problem with fancy-split's ! operator; bug?

problem with fancy-split's ! operator; bug?

[Matt Swift]

The following doesn't work.  You get an error when you try to trace
the split.  The same thing works with the `:' operator as expected.
What am I missing, or is this a bug?

(defun test-split (dummy)
  '(& "one" "two"))
(setq nnmail-split-fancy '(! test-split "dummy"))

Re: problem with fancy-split's ! operator; bug?

[Colin Rafferty]

Matt Swift writes:

> The following doesn't work.  You get an error when you try to trace
> the split.  The same thing works with the `:' operator as expected.
> What am I missing, or is this a bug?

> (defun test-split (dummy)
>   '(& "one" "two"))
> (setq nnmail-split-fancy '(! test-split "dummy"))

It looks good to me.  What is the error (traceback, please)?

-- 
Colin

Re: problem with fancy-split's ! operator; bug?

[Matt Swift]

[-- Attachment #1: Type: text/plain, Size: 748 bytes --]

>> On Wed, 02 Feb 2000 17:25:48 GMT,
>>    ">" == Colin Rafferty <colin.rafferty@msdw.com> wrote:

    >> Matt Swift writes:
    >> The following doesn't work.  You get an error when you try to trace
    >> the split.  The same thing works with the `:' operator as expected.
    >> What am I missing, or is this a bug?

    >> (defun test-split (dummy)
    >> '(& "one" "two"))
    >> (setq nnmail-split-fancy '(! test-split "dummy"))

    >> It looks good to me.  What is the error (traceback, please)?

You mean it works for you?

It's a quoting problem:

Signaling: (wrong-type-argument sequencep &)
  mapconcat(car (("two" . "two") ("one" . "one") (& . &)) ", ")
  gnus-summary-respool-query()
  call-interactively(gnus-summary-respool-query)


[-- Attachment #2: User settings --]
[-- Type: application/x-emacs-lisp, Size: 10040 bytes --]

Re: problem with fancy-split's ! operator; bug?

[Colin Rafferty]

Matt Swift writes:
> Colin Rafferty wrote:
>> Matt Swift writes:

>>> The following doesn't work.  You get an error when you try to trace
>>> the split.  The same thing works with the `:' operator as expected.
>>> What am I missing, or is this a bug?

>>> (defun test-split (dummy)
>>> '(& "one" "two"))
>>> (setq nnmail-split-fancy '(! test-split "dummy"))

>> It looks good to me.  What is the error (traceback, please)?

> You mean it works for you?

Oops, no it doesn't.  It simply looked correct to me, given the
documentation.

The problem is that the documentation does not match the code.

Here is the doc for `!':

  (! FUNCTION SPLIT): Call FUNCTION with the result of SPLIT.  The
    return value FUNCTION should be a split, which is then recursively
    processed.

However, here is the code:

     ;; Builtin ! operation.
     ((eq (car split) '!)
      (funcall (cadr split) (nnmail-split-it (caddr split))))

It should be this (to match `:'):

     ;; Builtin ! operation.
     ((eq (car split) '!)
      (nnmail-split-it (save-excursion (funcall (cadr split) (nnmail-split-it (caddr split))))))

I'm not sure that Lars should change the code to match documentation,
since there may be many people out there who depend on this.  Although
the documented method is much more useful.

-- 
Colin

Re: problem with fancy-split's ! operator; bug?

[Toby Speight]

>> Colin Rafferty <URL:mailto:colin@xemacs.org>

0> In article <vgvsnzanwqi.fsf@ms.com>,
0> > wrote:

>> However, here is the code:
>>
>>      ;; Builtin ! operation.
>>      ((eq (car split) '!)
>>       (funcall (cadr split) (nnmail-split-it (caddr split))))
>>
>> It should be this (to match `:'):
>>
>>      ;; Builtin ! operation.
>>      ((eq (car split) '!)
>>       (nnmail-split-it (save-excursion (funcall (cadr split) (nnmail-split-it (caddr split))))))
>>
>> I'm not sure that Lars should change the code to match documentation,
>> since there may be many people out there who depend on this.

Can you give an example of something that works with the former code,
that behaves differently in the latter?

Re: problem with fancy-split's ! operator; bug?

[Colin Rafferty]

Toby Speight writes:

>>> Colin Rafferty <URL:mailto:colin@xemacs.org>
0> In article <vgvsnzanwqi.fsf@ms.com>,
0> > wrote:

>>> However, here is the code:
>>> 
>>> ;; Builtin ! operation.
>>> ((eq (car split) '!)
>>>  (funcall (cadr split) (nnmail-split-it (caddr split))))
>>> 
>>> It should be this (to match `:'):
>>> 
>>> ;; Builtin ! operation.
>>> ((eq (car split) '!)
>>>  (nnmail-split-it (save-excursion (funcall (cadr split) (nnmail-split-it (caddr split))))))
>>> 
>>> I'm not sure that Lars should change the code to match documentation,
>>> since there may be many people out there who depend on this.

> Can you give an example of something that works with the former code,
> that behaves differently in the latter?

Just about anything.  Since the former code expects a list of groups
in which to put the mail, and the latter code expects a split to
parse, pretty much anything will owrk differently.

For example, if I had a .gnus that looked like this:

(defun test-split (dummy)
  '("one "two"))

(setq nnmail-split-fancy '(! test-split "dummy"))

Then in the former, the resulting split would be ("one" "two"), but in
the latter, the resulting split would be ("one").

-- 
Colin

Re: problem with fancy-split's ! operator; bug?

[Toby Speight]

>> Colin Rafferty <URL:mailto:colin.rafferty@msdw.com>

0> In article <vgvn1phktct.fsf@ms.com>,
0> > wrote:

>> Toby Speight writes:
>>
>>>> Colin Rafferty <URL:mailto:colin@xemacs.org>
0> In article <vgvsnzanwqi.fsf@ms.com>,
0> > wrote:

>> Can you give an example of something that works with the former
>> code, that behaves differently in the latter?
>>
>> Just about anything.  Since the former code expects a list of
>> groups in which to put the mail, and the latter code expects a
>> split to parse, pretty much anything will owrk differently.

Did this used to work at one time?  I have old code which I don't use
any more, which looks like

dot-gnus> (defun tms-split-lists (list)
dot-gnus>   (goto-char (point-min))
dot-gnus>   (let (groups name)
dot-gnus>     (while (re-search-forward tms-list-regexp nil t)
dot-gnus>       (setq name (match-string 1))
dot-gnus>       (or (member name groups)
dot-gnus>           (setq groups (cons name groups))))
dot-gnus>     (cons '& groups)))

In particular, note that I put '& at the beginning of the list.  This
worked in 5.5 or 5.7, I think.

Re: problem with fancy-split's ! operator; bug?

[Colin Rafferty]

Toby Speight writes:
> Colin Rafferty writes:
>> Toby Speight writes:

>>> Can you give an example of something that works with the former
>>> code, that behaves differently in the latter?

>> Just about anything.  Since the former code expects a list of
>> groups in which to put the mail, and the latter code expects a
>> split to parse, pretty much anything will owrk differently.

> Did this used to work at one time?  I have old code which I don't use
> any more, which looks like

I guess if you used to use it, then the it must have worked as
advertised at one point, and now it doesn't.

I'll submit a patch.

-- 
Colin