Re: Circular Lists that are not lists?

[Thorsten Jolitz]

"Pascal J. Bourguignon" <pjb@informatimago.com> writes:

Hi,

> Thorsten Jolitz <tjolitz@gmail.com> writes:
>
>> Hi List, 
>>
>> in a parse-tree produced calling 'org-element-parse-buffer' in
>> org-buffer "tmp.org" I see that the first first-level headline is tagged
>> with a :CATEGORY attribute:
>>
>> ,---------------------------------------
>> | (headline (:CATEGORY #1=\"tmp\" ...))
>> `---------------------------------------
>>
>> and that the other headlines in the parse-tree reference the attribute
>> value:
>>
>> ,-------------------------------
>> | (headline (:CATEGORY #1# ...))
>> `-------------------------------
>>
>> except when a category is explicitly specified for them:
>>
>> ,------------------------------------------------
>> | (headline (:CATEGORY \"My Category Name\" ...))
>> `------------------------------------------------
>>
>> Now ignore the Org-mode specific stuff, this question is really about
>> circular lists. 
>>
>> This seems to be a case where circular lists are applied to strings -
>> should they rather be called circular sequences? 
>>
>> How can I distinguish the three cases shown when mapping the parse-tree,
>> i.e. not on the textual level after printing the tree with (print-circle
>> t), but while mapping the nested list?
>>
>>  - case 1 :: value defined #1=\"tmp\"
>>
>>  - case 2 :: value referenced #1#
>>
>>  - case 3 :: custom value without circular characteristics
>>
>> ?
>>
>> When getting the property's values, they are all just strings ...
>
> #= and ## are just a way to refer to the same object when reading a
> sexp.
>
> This has nothing to do with circular lists, or circular structures,
> other than they're the only way to build such circular lists or circular
> structures at read time.
>
> But references can be used without anything circular.
>
>     (defvar *knights-saying* '(#1=Ni! #1# #1#))
>     *knights-saying* ; --> (Ni! Ni! Ni!)
>
> They can also be used without any list:
>
>     (defvar *recursive-vector* '#1=[a #1# vector])
>     (let ((print-level 4)
>           (print-circle nil))
>       (print *recursive-vector*))
>     ;; in any sane lisp system should print:
>     [a [a [a [a # vector] vector] vector] vector]
>     ;; but in emacs it doesn't
>     ; -> #1=[a #1# vector]
>
> And notice how there can't be any notion of circular vector!  But the
> structure built by this vector is "circular" meaning that walking it
> leads to a circle in the graph of objects and references.
>
> I won't give an example with defstruct since in emacs structures are
> actually vector, so we're reduced to the previous case.  But you may try
> it as an exercise, both in emacs lisp and in Common Lisp.
>
> Use:
>  (setf *print-length* 10 *print-level* 4 *print-circle* nil) 
> in Common Lisp.
>
> Circular lists are lists, in the sense that: (deftype list () '(or
> cons null))

Quite a lot of very interesting stuff to digest. Since this
'object-reference' syntax is presented as circular-list read-syntax in
the manual, I thought it was tightly bound to the concept of circular
lists (as the function `print-circle' suggests too), so I was surprised
to see it used with string objects. 

Thanks for info (Pascal and Teemu).

-- 
cheers,
Thorsten