[gambit-list] A rudimentary pipeline implementation with bridging threads

Jussi Piitulainen jpiitula at ling.helsinki.fi
Mon Jan 28 06:07:16 EST 2013


There's (shell-command "date | tr 0-9 D") that works but requires me
to build a string for the shell to parse and I don't think it lets me
control the input to and output from the pipe from Gambit.

Then there's
(read-line (open-process '(path: "sh" arguments: ("-c" "date | tr 0-9 D"))))
that probably does what I want except again I don't want to build that
string in the first place.

I want to do in Gambit what the shell does when it launches a
pipeline, while being able to write to and read from the resulting
pipeline. I know how to mix the two, but my goal is to use gsi
_instead_ of the shell.

(And no, I don't mean to create a cycle. Just a straight pipe.)

Thanks.


Mikael <mikael.rcv at gmail.com> writes:

> So you have a Gambit process A that you want to launch Unix processes B and
> C, such that A outputs to B, which outputs to C, which outputs back to A?
> 
> Probably you can just rely on Unix' piping primitives for that, something
> akin to open-process `(app: "/bin/sh" args: ("B" "|" "C"  or "B | C")
> ..redirection.. #t ?
> 
> 2013/1/28 Jussi Piitulainen <jpiitula at ling.helsinki.fi>
> 
> > Assuming my previous question came through, I asked about implementing
> > pipelines of processes in Gambit-C. The problem was that I could not
> > seem to make the output of one process to connect to the input of the
> > next on their own.
> >
> > I'm now experimenting with Gambit-C threads to bridge those gaps.
> > Here is a rudimentary implementation of what I want. The comments
> > indicate things that I think I know I still need to do. I haven't used
> > threads much before, and never in Scheme.
> >
> > The original question stands: am I overlooking some obvious way to do
> > such pipelines that Gambit-C already has? Because this is the easy 80%
> > of the work and another 80% sheer tedium probably remains :)
> >
> > (define (echo from to) ;should use a buffer
> >   (do ((line (read-line from) (read-line from)))
> >       ((eof-object? line))
> >     (display line to)
> >     (newline to)))
> >
> > (define (piece from to)
> >   (lambda ()
> >     (echo from to) ;should allocate a private buffer for echo
> >     (close-output-port to)  ;stopped using it as output port
> >     (process-status from))) ;wait for it - maybe time it out?
> >
> > (define (end from) ;not closing current output port
> >   (lambda ()
> >     (echo from (current-output-port))
> >     (process-status from)))
> >
> > ;actually, the end should be controlled by the called altogether! or
> > ;is it now?
> >
> > ;caller needs to close the first pipe segment as output port, and do I
> > ;need to force some output or close input ports? and errors should be
> > ;handled, including diagnostic output from the processes and failed
> > ;status of any, and not leave processes around on any sort of error if
> > ;possible. oh and do all those threads terminate nicely and always?
> >
> > (define (pipe first . rest)
> >   (let ((first (open-process first)))
> >     (do ((last first (car rest))
> >          (rest (map open-process rest) (cdr rest)))
> >         ((null? rest)
> >          (thread-start! (make-thread (end last)))
> >          ;;^altogether mistaken - leave it to the caller
> >          (values first last))
> >       (thread-start! (make-thread (piece last (car rest)))))))
> >
> > (define (test) ;has hardwired end to echo the output in current output
> >   (call-with-values ;^fix it so the called handles reading and waiting
> >       (lambda ()
> >         (pipe "date" '(path: "tr" arguments: ("0-9" "D"))))
> >     (lambda (first last)
> >       (close-output-port first))))
> >
> > _______________________________________________
> > Gambit-list mailing list
> > Gambit-list at iro.umontreal.ca
> > https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
> >



More information about the Gambit-list mailing list