assoc-delete-all

assoc-delete-all

[Lennart Borgman]

There is an assq-delete-all, but I am missing assoc-delete-all ... or, 
am I missing something (else)?

Re: assoc-delete-all

[Juanma Barranquero]

> There is an assq-delete-all, but I am missing assoc-delete-all ... or,
> am I missing something (else)?

If I had to guess, I'd say that is because `assq-delete-all' has only
one reasonable behavior, i.e., as it modifies the structure, you
usually are going to do

  (setq alist (assq-delete-all 'my-key alist))

or directly construct the alist, filter it through `assq-delete-all'
and then discard it.

With `assoc-delete-all' you have to decide whether you want a shallow
or a deep copy, for example:

  (defun assoc-delete-all (key alist)
   (let (l)
     (while alist
       (unless (and (consp (car alist))
                    (eq key (caar alist)))
         (setq l (cons (car alist) l)))
       (setq alist (cdr alist)))
     (nreverse l)))

will do what you want, but it will still share conses with the
original list. Other implementations can use `copy-tree' or
`copy-alist', but really, there's no one answer that is good for every
situation. Kent M. Pitman did a wonderful article about this issue (he
was speaking of copy) a long time ago: "The Best of Intentions: EQUAL
Rights--and Wrongs--in Lisp",
http://www.nhplace.com/kent/PS/EQUAL.html

-- 
                    /L/e/k/t/u

Re: assoc-delete-all

[Lennart Borgman]

Juanma Barranquero wrote:

> ...
>
>will do what you want, but it will still share conses with the
>original list. Other implementations can use `copy-tree' or
>`copy-alist', but really, there's no one answer that is good for every
>situation. Kent M. Pitman did a wonderful article about this issue (he
>was speaking of copy) a long time ago: "The Best of Intentions: EQUAL
>Rights--and Wrongs--in Lisp",
>http://www.nhplace.com/kent/PS/EQUAL.html
>
Thanks for the good lessons.

Re: assoc-delete-all

[Lute Kamstra]

Lennart Borgman <lennart.borgman.073@student.lu.se> writes:

> There is an assq-delete-all, but I am missing assoc-delete-all 

I guess nobody needed it before.  Do you want to use it?

Lute.

Re: assoc-delete-all

[Lennart Borgman]

Lute Kamstra wrote:

>Lennart Borgman <lennart.borgman.073@student.lu.se> writes:
>
>  
>
>>There is an assq-delete-all, but I am missing assoc-delete-all 
>>    
>>
>
>I guess nobody needed it before.  Do you want to use it?
>  
>
Thanks, but after reading Juanma's answer and thinking about it I 
realize that I do not need it really.

Re: assoc-delete-all

[Lennart Borgman]

Lute Kamstra wrote:

>Lennart Borgman <lennart.borgman.073@student.lu.se> writes:
>
>  
>
>>There is an assq-delete-all, but I am missing assoc-delete-all 
>>    
>>
>
>I guess nobody needed it before.  Do you want to use it?
>  
>
Thanks, but after reading Juanma's answer and thinking about it I 
realize that I do not need it really.

Re: assoc-delete-all

[Richard M. Stallman]

    There is an assq-delete-all, but I am missing assoc-delete-all ... or, 

We don't want to add assoc-delete-all merely for symmetry.
It would have to be needed in practice.

Re: assoc-delete-all

[Juri Linkov]

>     There is an assq-delete-all, but I am missing assoc-delete-all ... or, 
>
> We don't want to add assoc-delete-all merely for symmetry.
> It would have to be needed in practice.

Well, assoc-delete-all will help to fix the bug in dired-compare-directories.

Sometimes the number of marked files it reports is wrong.  That's
because it compares the file attributes of directories "." and "..",
but dired mark-functions don't mark them.

assoc-delete-all could filter out "." and ".." from the list of file
attributes returned by dired-files-attributes, which is an alist
with string keys

(("." "<full-name>" <attributes>)
 (".." "<full-name>" <attributes>)
 ("file" "<full-name>" <attributes>))

with this function

(assoc-delete-all "." (assoc-delete-all ".." file-alist1))

Even better would be to allow a test function like in assoc-default:

(assoc-delete-all nil file-alist1 (lambda (key) (string-match "^\.\.?$" key)))

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

Re: assoc-delete-all

[Juanma Barranquero]

On 7/4/05, Juri Linkov <juri@jurta.org> wrote:

> with this function
> 
> (assoc-delete-all "." (assoc-delete-all ".." file-alist1))
> 
> Even better would be to allow a test function like in assoc-default:
> 
> (assoc-delete-all nil file-alist1 (lambda (key) (string-match "^\.\.?$" key)))

That's one of these things that would be much easier if we could use CL:

 ; Not a particularly good implementation, but...
 (defun assoc-delete-all (key alist &rest keywords)
  (or keywords (setq keywords (list ':test 'eq )))
  (apply #'remove* key alist :key #'car keywords))

and you could do

 (assoc-delete-all "^\.\.?$" file-alist1
                   :test #'(lambda (test item)
                             (string-match test item)))

<sigh> :-)

-- 
                    /L/e/k/t/u

Re: assoc-delete-all

[Richard M. Stallman]

    Well, assoc-delete-all will help to fix the bug in dired-compare-directories.

    Sometimes the number of marked files it reports is wrong.  That's
    because it compares the file attributes of directories "." and "..",
    but dired mark-functions don't mark them.

Is the following patch not right?

*** dired-aux.el	03 Jul 2005 12:12:32 -0400	1.133
--- dired-aux.el	04 Jul 2005 15:01:37 -0400	
***************
*** 136,141 ****
--- 136,143 ----
                        (dired-file-set-difference
                         file-alist2 file-alist1
  		       predicate))))
+     (setq file-list1 (delete "." (delete ".." file-list1)))
+     (setq file-list2 (delete "." (delete ".." file-list2)))
      (dired-fun-in-all-buffers
       dir1 nil
       (lambda ()

Re: assoc-delete-all

[Juri Linkov]

> Is the following patch not right?
>
> *** dired-aux.el	03 Jul 2005 12:12:32 -0400	1.133
> --- dired-aux.el	04 Jul 2005 15:01:37 -0400	
> ***************
> *** 136,141 ****
> --- 136,143 ----
>                         (dired-file-set-difference
>                          file-alist2 file-alist1
>   		       predicate))))
> +     (setq file-list1 (delete "." (delete ".." file-list1)))
> +     (setq file-list2 (delete "." (delete ".." file-list2)))
>       (dired-fun-in-all-buffers
>        dir1 nil
>        (lambda ()

This doesn't work because file names in file-list1 and file-list2
are absolute.

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

Re: assoc-delete-all

[Juri Linkov]

>> with this function
>> 
>> (assoc-delete-all "." (assoc-delete-all ".." file-alist1))
>> 
>> Even better would be to allow a test function like in assoc-default:
>> 
>> (assoc-delete-all nil file-alist1 (lambda (key) (string-match "^\.\.?$" key)))
>
> That's one of these things that would be much easier if we could use CL:

Another thing that would be easier with CL is its `set-difference'
which is currently duplicated in dired-aux.el as `dired-file-set-difference'
with the specialization to work only on file attributes.

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