Steel Bank Common Lisp

On 09/19/2010 09:25 PM, Jerry James wrote:
> With SBCL 1.0.38, on an x86_64 Fedora 13 machine:
>
> * (read-from-string "Is it nil or t?" :start 6 :end 9)
>
> IS
> 3
>
*snip*
>
> I expected that to mean that only the characters between :start and
> :end would be examined.  Yet GCL 2.6.8pre, ECL 10.4.1, and Clisp 2.47
> all behave the same way on this machine, so ... what am I missing?

You're missing the wonderfully nonintuitive interaction between 
&OPTIONAL and &KEY. In your call, :START is taken as the value of 
ERROR-P, and 6 as the value of EOF-ERROR. Try this instead:

(read-from-string "Is it nil or t?" t nil :start 6 :end 9)
--
Brian Mastenbrook
brian@...
http://brian.mastenbrook.net/

On 20 September 2010 05:36, Brian Mastenbrook <brian@...> wrote:

> You're missing the wonderfully nonintuitive interaction between
> &OPTIONAL and &KEY. In your call, :START is taken as the value of
> ERROR-P, and 6 as the value of EOF-ERROR. Try this instead:
>
> (read-from-string "Is it nil or t?" t nil :start 6 :end 9)

Since pretty much everyone has been bitten by this at one time or
another, I went and added a style-warning about this mistake to SBCL
as of 1.0.42.46.

Cheers,

 -- Nikodemus

Nikodemus Siivola <nikodemus@...> writes:

> On 20 September 2010 05:36, Brian Mastenbrook <brian@...> wrote:
>
>> You're missing the wonderfully nonintuitive interaction between
>> &OPTIONAL and &KEY. In your call, :START is taken as the value of
>> ERROR-P, and 6 as the value of EOF-ERROR. Try this instead:
>>
>> (read-from-string "Is it nil or t?" t nil :start 6 :end 9)
>
> Since pretty much everyone has been bitten by this at one time or
> another, I went and added a style-warning about this mistake to SBCL
> as of 1.0.42.46.

Nice.  Thanks.   


By the way, if the user defines a function named like the few CL
functions that take both &optional and &key parameters, could you
disable the warning about them?

(shadow 'read-from-string)
(defun read-from-string (string &optional eof-error-p eof-value &key start end preserve-whitespace)
   ...)

should not give that warning IMO.
 

-- 
__Pascal Bourguignon__                     http://www.informatimago.com/