I am trying to learn CL using ClozureCL and am in the middle of Google's Lisp koans.
Spoiler warning: I'm giving away an answer to frame my question, because if I don't do so, submitted answers might not be as targeted.
The code here shows a value of x that passes.
(define-test test-guess-that-type!
(let ((x '(SIMPLE-ARRAY ARRAY (5 3 *))))
(assert-true (subtypep x '(SIMPLE-ARRAY T (* 3 *))))
(assert-true (subtypep x '(SIMPLE-ARRAY T (5 * *))))
(assert-true (subtypep x '(SIMPLE-ARRAY ARRAY *)))
(assert-true (typep (make-array '(5 3 9) :element-type 'STRING ) x))
(assert-true (typep (make-array '(5 3 33) :element-type 'VECTOR ) x))))
I didn't feel like I learned much though beyond the pattern used in type notation. I wanted to see if I can pass the test using a value of x in the form (type-of ...) so that I can relate actual values to types by example.
That said, here's my current uneducated guess. The assert I marked with ; <!> is failing for my first chosen value for x.
(define-test test-guess-that-type!
(let ((x (type-of (make-array '(5 3 33) :element-type 'VECTOR))))
(assert-true (subtypep x '(SIMPLE-ARRAY T (* 3 *))))
(assert-true (subtypep x '(SIMPLE-ARRAY T (5 * *))))
(assert-true (subtypep x '(SIMPLE-ARRAY ARRAY *)))
(assert-true (typep (make-array '(5 3 9) :element-type 'STRING ) x)) ; <!>
(assert-true (typep (make-array '(5 3 33) :element-type 'VECTOR ) x))))
My question is: If you are restricted to use of (type-of <val>) what <val> solves the koan?
Observations thus far:
(type-of x) is (SIMPLE-ARRAY T (5 3 33)), which is obviously not what I want. I want (SIMPLE-ARRAY ARRAY (5 3 *)) with vector elements.fixnum values, and setting :adjustable t makes the array "expressly adjustable", which apparently means the array is no longer a SIMPLE-ARRAY.I don't think there's a solution to this. Actually, it's possible that you might find a solution in some CL implementations, but there's no guarantee about it. The specification of TYPE-OF doesn't go into detail about the type specifier returned in most cases, merely requiring that
(typep object (type-of object))
must be true, along with a few other retrictions. But there's nothing specific to arrays.
I don't think any implementations will ever return type specifiers containing * in the dimensions for an array type. They'll likely either return a very general type specifier that omits the dimensions entirely, or a very specific one that contains the actual dimensions of the given array.
Nothing prevents an implementation from returning a type specifier like the one you're looking for, but it would be pretty perverse. Given a 3-dimensional array, why would it specifically choose to make the last dimension unspecified rather than one or all of the others?