[Haskell-beginners] Named Pipes Et Cetera

[matthew coolbeth]

Hello, all.

Three questions:

1> What, if anything, can I do in a Haskell program to do blocking
reads from a Linux FIFO? For the curious, there is a small program
below that doesn't block, but immediately terminates because of an EOF

2> How could I have found the answer to this question on my own? (The
documentation that Google found for me was not completely adequate)

3> Is this sort of question generally regarded as too unixy or not
haskelly enough to be appropriate for this mailing list?

Thanks.

--BEGIN PROGRAM--
module Main () where

import Control.Monad
import System.IO
import System.Posix.Files

pipeString = "/tmp/dld.fifo"

main :: IO ()
main = do
createNamedPipe pipeString $ unionFileModes ownerReadMode ownerWriteMode
pipe <- openFile pipeString ReadMode
forever (hGetLine pipe >>= putStrLn)

--
mac
_______________________________________________
Beginners mailing list
Begi...@haskell.org
http://www.haskell.org/mailman/listinfo/beginners

[Patrick LeBoutillier]

Matthew,

Does it return immediately when you start it or only when you try to
write to the pipe with some client?

Here's what I get on my Linux box:

$ rm /tmp/dld.fifo && ./pipe &
[1] 8603
$ echo -e "1\n2\n3" > /tmp/dld.fifo
1
2
3
pipe: /tmp/dld.fifo: hGetLine: end of file

Patrick

--
=====================
Patrick LeBoutillier
Rosemère, Québec, Canada

[matthew coolbeth]

That's interesting. This was using my source code?
On mine (also a Linux box), it exited immediately, giving me no
opportunity to write to the pipe.

[Magnus Therning]

On Tue, Nov 23, 2010 at 12:55, matthew coolbeth <mac0...@engr.uconn.edu> wrote:
> Hello, all.
>
> Three questions:
>
> 1> What, if anything, can I do in a Haskell program to do blocking
> reads from a Linux FIFO?  For the curious, there is a small program
> below that doesn't block, but immediately terminates because of an EOF
>
> 2> How could I have found the answer to this question on my own? (The
> documentation that Google found for me was not completely adequate)
>
> 3> Is this sort of question generally regarded as too unixy or not
> haskelly enough to be appropriate for this mailing list?

The behaviour is as expected from reading the documentation on
hGetLine (http://hackage.haskell.org/packages/archive/haskell98/latest/doc/html/IO.html#v:hGetLine).
The same seems to be true for all functions in IO, so I suspect
you'll have to go down to a lower level where you have more control
over this behaviour, e.g. by using th fd* functions in
System.Posix.IO.

/M

--
Magnus Therning                      OpenPGP: 0xAB4DFBA4
email: mag...@therning.org   jabber: mag...@therning.org
twitter: magthe               http://therning.org/magnus

[Magnus Therning]

On Tue, Nov 23, 2010 at 14:09, Patrick LeBoutillier
<patrick.le...@gmail.com> wrote:
> Matthew,
>
> Does it return immediately when you start it or only when you try to
> write to the pipe with some client?
>
> Here's what I get on my Linux box:
>
> $ rm /tmp/dld.fifo && ./pipe &
> [1] 8603
> $ echo -e "1\n2\n3" > /tmp/dld.fifo
> 1
> 2
> 3
> pipe: /tmp/dld.fifo: hGetLine: end of file

Would you run 'uname -a' and 'ghc --version'?

I don't see that behaviour here

% uname -a
Linux bryma 2.6.35-ARCH #1 SMP PREEMPT Sat Oct 30 21:22:26 CEST 2010
x86_64 Intel(R) Core(TM)2 Quad CPU Q9450 @ 2.66GHz GenuineIntel
GNU/Linux
% ghc --version
The Glorious Glasgow Haskell Compilation System, version 6.12.3

% rm /tmp/dld.fifo && ./pipe&
[1] 30762
% pipe: /tmp/dld.fifo: hGetLine: end of file

[1] + exit 1 ./pipe

/M

--
Magnus Therning                      OpenPGP: 0xAB4DFBA4
email: mag...@therning.org   jabber: mag...@therning.org
twitter: magthe               http://therning.org/magnus

[Daniel Fischer]

I get the same with 6.12.3 and 7.0.1.

