From: "Ara.T.Howard" <ahoward@...>
Date: 2004-04-17T01:24:11+09:00
Subject: Re: What's "io/nonblock"?

On Fri, 16 Apr 2004, Michael Granger wrote:

> On Apr 16, 2004, at 7:15 AM, Gavin Sinclair wrote:
> 
> > On Friday, April 16, 2004, 10:50:55 PM, Gavin wrote:
> >
> >> Ara's "session" package tries to require 'io/nonblock' and fails on my
> >> machine.  Google reveals nothing very instructive.  Anyone know what
> >> it is?
> >
> > [...]
> > installing io/wait
> > make[1]: Entering directory `/home/gavin/Projects/ruby18/ext/io/wait'
> > make[1]: Nothing to be done for `install'.
> > make[1]: Leaving directory `/home/gavin/Projects/ruby18/ext/io/wait'
> >
> > Problem, that.
> 
> A problem, perhaps, but io/nonblock.rb is installed from 
> ext/io/wait/lib/, and is, as its name suggests, a plain Ruby file. It 
> doesn't appear to require io/wait, either, so you should be okay with 
> just installing it by itself to get 'session' working.

would do - but __i__ require io/wait as well.  the truth is i currently don't
really need both: my original design used nonblock, but this is broken in
threads, the lastest uses IO.ready?, which you get from io/wait:


from the source

<snip>
        # this is the desired call - but threads block on it!
          #iodat[:io].nonblock{ buf = iodat[:io].read }

        # 2.1.5 
          #sleep(0.042) and Thread.pass until iodat[:io].ready?
          #buf << iodat[:io].getc while iodat[:io].ready?

        # 2.1.6
        # decided to go ahead and read lines since
        # - loop only terminates when cmd_end_pat has been found and this
        #   must be newline terminated
        # - although this will block mid-line in the case of parital lines we
        #   know that, eventually, a newline will appear and this delay is
        #   acceptable so we can make larger (line buffered) reads instead of
        #   character ones.  eventually i'd like to use io.nonblock{} to read 
        #   'as much as is there' but this does not play well with threads at the
        #   moment
          buf << iodat[:io].gets while iodat[:io].ready?
</snip>


the semantics i'd like to have are 'read as much data as is possible, even a
single character, but don't block on the read'.  this is what

  iodat[:io].nonblock{ buf = iodat[:io].read }

should do. but in a thread it blocks.  there was a few postings on this 2
months ago by me, nobu i think, and tanaka akira if you feel like searching...

in any case = you could acutally commnet out the 'require "io/nonblock"' and
it would be fine.  but i hope to use this in future versions...

weird that you were lacking it - my old 18 seems to have it...

oh well.

cheers.

-a
-- 
===============================================================================
| EMAIL   :: Ara [dot] T [dot] Howard [at] noaa [dot] gov
| PHONE   :: 303.497.6469
| ADDRESS :: E/GC2 325 Broadway, Boulder, CO 80305-3328
| URL     :: http://www.ngdc.noaa.gov/stp/
| TRY     :: for l in ruby perl;do $l -e "print \"\x3a\x2d\x29\x0a\"";done 
===============================================================================