Re: [Chicken-users] fix for read-string

[felix]

Joerg F. Wittenberger wrote:
> Hi all,
>
> the read-string procedure in extras.scm is broken when it comes to
> length terminated reads (one of those nice off-by-one's).  Here a
> fixed version:
>
> (define read-string
>   (let ([read-char read-char]
>         [write-char write-char]
>         [open-output-string open-output-string]
>         [get-output-string get-output-string] )
>     (lambda (n . port)
>       (let ([p (if (pair? port) (car port) ##sys#standard-input)]
>             [str (open-output-string)] )
>         (when n (##sys#check-exact n 'read-string))
>         (let loop ([n n])
>           (or (and (eq? n 0) (get-output-string str))
>               (let ([c (read-char p)])
>                 (cond [(eof-object? c) (get-output-string str)]
>                       [else
>                        (write-char c str)
>                        (loop (and n (fx- n 1))) ] ) )) ) ) ) ) )

Ok, ok. Fixed. (geez - I'm having trouble keeping up... ;-)

> One remark: I'd tend to use `if' instead of the `or/and' construction
> I used here, but I noticed (in some different context) that `or/and'
> is faster with chicken than `if'.  To me this is something between
> surprizing, strange and suspicious.  I'm not sure whether my test
> benchmark was broken or that's really the truth, anybody who can
> comment?

Well, the code that comes out of different combinations of
if/and/or can influence subsequent compiler optimizations.
I suggest using the construct that shows the programmers
intention most clearly and tune for speed only if it makes
a noticable difference.


cheers,
felix