent modifying this alist, but allowing getting
data from it in Lisp.  This will allow implementing equivalent
functionality in other related Lisp functions as well.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> If there is no equivalent of Vbuffer_alist in Lisp, then
> internal-complete-buffer-sans-current will produce wrong results.

I don't understand what's the problem.  Your previously posted code
should work just fine, for example.  Or you could define it on top of
the internal-complete-buffer code, using the `predicate' argument.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> If there is no equivalent of Vbuffer_alist in Lisp, then
>> internal-complete-buffer-sans-current will produce wrong results.
>
> I don't understand what's the problem.  Your previously posted code
> should work just fine, for example.

I understood that the reason not to move `internal-complete-buffer'
to Lisp is because it uses the internal variable Vbuffer_alist.

Since these two functions are equivalent in behavior with the exception
that one of them omits the current buffer, I thought that it would be
better to have equivalent implementations by having both in Lisp with
removing possible problems of using Vbuffer_alist that you mentioned.

> Or you could define it on top of the internal-complete-buffer code,
> using the `predicate' argument.

Do you mean something like

(defun internal-complete-buffer-sans-current (string predicate flag)
  "Perform completion on buffer names excluding the current buffer.
Like `internal-complete-buffer', but removes the current buffer from
the completion list."
  (let ((current-buffer (buffer-name (other-buffer (current-buffer) t))))
    (internal-complete-buffer
     string (lambda (buffer) (not (equal current-buffer buffer))) flag)))

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>>> If there is no equivalent of Vbuffer_alist in Lisp, then
>>> internal-complete-buffer-sans-current will produce wrong results.
>> 
>> I don't understand what's the problem.  Your previously posted code
>> should work just fine, for example.

> I understood that the reason not to move `internal-complete-buffer'
> to Lisp is because it uses the internal variable Vbuffer_alist.

> Since these two functions are equivalent in behavior with the exception
> that one of them omits the current buffer, I thought that it would be
> better to have equivalent implementations

Actually it's better in that case to have one of the two implemented as
a "diff" over the other.

> by having both in Lisp with removing possible problems of using
> Vbuffer_alist that you mentioned.

Yes, except that your proposal doesn't remove possible problems with using
Vbuffer_alist (that problem is if someone gets hold of this list
and does things like "rplcd" (i.e. setcdr) on it).

>> Or you could define it on top of the internal-complete-buffer code,
>> using the `predicate' argument.

> Do you mean something like

> (defun internal-complete-buffer-sans-current (string predicate flag)
>   "Perform completion on buffer names excluding the current buffer.
> Like `internal-complete-buffer', but removes the current buffer from
> the completion list."
>   (let ((current-buffer (buffer-name (other-buffer (current-buffer) t))))
>     (internal-complete-buffer
>      string (lambda (buffer) (not (equal current-buffer buffer))) flag)))

Yes, except it should not ignore `predicate'.
You could define it in terms of completion-table-with-predicate (100%
untested code, of course):

(defun internal-complete-buffer-except (&optional buf)
  (lexical-let ((except (if (stringp buf) buf (buffer-name buf))))
    (apply-partially 'completion-table-with-predicate
                     'internal-complete-buffer
                     (lambda (name) (not (equal name except)))
                     nil)))


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> by having both in Lisp with removing possible problems of using
>> Vbuffer_alist that you mentioned.
>
> Yes, except that your proposal doesn't remove possible problems with using
> Vbuffer_alist (that problem is if someone gets hold of this list
> and does things like "rplcd" (i.e. setcdr) on it).

That is why I propose creating a function `buffer-alist' that returns
a copy of Vbuffer_alist, not using this variable directly.  So modifications
on this alist would be impossible.

