avoid duplicates in history

avoid duplicates in history

[Dan Nicolaescu]

One feature of tcsh that I miss in emacs is "histdup=erase". When
histdup is set that way when a command is entered in history all
previous occurrences of the same command will be erased. 
bash3 will also implement this feature, it will be called
"erasedups". 

Here is an implementation of this feature for Emacs. 

If this is acceptable, I will provide a complete patch with
documentation. 

Feedback would be appreciated. 
Is the history-erasedups name ok? 

Thanks.


*** minibuf.c	27 May 2004 23:22:35 -0700	1.266
--- minibuf.c	01 Jun 2004 09:47:22 -0700	
***************
*** 61,66 ****
--- 61,71 ----
  
  Lisp_Object Qhistory_length, Vhistory_length;
  
+ /* No duplicates in history.  */
+ 
+ int history_erasedups;
+ 
+ 
  /* Fread_minibuffer leaves the input here as a string. */
  
  Lisp_Object last_minibuf_string;
***************
*** 742,748 ****
  	{
  	  Lisp_Object length;
  
! 	  histval = Fcons (histstring, histval);
  	  Fset (Vminibuffer_history_variable, histval);
  
  	  /* Truncate if requested.  */
--- 747,754 ----
  	{
  	  Lisp_Object length;
  
! 	  if (history_erasedups) Fdelete (histstring, histval);
!           histval = Fcons (histstring, histval);
  	  Fset (Vminibuffer_history_variable, histval);
  
  	  /* Truncate if requested.  */
***************
*** 2631,2636 ****
--- 2637,2647 ----
  property of a history variable overrides this default.  */);
    XSETFASTINT (Vhistory_length, 30);
  
+   DEFVAR_BOOL ("history-erasedups", &history_erasedups,
+ 	       doc: /* *Non-nil means delete duplicates in history.
+   */);
+   history_erasedups = 0;
+   
    DEFVAR_LISP ("completion-auto-help", &Vcompletion_auto_help,
  	       doc: /* *Non-nil means automatically provide help for invalid completion input.  */);
    Vcompletion_auto_help = Qt;

Re: avoid duplicates in history

[Juri Linkov]

Dan Nicolaescu <dann@ics.uci.edu> writes:
> Feedback would be appreciated. 

I think this is a useful feature.

> Is the history-erasedups name ok? 

There are similar variables `comint-input-ignoredups' and
`eshell-hist-ignoredups', so it could be rather named as
`history-ignoredups'.

OTOH, by functionality it is very like `delete-dups' so the name
`history-deletedups' is no worse.

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

Re: avoid duplicates in history

[Miles Bader]

Juri Linkov <juri@jurta.org> writes:

> Dan Nicolaescu <dann@ics.uci.edu> writes:
>> Feedback would be appreciated. 
>
> I think this is a useful feature.
>
>> Is the history-erasedups name ok? 
>
> There are similar variables `comint-input-ignoredups' and
> `eshell-hist-ignoredups', so it could be rather named as
> `history-ignoredups'.
>
> OTOH, by functionality it is very like `delete-dups' so the name
> `history-deletedups' is no worse.

Yes it is, it runs two words together for no reason.  Just use
`history-delete-duplicates'.

-Miles
-- 
Occam's razor split hairs so well, I bought the whole argument!

Re: avoid duplicates in history

[Dan Nicolaescu]

Miles Bader <miles@lsi.nec.co.jp> writes:

  > Yes it is, it runs two words together for no reason.  Just use
  > `history-delete-duplicates'.

How about the default? Should this be turned on by default? 

I even got some suggestions that deleting duplicates should be the
only way to do it. 

(It would be nice to have a decision about this, so I can write the
final docs and patch). 

        --dan

Re: avoid duplicates in history

[Juri Linkov]

Dan Nicolaescu <dann@ics.uci.edu> writes:
> How about the default? Should this be turned on by default? 

It seems reasonable to turn it on by default.

> (It would be nice to have a decision about this, so I can write the
> final docs and patch). 

When you'll write the final patch, please don't forget to make this
option customizable by adding it to cus-start.el somewhere near
`history-length'.

And please consider the situation where your patch doesn't work.
There is the following condition in `read_minibuf':

	      /* Don't duplicate the most recent entry in the history.  */
	      && NILP (Fequal (histstring, Fcar (histval)))))

which prevents removing all duplicate elements from the history
with your patch.

