`span` doesn't work well with “local input streams”

[augfab]

Hi,

I'm using :com.inuoe.jzon v1.1.4 on SBCL 2.2.9 on Ubuntu 24.04.2 on x86_64.

I don't know exactly what to call those “local input streams” created with constructs such as cl:with-input-from-string or flexi-streams:with-input-from-sequence; in any case, I was expecting the following to work, alas it doesn't:

CL-USER> (cl:with-input-from-string (input "{}") (com.inuoe.jzon:span input :end 2))
; Evaluation aborted on #<TYPE-ERROR expected-type:
;              (OR FILE-STREAM SYNONYM-STREAM BROADCAST-STREAM)
;              datum: #<SB-IMPL::STRING-INPUT-STREAM {10042F0C63}>>.
CL-USER> (flexi-streams:with-input-from-sequence (input #(91 93)) (com.inuoe.jzon:span input :end 2))
; Evaluation aborted on #<TYPE-ERROR expected-type:
;              (OR FILE-STREAM SYNONYM-STREAM BROADCAST-STREAM)
;              datum:
;              #<FLEXI-STREAMS::VECTOR-INPUT-STREAM {1004F53D63}>>.

The error seems to come from line 1083 of src/jzon.lisp (in definition of span):

        (unless (<= 0 start end (file-length in))

In SBCL file-length is very strict about the type of its argument.

Am I doing something wrong? (Relative newcomer to Common Lisp 😅)

Thanks for your work on this package 👍

[augfab]

This came up while I was hacking a little program to do some simplistic RPC, the parent process runs the program in a subprocess, sending it the size (number of bytes as little-endian 32-bits unsigned integer) followed by the JSON payload; the subprocess responds with size+payload; rinse and repeat.

I was interested in using com.inuoe.jzon:span to constrain the size of the data to read from the input stream 😄

[augfab]

The problem came up while testing out small inputs at the REPL. I've found a way to solve the problem, I read-sequence into a vector of bytes of the right size, then call com.inuoe.jzon:parse on this vector. This works from the REPL, so I don't have a problem anymore!

Not sure the original use case is interesting or important to support, feel free to close the issue if you think it isn't.

Thanks.

[Zulu-Inuoe]

Sorry @augfab ! You're right that this should Just Work. I'm surprised I hadn't run into it before. Good catch on the file-length bit