>>> Or you could define it on top of the internal-complete-buffer code,
>>> using the `predicate' argument.
>
>> Do you mean something like
>
>> (defun internal-complete-buffer-sans-current (string predicate flag)
>>   "Perform completion on buffer names excluding the current buffer.
>> Like `internal-complete-buffer', but removes the current buffer from
>> the completion list."
>>   (let ((current-buffer (buffer-name (other-buffer (current-buffer) t))))
>>     (internal-complete-buffer
>>      string (lambda (buffer) (not (equal current-buffer buffer))) flag)))
>
> Yes, except it should not ignore `predicate'.
> You could define it in terms of completion-table-with-predicate (100%
> untested code, of course):
>
> (defun internal-complete-buffer-except (&optional buf)
>   (lexical-let ((except (if (stringp buf) buf (buffer-name buf))))
>     (apply-partially 'completion-table-with-predicate
>                      'internal-complete-buffer
>                      (lambda (name) (not (equal name except)))
>                      nil)))

This works correctly in the following patch:

Index: lisp/minibuffer.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/minibuffer.el,v
retrieving revision 1.18
diff -c -r1.18 minibuffer.el
*** lisp/minibuffer.el	19 Apr 2008 03:34:02 -0000	1.18
--- lisp/minibuffer.el	19 Apr 2008 22:46:29 -0000
***************
*** 758,763 ****
--- 758,772 ----
                              'completion--file-name-table)
    "Internal subroutine for `read-file-name'.  Do not call this.")
  
+ (defun internal-complete-buffer-except (&optional buffer)
+   "Perform completion on all buffers excluding BUFFER.
+ Like `internal-complete-buffer', but removes BUFFER from the completion list."
+   (lexical-let ((except (if (stringp buffer) buffer (buffer-name buffer))))
+     (apply-partially 'completion-table-with-predicate
+ 		     'internal-complete-buffer
+ 		     (lambda (name) (not (equal name except)))
+ 		     nil)))
+ 
  (provide 'minibuffer)
  
  ;; arch-tag: ef8a0a15-1080-4790-a754-04017c02f08f

Index: lisp/files.el
===================================================================
RCS file: /sources/emacs/emacs/lisp/files.el,v
retrieving revision 1.971
diff -c -r1.971 files.el
*** lisp/files.el	19 Apr 2008 03:33:12 -0000	1.971
--- lisp/files.el	19 Apr 2008 22:46:42 -0000
***************
*** 1039,1044 ****
--- 1039,1054 ----
      (rename-file encoded new-encoded ok-if-already-exists)
      newname))
  \f
+ (defun read-buffer-to-switch (prompt)
+   "Read the name of a buffer to switch to and return as a string.
+ It is intended for `switch-to-buffer' family of commands since they
+ need to omit the name of current buffer from the list of defaults."
+   (minibuffer-with-setup-hook
+       (lambda ()
+ 	(set (make-local-variable 'minibuffer-completion-table)
+ 	     (internal-complete-buffer-except (other-buffer (current-buffer) t))))
+     (read-buffer prompt (other-buffer (current-buffer)))))
+ 
  (defun switch-to-buffer-other-window (buffer &optional norecord)
    "Select buffer BUFFER in another window.
  If BUFFER does not identify an existing buffer, then this function
***************
*** 1053,1059 ****
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive "BSwitch to buffer in other window: ")
    (let ((pop-up-windows t)
  	;; Don't let these interfere.
  	same-window-buffer-names same-window-regexps)
--- 1063,1070 ----
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive
!    (list (read-buffer-to-switch "Switch to buffer in other window: ")))
    (let ((pop-up-windows t)
  	;; Don't let these interfere.
  	same-window-buffer-names same-window-regexps)
***************
*** 1067,1073 ****
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive "BSwitch to buffer in other frame: ")
    (let ((pop-up-frames t)
  	same-window-buffer-names same-window-regexps)
      (prog1
--- 1078,1085 ----
  
  This uses the function `display-buffer' as a subroutine; see its
  documentation for additional customization information."
!   (interactive
!    (list (read-buffer-to-switch "Switch to buffer in other frame: ")))
    (let ((pop-up-frames t)
  	same-window-buffer-names same-window-regexps)
      (prog1
***************
*** 1978,1984 ****
    (if (fboundp 'ucs-set-table-for-input) ; don't lose when building
        (ucs-set-table-for-input)))
  
! (defcustom auto-mode-case-fold nil
    "Non-nil means to try second pass through `auto-mode-alist'.
  This means that if the first case-sensitive search through the alist fails
  to find a matching major mode, a second case-insensitive search is made.
--- 1990,1996 ----
    (if (fboundp 'ucs-set-table-for-input) ; don't lose when building
        (ucs-set-table-for-input)))
  
! (defcustom auto-mode-case-fold t
    "Non-nil means to try second pass through `auto-mode-alist'.
  This means that if the first case-sensitive search through the alist fails
  to find a matching major mode, a second case-insensitive search is made.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> This works correctly in the following patch:

Looks good, feel free to install.


