Steel Bank Common Lisp

Cyrus Harmon <ch-sbcl@...> writes:

> I'm exploring the relatively new bivalent streams in SBCL and I am  
> trying to do the following:
>
> (defun bivalent-stream-test-1 ()
>   (with-open-file (foo #p"foo"
>                        :direction :output :if-does-not-exist :create
>                        :if-exists :overwrite
>                        :element-type :default)
>     (format foo "bogus!~&")
>     (write-byte 131 foo)
>     (write-sequence #(137 80 78 71 13 10 26 10) foo)))
>
> The format and the write-byte succeed, but the write-sequence fails  
> with the error:
>
> The value 137 is not of type CHARACTER.

Does it work if you use a vector specialized on (unsigned-byte 8)
rather than a (vector t)?

Cheers,

Christophe

Why, yes it does. Interesting. I'd argue that the default behavior  
should be as in the ub8 case, but this might help.

Should there be some sort of field in the stream struct for bivalent- 
p or some such?

Thanks,

Cyrus

On Sep 7, 2005, at 9:21 AM, Christophe Rhodes wrote:

> Cyrus Harmon <ch-sbcl@...> writes:
>
>
>> I'm exploring the relatively new bivalent streams in SBCL and I am
>> trying to do the following:
>>
>> (defun bivalent-stream-test-1 ()
>>   (with-open-file (foo #p"foo"
>>                        :direction :output :if-does-not-exist :create
>>                        :if-exists :overwrite
>>                        :element-type :default)
>>     (format foo "bogus!~&")
>>     (write-byte 131 foo)
>>     (write-sequence #(137 80 78 71 13 10 26 10) foo)))
>>
>> The format and the write-byte succeed, but the write-sequence fails
>> with the error:
>>
>> The value 137 is not of type CHARACTER.
>>
>
> Does it work if you use a vector specialized on (unsigned-byte 8)
> rather than a (vector t)?
>
> Cheers,
>
> Christophe
>

Cyrus Harmon <ch-sbcl@...> writes:

> Why, yes it does. Interesting. I'd argue that the default behavior  
> should be as in the ub8 case, but this might help.

  (write-sequence #(#\a #\b #\c) stream)
  (write-sequence #(1 2 3) stream)
and, worst of all
  (write-sequence #(#\a 2 #\c) stream)
are all potentially valid operations on bivalent streams... we don't
support them as yet, though.

> Should there be some sort of field in the stream struct for
> bivalent-p or some such?

Dunno.  We have managed to do without it so far...

Cheers,

Christophe