Re: [Chicken-users] chicken-setup -test

[Kon Lovett]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


On May 25, 2007, at 8:57 PM, Ivan Raikov wrote:

>   Yes, I actually meant testbase, but I wasn't paying attention when
> typing. By the way, if I want to migrate my test code from testeez to
> testbase, what should I use in place of the test-define macro? In
> testeez, it is a convenient way to define constants local to a test,
> but it is not clear to me how to do this in testbase.

Hi Ivan,

Depends where you are in your test container hierarchy.

At the top:

; From testeez document:
(define-test %foo:test "Foo Station"
  (initial
    ; "Bar function" - sorry there isn't any keyed repository for  
these things.
    (define (bar x) (+ x 42))) )

  ; Bit of problem if this needed some 'bar' in the outer scope!
  (test/equal "Put two and two together" (+ 2 2) 4)

  (test/equal "Bar scene" (bar 69) 0)

  (test/eqv   "Full circle" (* (bar -21) 2) 42)

  (test/eqv   "Multiple"
              (values (+ 2 2) (string #\h #\i) (char-upcase #\p))
              (values 4 "hi" #\P))))
) ;%foo:test

Inside a test-case, test-suite, etc. (some kind of test container):

  ... outer ...
  (test/case "Foo Station" (
      [bar (lambda (x) (+ x 42))]
      ... )
    ... tests ...
  )
  ... outer ...

or

  ... outer ...
  (test-letrec ([bar (lambda (x) (+ x 42))] ...)
    ...
    (some test) )
  ... outer ...

(I know, too many ways to do not quite the same thing.)

That said a 'test-define' macro isn't in the cards. While the name is  
open, the body of a test container isn't, open that is. It is  
actually a list of thunks, so there is no new scope once you enter  
the clause body proper. (See below.)

(Really need a tutorial. In my copious spare time.)

> Also, something
> that testeez is lacking, but I would like to be able to do is test
> loop invariants, e.g.:
>
>
>  (some tests ... )
>
>  (do ((i min (+ 1 i))) ((> i max))
>      (some code ...)
>      (test/equal  (some test ...) #t))
>
>
> So if any of the tests in the loop fail, then there should be a way to
> indicate that in the overall test stats. Is it possible to do this in
> testbase?

Hmm, noo, but a good idea. I have thought of it.

The problem is each test is macroexpanded as a thunk in a list owned  
by the enclosing test container. The machinery assumes each non-test- 
container (i.e. an expectation) in a test container will have an  
atomic value - a test result. This is one of the reasons for forms  
like 'side-effect' and 'teardown'.

That said it becomes of issue of a "generative" test container.  
Something like:

(test/collector
  (do ((i min (+ 1 i))) ((> i max))
    (some code ...)
    (test-collect (test/equal (some test ...) #t))))

(test-collector TSTNAM DTOR ESCR [(warn WARNING)] [((VAR VAL) ...)]  
CLAUSE ...)
(test/collector [TSTNAM] [((VAR VAL) ...)] CLAUSE ...)
(test-collect CLAUSE)

The CLAUSE-body of a test-collector would function like a 'test- 
case' (i.e. fail container on 1st test failure; assuming 'test-case- 
evaluate-thru-failure' is #f). The collector container local form,  
'test-collect', would append each test result as it is returned by  
the CLAUSE. When the 'test-collector' container exits its' result  
list then becomes a node in the results tree of the parent container.

Hmm. I will look into this.

BTW, one benefit of using 'expectations' is the ability to define new  
ones:

(define-expect-unary procedure?)
...
(expect-procedure (make-format-function ...))

(define-expect-binary array-equal?)
...
(expect-array-equal (array '(3) 'a 'b 'c) (subarray ra 0 #f))

(define-expect-binary =)
(define-expect-binary/values =)
...
(expect-= 20.0 (ldexp 5.0 2))
(expect-=/values (values 5.0 0.5) (modf 5.5))

Best Wishes,
Kon

P.S. I inherited the architecture from 'test-infrastructure', and  
never made any fundamental changes. One big change would be to shift  
to a 'collector' style for every test form. The collector/collect  
machinery would be invisible, essentially "posting" each test result  
node as it is generated. Kinda late in the game though.

>   -Ivan
>
>
> Kon Lovett <address@hidden> writes:
>
> > On May 24, 2007, at 7:28 PM, Ivan Raikov wrote:
> >
> > >   Thanks for that useful addition. So if the tests require an
> > > additional egg (such as test-infrastructure or testeez), should that
> > > be added as a dependency in the .meta file?
> >
> > Yes, but haven't done it myself yet.
> >
> > (Note 'test-infrastructure' is obsolete.)
> >
> > Best Wishes,
> > Kon

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iEYEARECAAYFAkZX0joACgkQJJNoeGe+5O7p4gCfd4QsZHgL/AlSbz2fAox1IKhP
l4YAoILTxrpAvuI4ip8dXPdbkFpZJQyj
=1ivV
-----END PGP SIGNATURE-----