[Haskell-cafe] Re: FFI and callbacks

[John Goerzen]

On Thu, Jul 21, 2005 at 11:07:15AM +0400, Bulat Ziganshin wrote:
> Hello Simon,
> 
> Thursday, July 21, 2005, 1:16:10 AM, you wrote:
> 
> SM> from a single thread.
> 
> you can either:
> 1) made all calls from single thread
> 2) put all calls in "withMVar lock", where `lock` is a global MVar

OK, that makes sense.

> JG> 1. It seems that there is no function that says "block the current
> JG> thread until the thread given by ThreadId dies"
> 
> GHC concurrency stuff is very basic and low-level. there is several
> libs which adds some more high-level abilities, including MissingH (see
> Child.hs) and some functions in http://www-i2.informatik.rwth-aachen.de/~stolz/Haskell/CA.hs

Err, got a bit of egg on my face for that one.  Would have hoped I'd
have remembered what's in the library I maintain ;-)

But actually, looking at it now, it's not a plain ThreadId that it waits
for, but rather a Child, which is:

data Child a = Child ThreadId (MVar a)

Great if you use the Child.hs (which, for the record, is one of Peter
Simons' works), but if somebody isn't using it, it may be nice to have
another option.  I'll look into Simon's recipe and see if I  can
understand it ;-)

FWIW, I think Python has some really great threading primitives.  From a
brief look at what's in Haskell, I think they are all implementable in
terms of the low-level features Haskell has (I'll trust you all on the
waiting for a thread thing).  I may write something up along those
lines, stick it in MissingH and offer it up for fptools if anyone is
interested.

Also, I think it's annoying that ThreadId is not showable.  This could
be very useful for debugging.

Python's stuff is at
http://www.python.org/doc/current/lib/module-threading.html

-- John