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Richard Stallman]

    That is why I propose creating a function `buffer-alist' that returns
    a copy of Vbuffer_alist, not using this variable directly.  So modifications
    on this alist would be impossible.

(defun buffer-alist ()
  (mapcar (lambda (b) (cons (buffer-name b) b))
          (buffer-list)))

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> This works correctly in the following patch:
>
> Looks good, feel free to install.

Ah, I accidentally tested this with a Lisp version of
`internal-complete-buffer' that I created earlier.
But when I restored the C version of `internal-complete-buffer',
it doesn't work with it.

I don't understand why it works with with the equivalent
Lisp version, but not with the C version.  The C version
returns a full buffer list disregarding the `predicate' arg.
This is a working Lisp version for your reference:

(defun internal-complete-buffer (string predicate flag)
  (let ((table (mapcar 'buffer-name (buffer-list))))
    (cond
     ((null flag) (try-completion string table predicate))
     ((eq flag t) (all-completions string table predicate t))
     (t (test-completion string table predicate)))))

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

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>     That is why I propose creating a function `buffer-alist' that returns
>     a copy of Vbuffer_alist, not using this variable directly.  So modifications
>     on this alist would be impossible.
>
> (defun buffer-alist ()
>   (mapcar (lambda (b) (cons (buffer-name b) b))
>           (buffer-list)))

If this is so simple, then what problems would cause moving
`internal-complete-buffer' to Lisp?

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>> That is why I propose creating a function `buffer-alist' that returns
>> a copy of Vbuffer_alist, not using this variable directly.  So modifications
>> on this alist would be impossible.
>> 
>> (defun buffer-alist ()
>> (mapcar (lambda (b) (cons (buffer-name b) b))
>> (buffer-list)))

> If this is so simple, then what problems would cause moving
> `internal-complete-buffer' to Lisp?

None (other than a very minor reduction in consing), but I'm not sure
what would be the benefits,


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

> Ah, I accidentally tested this with a Lisp version of
> `internal-complete-buffer' that I created earlier.
> But when I restored the C version of `internal-complete-buffer',
> it doesn't work with it.

> I don't understand why it works with with the equivalent
> Lisp version, but not with the C version.  The C version
> returns a full buffer list disregarding the `predicate' arg.
> This is a working Lisp version for your reference:

That is strange indeed.  Could you investigate?


        Stefan

Re: C-r and C-s in minibuffer should search completion

[Juri Linkov]

>> Ah, I accidentally tested this with a Lisp version of
>> `internal-complete-buffer' that I created earlier.
>> But when I restored the C version of `internal-complete-buffer',
>> it doesn't work with it.
>
>> I don't understand why it works with with the equivalent
>> Lisp version, but not with the C version.  The C version
>> returns a full buffer list disregarding the `predicate' arg.
>> This is a working Lisp version for your reference:
>
> That is strange indeed.  Could you investigate?

It is very inconvenient to debug higher-order functions when part
of functionality is implemented in C.  This is another reason to
move completion functions to Lisp.

Nevertheless, I've found that this was caused by internal-complete-buffer
that really uses a cons pair of (buffer-name . buffer) for the
buffer name argument.  I installed a fixed version of
internal-complete-buffer-except.

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

Re: C-r and C-s in minibuffer should search completion

[Stefan Monnier]

>>> Ah, I accidentally tested this with a Lisp version of
>>> `internal-complete-buffer' that I created earlier.
>>> But when I restored the C version of `internal-complete-buffer',
>>> it doesn't work with it.
>> 
>>> I don't understand why it works with with the equivalent
>>> Lisp version, but not with the C version.  The C version
>>> returns a full buffer list disregarding the `predicate' arg.
>>> This is a working Lisp version for your reference:
>> 
>> That is strange indeed.  Could you investigate?

> It is very inconvenient to debug higher-order functions when part
> of functionality is implemented in C.  This is another reason to
> move completion functions to Lisp.

> Nevertheless, I've found that this was caused by internal-complete-buffer
> that really uses a cons pair of (buffer-name . buffer) for the
> buffer name argument.  I installed a fixed version of
> internal-complete-buffer-except.

Thanks,


        Stefan