$ uname -a
Linux linux-mkk1 2.6.27.54-0.1-pae #1 SMP 2010-10-19 18:40:07 +0200 i686
i686 i386 GNU/Linux

>
>
> /M

[Patrick LeBoutillier]

On Tue, Nov 23, 2010 at 9:13 AM, matthew coolbeth
<mac0...@engr.uconn.edu> wrote:
> That's interesting.  This was using my source code?

Yes, as is. I compiled it with "ghc --make pipe.hs".

I'm running Fedora 10 x64 with:
$ ghc --version
The Glorious Glasgow Haskell Compilation System, version 6.10.1

Patrick

[Patrick LeBoutillier]

Here's the uname -a output:

$ uname -a
Linux pc207 2.6.30.10-105.2.23.fc11.x86_64 #1 SMP Thu Feb 11 07:06:34
UTC 2010 x86_64 x86_64 x86_64 GNU/Linux

Patrick

>>> Rosem�re, Qu�bec, Canada

>>>
>>
>>
>> --
>> mac
>>
>
>
>
> --
> =====================
> Patrick LeBoutillier

> Rosem�re, Qu�bec, Canada
>

--
=====================
Patrick LeBoutillier
Rosem�re, Qu�bec, Canada

[Brandon S Allbery KF8NH]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 11/23/10 07:55 , matthew coolbeth wrote:
> Hello, all.
>
> Three questions:
>
> 1> What, if anything, can I do in a Haskell program to do blocking
> reads from a Linux FIFO? For the curious, there is a small program
> below that doesn't block, but immediately terminates because of an EOF
>
> 2> How could I have found the answer to this question on my own? (The
> documentation that Google found for me was not completely adequate)

This isn't specific to Haskell, but rather to Unix/POSIX FIFOs. In general,
they often don't do what you want. :)

The short version is: the "server" end should open read-write, because the
last close by a writer will send an EOF and tear down the connection so that
it can't be reused without closing and reopening. This does mean you can't
actually receive EOFs if you intend to block on the FIFO, so design whatever
protocol you plan to use with this in mind.

That said, it sounds like GHC's I/O manager is handling this incorrectly;
what you wrote should in fact block waiting for a writer, output everything
it reads, then exit when the writer goes away. If it's GHC7, then likely
someone broke FIFO handling (again... very few people seem to understand the
care and feeding of FIFOs) in the new I/O manager.

- --
brandon s. allbery [linux,solaris,freebsd,perl] all...@kf8nh.com
system administrator [openafs,heimdal,too many hats] all...@ece.cmu.edu
electrical and computer engineering, carnegie mellon university KF8NH
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.10 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkz5TBkACgkQIn7hlCsL25URDQCdGisbjpa65EeE83CZcyXRLuru
Py4AoIB17qhlagBd3JxggakvjV23+6pC
=mphl
-----END PGP SIGNATURE-----

[Brandon S Allbery KF8NH]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 11/23/10 10:36 , Patrick LeBoutillier wrote:
> The Glorious Glasgow Haskell Compilation System, version 6.10.1

IIRC this was in fact broken in 6.10.1; upgrade at least to 6.10.3.

- --
brandon s. allbery [linux,solaris,freebsd,perl] all...@kf8nh.com
system administrator [openafs,heimdal,too many hats] all...@ece.cmu.edu
electrical and computer engineering, carnegie mellon university KF8NH
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.10 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkz5THAACgkQIn7hlCsL25VOEQCgp/fPxHTsFb6vAeaO1GGonZDc
dh8AmgIy4l0khXTHliySwRSnz804tu0Z
=m7l1
-----END PGP SIGNATURE-----

[matthew coolbeth]

As Mr Therning suggested, I switched over from System.IO.openFile to
openFd in System.Posix.

This gives the blocking behavior that I want. The only downside was
having to provide my own readLine.

Just so everyone knows.

thanks,

mac

--
mac