This condition is redundant when `history-delete-duplicates' is t.

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

Re: avoid duplicates in history

[Dan Nicolaescu]

Juri Linkov <juri@jurta.org> writes:

  > Dan Nicolaescu <dann@ics.uci.edu> writes:

  > When you'll write the final patch, please don't forget to make this
  > option customizable by adding it to cus-start.el somewhere near
  > `history-length'.

Thanks, will do, I didn't know about that. 

  > And please consider the situation where your patch doesn't work.
  > There is the following condition in `read_minibuf':
  > 
  > 	      /* Don't duplicate the most recent entry in the history.  */
  > 	      && NILP (Fequal (histstring, Fcar (histval)))))
  > 
  > which prevents removing all duplicate elements from the history
  > with your patch.
  > 
  > This condition is redundant when `history-delete-duplicates' is t.

I believe this is fine, that condition just avoids adding consecutive
identical entries, there's no need to do anything when that condition
is false.

Re: avoid duplicates in history

[Juri Linkov]

Dan Nicolaescu <dann@ics.uci.edu> writes:
> Juri Linkov <juri@jurta.org> writes:
>   > This condition is redundant when `history-delete-duplicates' is t.
>
> I believe this is fine, that condition just avoids adding consecutive
> identical entries, there's no need to do anything when that condition
> is false.

It can't delete duplicates if the history contains several of them.
This is possible when the value of `history-delete-duplicates' is
changed during the Emacs session.  But when it is set in .emacs, your
patch will guarantee that no duplicates are added to the history.  So
for the case when the value `history-delete-duplicates' isn't changed
during the Emacs session, your patch is ok.

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

Re: avoid duplicates in history

[Richard Stallman]

    > How about the default? Should this be turned on by default? 

    It seems reasonable to turn it on by default.

Absolutely not!  It would be a great surprise, and it would mean
you could not look around at your error messages without changing
other buffers on the screen.

This is a useful feature, but only when people ask for it.

Re: avoid duplicates in history

[Stefan Monnier]

>> How about the default? Should this be turned on by default? 
>     It seems reasonable to turn it on by default.

> Absolutely not!  It would be a great surprise, and it would mean
> you could not look around at your error messages without changing
> other buffers on the screen.

Huh?  I don't understand why removing duplicates from the history
would ever have any such effect.
Could you explain what use case you're thinking of ?


        Stefan

removing duplicate from history by default [was Re: avoid duplicates in history]

[Dan Nicolaescu]

Let's reopen the discussion started with the message here:

http://lists.gnu.org/archive/html/emacs-devel/2004-06/msg00021.html

I would like to prepare a patch that can be checked in. 

There was one question left, if removing duplicates from history
should be turned on by default. 
IMHO duplicates should be removed from history by default, most users
are more interested in the unique data present in history. If an exact
sequence of events is desired M-x list-command-history can do that.

[I reformated the last message from Stephan in that thread  to make it
clear who said what]:

Dan> How about the default? Should this be turned on by default? 
Juri>     It seems reasonable to turn it on by default.

RMS> Absolutely not!  It would be a great surprise, and it would mean
RMS> you could not look around at your error messages without changing
RMS> other buffers on the screen.

Stephan> Huh?  I don't understand why removing duplicates from the history
Stephan> would ever have any such effect.
Stephan> Could you explain what use case you're thinking of ?

Thanks
                --dan

Re: removing duplicate from history by default [was Re: avoid duplicates in history]

[Richard Stallman]

    Let's reopen the discussion started with the message here:

    http://lists.gnu.org/archive/html/emacs-devel/2004-06/msg00021.html

No, please let's not.  What we need to do now is fix the problems
that prevent a release.  Please let's not take up other changes
which are not necessary now.

In order to make a new release, we need to check the manual for errors.
Luc Teirlinck has done a lot of work on this, but hardly anyone else has.

Could you check one chapter of the Lisp manual now?

Re: removing duplicate from history by default

[Juri Linkov]

Richard Stallman <rms@gnu.org> writes:
>     Let's reopen the discussion started with the message here:
>
>     http://lists.gnu.org/archive/html/emacs-devel/2004-06/msg00021.html
>
> No, please let's not.  What we need to do now is fix the problems
> that prevent a release.  Please let's not take up other changes
> which are not necessary now.

