Eventmachine talk - Jan 2008 - Handling events over a socket with no delimiter?

Hi guys,

Is there a sanctioned way of modifying the socket-level loop (in Ruby)
which eventually passes chunks of data to receive_data()? I''m having
to parse a protocol that doesn''t always use a delimiter between
events: sometimes it''s "\r\n\r\n" and sometimes it wants you
to
read_nonblock(1) until nothing else exists on the wire (effectively
raising EAGAIN), then handle it.

Yes, it''d help if you felt sorry for me.  ;)

Jay Phillips

so it doesn''t block read after sending "\r\n\r\n"?
Maybe hack the c code? :)
Take care.

On Jan 8, 2008 10:52 AM, Jay Phillips <jicksta at gmail.com>
wrote:
> Hi guys,
>
> Is there a sanctioned way of modifying the socket-level loop (in Ruby)
> which eventually passes chunks of data to receive_data()? I''m
having
> to parse a protocol that doesn''t always use a delimiter between
> events: sometimes it''s "\r\n\r\n" and sometimes it wants
you to
> read_nonblock(1) until nothing else exists on the wire (effectively
> raising EAGAIN), then handle it.
>
> Yes, it''d help if you felt sorry for me.  ;)
>
> Jay Phillips
> _______________________________________________
> Eventmachine-talk mailing list
> Eventmachine-talk at rubyforge.org
> http://rubyforge.org/mailman/listinfo/eventmachine-talk
>


--
-Roger Pack
For God hath not given us the spirit of fear; but of power, and of
love, and of a sound mind" -- 2 Timothy 1:7

On Jan 8, 2008 10:52 AM, Jay Phillips <jicksta at gmail.com> wrote:
> Is there a sanctioned way of modifying the socket-level loop (in Ruby)
> which eventually passes chunks of data to receive_data()? I''m
having
> to parse a protocol that doesn''t always use a delimiter between
> events: sometimes it''s "\r\n\r\n" and sometimes it wants
you to
> read_nonblock(1) until nothing else exists on the wire (effectively
> raising EAGAIN), then handle it.
>
> Yes, it''d help if you felt sorry for me.  ;)
How do you know when using this protocol if you look for \r\n\r\n or
if you keep reading?  How do you know that there really is no more
data, and that isn''t just a delay caused by something out on the
network?


Kirk Haines

From: "Jay Phillips" <jicksta at
gmail.com>
>
> Is there a sanctioned way of modifying the socket-level loop (in Ruby)
> which eventually passes chunks of data to receive_data()? I''m
having
> to parse a protocol that doesn''t always use a delimiter between
> events: sometimes it''s "\r\n\r\n" and sometimes it wants
you to
> read_nonblock(1) until nothing else exists on the wire (effectively
> raising EAGAIN), then handle it.
It''s not OpenNap, is it?  :(

What you''re describing sounds similar to the absolute horror I
recall seeing in that protocol, where they believe that if they
send data over TCP, that when the remote end select''s true for
read, all the data will have arrived in one "packet".

I read about that years ago, and I still go into fits of nausea
every time I think about it.

But... if what you''re asking about is like that, what is the
snag with the EventMachine semantics?  Doesn''t receive_data get
called as soon as the reactor gets data from the Kernel?  And
doesn''t receive_data pass along all that data that was available
at that point?

I''m not understanding the problem, I guess...


Regards,

Bill

LineAndTextProtocol is designed to do that sort of thing... it lets you flip
it out of line mode into binary mode

There''s also LineAndText2 but I don''t really understand what
it does
differently.

On Jan 8, 2008 10:52 AM, Jay Phillips <jicksta at gmail.com> wrote:
> Hi guys,
>
> Is there a sanctioned way of modifying the socket-level loop (in Ruby)
> which eventually passes chunks of data to receive_data()? I''m
having
> to parse a protocol that doesn''t always use a delimiter between
> events: sometimes it''s "\r\n\r\n" and sometimes it wants
you to
> read_nonblock(1) until nothing else exists on the wire (effectively
> raising EAGAIN), then handle it.
>
> Yes, it''d help if you felt sorry for me.  ;)
>
> Jay Phillips
> _______________________________________________
> Eventmachine-talk mailing list
> Eventmachine-talk at rubyforge.org
> http://rubyforge.org/mailman/listinfo/eventmachine-talk
>


--
Tony Arcieri
ClickCaster, Inc.
tony at clickcaster.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080108/296e2f1e/attachment.html

> But... if what you''re asking about is like that, what is the
> snag with the EventMachine semantics?  Doesn''t receive_data get
> called as soon as the reactor gets data from the Kernel?  And
> doesn''t receive_data pass along all that data that was available
> at that point?
receive_data receives ''up to but not necessary all of kernel
data'' --
at least on mac os x it tends to receive the size of the receive
buffer, I think, so no conflict, but if there is somehow a lot of
input, it will read some of it call receive_data, then reloop and call
receive_data again.  That part of the code you might be able to tweak,
but it might be nontrivial.

Whoops missed the bit about a nonblocking read at the end... what the crap
protocol?  Sounds like it should be using UDP or something...

On Jan 8, 2008 11:46 AM, Tony Arcieri <tony at clickcaster.com> wrote:
> LineAndTextProtocol is designed to do that sort of thing... it lets you
> flip it out of line mode into binary mode
>
> There''s also LineAndText2 but I don''t really understand
what it does
> differently.
>
>
> On Jan 8, 2008 10:52 AM, Jay Phillips <jicksta at gmail.com> wrote:
>
> > Hi guys,
> >
> > Is there a sanctioned way of modifying the socket-level loop (in Ruby)
> > which eventually passes chunks of data to receive_data()? I''m
having
> > to parse a protocol that doesn''t always use a delimiter
between
> > events: sometimes it''s "\r\n\r\n" and sometimes it
wants you to
> > read_nonblock(1) until nothing else exists on the wire (effectively
> > raising EAGAIN), then handle it.
> >
> > Yes, it''d help if you felt sorry for me.  ;)
> >
> > Jay Phillips
> > _______________________________________________
> > Eventmachine-talk mailing list
> > Eventmachine-talk at rubyforge.org
> > http://rubyforge.org/mailman/listinfo/eventmachine-talk
> >
>
>
>
> --
> Tony Arcieri
> ClickCaster, Inc.
> tony at clickcaster.com



--
Tony Arcieri
ClickCaster, Inc.
tony at clickcaster.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080108/9ad3d98d/attachment-0001.html