[plt-scheme] When is flat-named-contract necessary/useful?

From: Richard Cobbe (cobbe at ccs.neu.edu)
Date: Thu Jul 10 18:54:09 EDT 2008

I have a question about the pragmatics of the contract system in PLT 4.

I've never been completely sure exactly when `flat-named-contract' is
helpful.  Section 7.3.1 of the PLT Scheme Guide suggests that it helps make
contract violation errors more readable, and certainly the second version
of the `myaccount' module from that section does indeed produce a clearer
error message when the contract on `deposit' is violated.

But it doesn't look like `flat-named-contract' is necessary to get the
clearer error message, as the following version of `myaccount' seems to
work just as well:

    #lang scheme

    (define (amount? x) (and (number? x) (integer? x) (>= x 0)))

      [deposit (amount? . -> . any)])

    (define this 0)
    (define (deposit a) (set! this (+ a this)))

Now, the error message that you get on evaluating `(deposit -10)' in a
different module says "expected <amount?>" rather than "expected
<amount>".  Is that the only difference?

Are there other situations in which flat-named-contract gets you more than
an error message without extra question marks?  Perhaps more important from
a practical standpoint, are there situations where writing an arrow
contract (or ->*, or ->d, etc.) and providing a predicate name as a
subcontract will *not* result in error messages involving the predicate's
name where using a contract defined with `flat-named-contract' would?

These questions are related to the question of where the contract system
gets the strings it uses for the expected value in error messages.  While
I've never completely understood this either, we can leave the general
explanation for another time.



Posted on the users mailing list.