Also see proc-doc and friends for a "creator 3" possibility! Robby
On Saturday, April 20, 2013, Matthias Felleisen wrote: > > Creator 1: Sure enough this was a miscommunication. The creator of > file->string added the check in the its implementation and then used the > check to document the function. If the check is inside the function, as is > the case here, there is no cross-checking between the docs and the code. > > Creator 2: However, Racket provides a contract system. If the creator of > file->string had used it, he would have done it this way: > > (provide > (contract-out > (file->string (-> path-string? string?)))) > > and in principle, we could then have documented the function straight from > this export statement, meaning the documentation and the code would be in > sync. > > People who act like creator 1 tend to write docs as if they had acted like > creator 2, writing docs as if you had used the contract system. Even if you > are a creator 2 kind of person, you may manually create the docs with a > copy of the contract instead of linking it directly to the contract. > (Java's doc system kind of forces the reuse of types in the docs, so we're > a step down from there in this regard.) > > -- Matthias > > > > > > On Apr 20, 2013, at 10:40 AM, Manfred Lotz wrote: > > > On Sat, 20 Apr 2013 10:19:54 -0400 > > Matthias Felleisen <matth...@ccs.neu.edu <javascript:;>> > > wrote: > > > >> > >> On Apr 20, 2013, at 12:37 AM, Manfred Lotz wrote: > >> > >>> My understanding is that parameter path-string? will be > >>> checked at run time but only if the programmer who coded e.g. > >>> file->string did call path-string?. > >> > >> > >> No, the function will check pathness. Try file->string with a string > >> that contains \0. > > > > Either I did not understand or I just did not express myself correctly. > > > > file->string uses file->x > > > > (define (file->string f #:mode [mode 'binary]) > > (file->x 'file->string f mode read-string string-append)) > > ..... > > and file->x > > > > (define (file->x who f file-mode read-x x-append) > > (check-path who f) > > ..... > > > > checks explicitly for path-string? via check-path > > > > (define (check-path who f) > > (unless (path-string? f) > > (raise-argument-error who "path-string?" f))) > > > > > > If the programmer would have forgotten to code check-path in file->x > > then there would be no check. Therefore my understanding is that the > > language itself cannot enforce the predicate (as it has not a type > > system like Haskell or Ocaml). This is that enforcement of the > > predicate relies on proper programming. > > > > Or did I miss anything? > > > > > > -- > > Manfred > > > > > > > > > > > > > > > > ____________________ > > Racket Users list: > > http://lists.racket-lang.org/users > > > ____________________ > Racket Users list: > http://lists.racket-lang.org/users >
____________________ Racket Users list: http://lists.racket-lang.org/users
