Re: memory.text

[Dirk Herrmann]

On 24 Nov 2001, Marius Vollmer wrote:

> Improving memory handling in Guile
> ----------------------------------
[...]

Very nice.  I like the concept as well as the function names you are
suggesting.


> Cell allocation and initialization
> ----------------------------------
[...]
>     SCM scm_newcell_init (scm_bits_t car, scm_bits_t cdr)
>     {
>       SCM z;
>       if (SCM_NULLP (scm_freelist))
>         z = scm_gc_for_newcell (&scm_master_freelist,
>                                 &scm_freelist);
>       else
>         {
>           z = scm_freelist;
>           scm_freelist = SCM_FREE_CELL_CDR (scm_freelist);
>         }
>       SCM_SET_CELL_WORD_1 (z, cdr);
>       SCM_SET_CELL_WORD_0 (z, car);
>       scm_remember_upto_here (cdr);
>       return into;
>     }

What about naming the function scm_make_cell, or scm_new_cell?  If that
will be the only way to get a new cell in the future, the _init postfix
will be unnecessary.  Further, scm_make_<foo> seems to fit better into the
common way of naming functions in scheme.

There is, however, something we should think about:  car and cdr are, in
your example, of type scm_t_bits.  If the new cell won't be a pair we
should not use the terms car and cdr, but rather word0 and word1 or
something.  Further, we should not use word1 as an argument for
scm_remember_upto_here, since that only accepts SCM values.  It is,
however, not clean to coerce word1 to a SCM value with SCM_PACK either,
since that turns something into a SCM value that not necessarily
represents a real SCM object.  Hmmm?

Best regards
Dirk Herrmann