Re: [Sbcl-help] string type uncertainty?

[Christophe R. <cs...@ca...>]

Pascal Bourguignon <pj...@in...> writes:

>> As an example: consider the type (SINGLE-FLOAT 0.0 1.0).  If you ask
>> sbcl to create an array :ELEMENT-TYPE '(SINGLE-FLOAT 0.0 1.0), it will
>> give you back an array specialized to hold general SINGLE-FLOATs, the
>> layout of which looks something like this in memory:
>> 
>>   [ header | length | element0 | element1 | element2 ]
>> 
>> where the element<n>s are unboxed single floats (i.e. they don't have
>> any header; they are just the bare IEEE single float in bits).  This
>> is what is meant by a specialized array on some type.
>
> Yes, but  the type inference  engine should remember that  elements of
> this  array are  always  between 0.0  and  1.0, and  could apply  this
> knowledge to (AREF (THE (ARRAY (SINGLE-FLOAT 0.0 1.0) PROBS) 0).

I don't think that implementation strategy is as inexpensive as you
seem to think it is.  Consider:
  (defun foo (x)
    (declare (type (simple-array (single-float 0.0 1.0) (*)) x))
    (bar x)
    (aref x 0))

Is the "type inference engine" entitled to assume that the return
value from FOO will be a single float in the range 0.0 to 1.0?  Well,
maybe, by some lights, but BAR here is an arbitrary function that
could modify the contents of X in an arbitrary way.  To provide
reasonable diagnostics in the case of violation of this constraint, we
have to execute _runtime_ checks in the presence of any modification
inside BAR (and remember, we can change BAR's function binding at
will, so we have to remember that we have to reinsert runtime checks
on recompiling BAR).

>> You might wish to consider what you would need to do to provide a
>> specialization of array to hold objects of type (ARRAY BASE-CHAR).
>
> I  think that  since objects  of type  (ARRAY BASE-CHAR  *) may  be of
> different sizes,  an array  of such  objects needs to  be an  array of
> pointers to  them.  An  (ARRAY (ARRAY BASE-CHAR  10) *) could  put the
> BASE-CHAR inline:
>
>    [ header | length |t|e|n|c|h|a|r|a|c|t|t|e|n|c|h|a|r|a|c|t|...]

No, you can't do this.  Arrays are objects which have identity; in
other words, you can use EQL to ask whether two arrays are in fact the
same object.  With the above implementation scheme, there would be no
way of, say, storing the same array twice in one of these arrays of
arrays.

Cheers,

Christophe
-- 
http://www-jcsu.jesus.cam.ac.uk/~csr21/       +44 1223 510 299/+44 7729 383 757
(set-pprint-dispatch 'number (lambda (s o) (declare (special b)) (format s b)))
(defvar b "~&Just another Lisp hacker~%")    (pprint #36rJesusCollegeCambridge)