Re: Difference between list and cons*

[David Kastrup]

David Kastrup <address@hidden> writes:

> Carl Sorensen <address@hidden> writes:
>
>> David,
>>
>> I noticed that in a recent patch you changed from constructing property
>> lists with list to using cons*.
>>
>> I'm not questioning your decision.  However, I'd like to gain some
>> understanding.  Why was it necessary (or at least desirable) to make the
>> change?  As far as I understand, the difference between list and cons* is
>> that list puts '() at the tail of the list, while cons* puts the last
>> argument at the tail of the list.
>
> And that's exactly what this change was about.  In this particular case,
> I had constructed a list where the last element was property-path (or
> something like it).  But the actual form of the entry used elsewhere was
> such that the _tail_ of the list represented property-path, not the last
> element.  While I was trying out something else, some parts of LilyPond
> (I think the documentation formatter) blew up around me because I had
> missed that difference.
>
>> Any light you could shed on this would be helpful to me.
>
> Does the above help?

Just reread this, and noticed you were talking about "property lists"
here.  There are no property lists here.  Context mods have a structure
that starts with a key symbol in the car, and the data following
afterwards.  If you take a look at where they get interpreted, you'll
see in context-property.cc stuff like
      if (type == ly_symbol2scm ("push"))
        {
          SCM val = scm_cadr (entry);
          SCM grob_prop_path = scm_cddr (entry);
          sloppy_general_pushpop_property (tg, context_prop, grob_prop_path, 
val);
        }

Now you see that we are referencing scm_cadr (entry) and scm_cddr
(entry) here for value and property-path, respectively.  If
property-path were the last _element_ of the list, it would be at
scm_caddr (entry), with scm_cdddr (entry) being '().

-- 
David Kastrup