Concentrating on making the next release is undoubtedly the top priority,
but the proposed change is so small (but still very useful), that it's
better to install it now than to raise the same question again in the future.

Dan Nicolaescu <dann@godzilla.ics.uci.edu> writes:

> There was one question left, if removing duplicates from history
> should be turned on by default. 

Let's not turn it on by default.  Even shells don't set a variable to
remove duplicates by default, isn't?  And so this change will not
change the current Emacs behavior.

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

Re: removing duplicate from history by default

[Dan Nicolaescu]

Juri Linkov <juri@jurta.org> writes:

  > Richard Stallman <rms@gnu.org> writes:
>>     Let's reopen the discussion started with the message here:
>>
>>     http://lists.gnu.org/archive/html/emacs-devel/2004-06/msg00021.html
>>
>> No, please let's not.  What we need to do now is fix the problems
>> that prevent a release.  Please let's not take up other changes
>> which are not necessary now.
>
  > Concentrating on making the next release is undoubtedly the top priority,
  > but the proposed change is so small (but still very useful), that it's
  > better to install it now than to raise the same question again in the futur\
e.

Agreed, the code changes are minimal: 1 line of code added + glue code
to declare a lisp variable. Plus the code already exists, I have used
it for a few months with no problems.

  > Dan Nicolaescu <dann@godzilla.ics.uci.edu> writes:
>
>> There was one question left, if removing duplicates from history
>> should be turned on by default.
>
  > Let's not turn it on by default.  Even shells don't set a variable to
  > remove duplicates by default, isn't?  

Right. 

  > And so this change will not change the current Emacs behavior.

Sure, a complete patch is attached. Please install it if it's OK. 

2004-09-05  Dan Nicolaescu  <dann@ics.uci.edu>

	* minibuf.c (history_delete_duplicates): New variable.
	(read_minibuf): Use it.
	(syms_of_minibuf): Create the corresponding lisp variable.


Index: NEWS
===================================================================
RCS file: /cvsroot/emacs/emacs/etc/NEWS,v
retrieving revision 1.1019
diff -c -3 -p -c -r1.1019 NEWS
*** NEWS	2 Sep 2004 16:39:19 -0000	1.1019
--- NEWS	5 Sep 2004 17:54:45 -0000
*************** from the file name or buffer contents.
*** 1792,1797 ****
--- 1798,1807 ----
  This option can be disabled, to avoid the normal behavior of isearch
  which puts calls to `isearch-resume' in the command history.
  
+ ** New user option `history-delete-duplicate'.
+ If set to t when adding a new history entry, all previous identical
+ entries are deleted. 
+ 
  ---
  ** Lisp mode now uses font-lock-doc-face for the docstrings.
  

Index: minibuf.c
===================================================================
RCS file: /cvsroot/emacs/emacs/src/minibuf.c,v
retrieving revision 1.271
diff -c -3 -p -c -r1.271 minibuf.c
*** minibuf.c	14 Aug 2004 12:46:46 -0000	1.271
--- minibuf.c	5 Sep 2004 17:40:26 -0000
*************** Lisp_Object Vcompletion_auto_help;
*** 61,66 ****
--- 61,71 ----
  
  Lisp_Object Qhistory_length, Vhistory_length;
  
+ /* No duplicates in history.  */
+ 
+ int history_delete_duplicates;
+ 
+ 
  /* Fread_minibuffer leaves the input here as a string. */
  
  Lisp_Object last_minibuf_string;
*************** read_minibuf (map, initial, prompt, back
*** 749,755 ****
  	{
  	  Lisp_Object length;
  
! 	  histval = Fcons (histstring, histval);
  	  Fset (Vminibuffer_history_variable, histval);
  
  	  /* Truncate if requested.  */
--- 754,761 ----
  	{
  	  Lisp_Object length;
  
! 	  if (history_delete_duplicates) Fdelete (histstring, histval);
!           histval = Fcons (histstring, histval);
  	  Fset (Vminibuffer_history_variable, histval);
  
  	  /* Truncate if requested.  */
*************** just after a new element is inserted.  S
*** 2668,2673 ****
--- 2674,2685 ----
  property of a history variable overrides this default.  */);
    XSETFASTINT (Vhistory_length, 30);
  
+   DEFVAR_BOOL ("history-delete-duplicates", &history_delete_duplicates,
+ 	       doc: /* *Non-nil means delete duplicates in history.
+ If set to t when adding a new history entry, all previous identical
+ entries are deleted.  */);
+   history_delete_duplicates = 0;
+   
    DEFVAR_LISP ("completion-auto-help", &Vcompletion_auto_help,
  	       doc: /* *Non-nil means automatically provide help for invalid completion input.  */);
    Vcompletion_auto_help = Qt;

Re: removing duplicate from history by default

[Richard Stallman]

I won't object to this simple change.
Someone please install it; let's not spend more time discussing it.

