Hi Matthias
Thanks for the explanation - it helps. I played with the kind of
post-condition "broken contract" example that you gave, and find it less
problematic, since the blame is local to the function / supplier.
If the error in my original example was reformatted by moving the at clause
to the top (easy?) and supplemented with the line number of the guilty
client [or supplier] code (hard?) I think it would reduce the confusion and
improve the utility. I.e.
twice: contract violation
at: .../blame-test.rkt - line 2, column 18
expected: number?
given: 'foo
in: the 1st argument of
(-> number? number?)
contract from: (function twice)
blaming: .../blame-test.rkt - line 6, column 1
What do you think?
Kind regards
Dan
On Mon, Sep 2, 2013 at 11:48 PM, Matthias Felleisen <matth...@ccs.neu.edu>wrote:
>
> On Sep 2, 2013, at 8:22 AM, Daniel Prager wrote:
>
> Consider this file:
>
> --------------------------------------------
> #lang racket
> (define/contract (twice x)
> (-> number? number?)
> (* 2 x))
>
> (twice 'foo)
> --------------------------------------------
>
> When run in DrRacket 5.3.6 this yields the following error:
>
> twice: contract violation
> expected: number?
> given: 'foo
> in: the 1st argument of
> (-> number? number?)
> contract from: (function twice)
> blaming: .../blame-test.rkt
> at: .../blame-test.rkt: 2.18
>
>
> The final line attributes blame to the twice function (line 2, col 18),
> but clearly the blame in this instance lies with the client code (line 6),
> consistent with the first six lines of the error message.
>
>
> This is a misinterpretation of the last line. If you read the entire
> message as a sentence, you will see that the "at: ..." means "this is where
> you find the contract that was broken". The "expected:"/"given" pairing
> means that the function was called, it expected a number, and the caller
> applied it to the symbol 'foo instead.
>
> If you run this program
>
> #lang racket
> (define/contract (twice x)
> (-> number? number?)
> 'foo)
>
> (twice 2)
>
> the error message will say
>
> twice: broke its contract
> promised: number?
> produced: 'foo
> in: the range of
> (-> number? number?)
> contract from: (function twice)
> blaming: (function twice)
> at: unsaved-editor665:2.18
>
> meaning that foo was called, it promised to return a number, but it
> produced the symbol 'foo instead.
>
> ;; ---
>
> I will admit that you're not the first one who has tripped over this error
> message [format] and that you aren't the last one. But changing it isn't
> easy and adding information isn't easy. The problem is that we wish to keep
> the messages uniform across all functions but not all functions can benefit
> from the actual contract system.
>
> -- Matthias
>
____________________
Racket Users list:
http://lists.racket-lang.org/users