Re: avoid duplicates in history

[Richard Stallman]

      > Yes it is, it runs two words together for no reason.  Just use
      > `history-delete-duplicates'.

    How about the default? Should this be turned on by default? 

Please do not change the default.

Re: avoid duplicates in history

[Juri Linkov]

Dan Nicolaescu <dann@ics.uci.edu> writes:
> Miles Bader <miles@lsi.nec.co.jp> writes:
>   > Yes it is, it runs two words together for no reason.  Just use
>   > `history-delete-duplicates'.
>
> How about the default? Should this be turned on by default? 
>
> I even got some suggestions that deleting duplicates should be the
> only way to do it. 
>
> (It would be nice to have a decision about this, so I can write the
> final docs and patch). 

This is a good feature even without turning it on by default.  I think
it should be installed after renaming to history-delete-duplicates and
updating the Emacs manual and NEWS.

Meanwhile, I'd like to propose a related feature.  It would be useful
to be able to delete some old history elements from the history list.
I think a good key binding for this command in the minibuffer is M-k.

Index: lisp/simple.el
===================================================================
RCS file: /cvsroot/emacs/emacs/lisp/simple.el,v
retrieving revision 1.648
diff -u -r1.648 simple.el
--- lisp/simple.el	13 Jun 2004 22:00:17 -0000	1.648
+++ lisp/simple.el	26 Jun 2004 20:16:59 -0000
@@ -1106,6 +1183,24 @@
   (interactive "p")
   (next-complete-history-element (- n)))
 
+(defun delete-history-element ()
+  "Delete the current minibuffer history element from the history.
+After deleting the element the history position is changed either
+to the the previous history element, or to the next history element
+if the deleted element was the last in the history list."
+  (interactive)
+  (cond
+   ((= minibuffer-history-position 1)
+    (set minibuffer-history-variable
+         (cdr (symbol-value minibuffer-history-variable))))
+   ((> minibuffer-history-position 1)
+    (setcdr (nthcdr (- minibuffer-history-position 2)
+                    (symbol-value minibuffer-history-variable))
+            (nthcdr minibuffer-history-position
+                    (symbol-value minibuffer-history-variable)))))
+  (condition-case nil (next-history-element     1) (error nil))
+  (condition-case nil (previous-history-element 1) (error nil)))
+
 ;; For compatibility with the old subr of the same name.
 (defun minibuffer-prompt-width ()
   "Return the display width of the minibuffer prompt.

Index: lisp/bindings.el
===================================================================
RCS file: /cvsroot/emacs/emacs/lisp/bindings.el,v
retrieving revision 1.132
diff -u -r1.132 bindings.el
--- lisp/bindings.el	12 Jun 2004 10:15:11 -0000	1.132
+++ lisp/bindings.el	26 Jun 2004 20:20:37 -0000
@@ -658,7 +658,8 @@
   (define-key map [prior] 'previous-history-element)
   (define-key map [up]    'previous-history-element)
   (define-key map "\es"   'next-matching-history-element)
-  (define-key map "\er"   'previous-matching-history-element))
+  (define-key map "\er"   'previous-matching-history-element)
+  (define-key map "\ek"   'delete-history-element))
 
 (define-key global-map "\C-u" 'universal-argument)
 (let ((i ?0))

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

Re: avoid duplicates in history

[Richard Stallman]

    Meanwhile, I'd like to propose a related feature.  It would be useful
    to be able to delete some old history elements from the history list.
    I think a good key binding for this command in the minibuffer is M-k.

I think that is too obscure to be included in Emacs;
the code is short, but the cost in conceptual complexity
is too great.  So please don't install it.

Those who would like it can put it in their .emacs files.