thread safe implementations

[Umberto Nicoletti]

Hi Dan,

On Thu, 2004-11-04 at 16:12, Dan Barron wrote:
> Umberto,
>
> I'm not going to try to recreate the functionality of the cgi - I'm not
> fully sure yet what absolute path I will take - I really have not done much
> threaded java code.  I've looked at the java mapscript - and your map
> server wiki entries for compiling java mapscript - I'm not convinced I have
> a good compile yet (platform is Fedora Core 2).  Is the java mapscript
> viable at this point?  I have read posts from earlier this year saying it
> looked abandoned?

I think it has rather improved since last year. I have done some load
testing and did not found evident issues, even though threading errors
are rather difficult to spot.

A random thought on this: disabling the use-threads compile flag and
running the load tests afterwards shows many errors in the log files.
With use-threads enabled the errors go away.

>
> I would really like to get to an all java solution if possible - I
> currently use a combination of java/php (java opens a connection to a php
> script then parses the returned data) - not the best but it seems to work
> fine under light loads.
>
> I'd be interested to hear more about what you are up to - it sounds
> interesting.  I would really like to just build a threaded front end (maybe
> a thread pool?) to the map server functionality via java mapscript and
> nothing more.

See my previous posts for this. I might be able to put a demo together
if some others are interested. But be warned: it is just mapserver as a
Java webapp.

Regards,
umberto

>
> Thanks,
> Dan
>
> At 11:51 PM 11/3/2004, unicoletti at prometeo.it wrote:
> >Dan,
> >if by creating a java implementation you mean writing a web app to deliver the
> >same (or  subset) of the functionality the cgi does, then I am writing one.
> >
> >I think I am one of the few java mapscript users around :-)
> >
> >Regards,
> >umberto
> >
> >Quoting Sean Gillies <sgillies at FRII.COM>:
> >
> > > Dan,
> > >
> > > It's not clear to me what you mean by creating a java implementation
> > > on your own.  Implementation of what?  Do you mean a programming API
> > > like the php mapscript module?
> > >
> > > If this is what you mean, you should look at the Java stuff that
> > > you can generate using SWIG.  In mapserver versions 4.2.4+ and in
> > > the 4.4.0 beta1 there is a Java makefile under mapserver/mapscript/java.
> > > It seems to work, but as I've said -- I'm not a Java user and so am
> > > not beating on the Java mapscript module heavily enough or often
> > > enough to say anything about its readiness.
> > >
> > > Sean
> > >
> > > On Nov 3, 2004, at 10:37 AM, Dan Barron wrote:
> > >
> > > > Hmmm - I was really hoping someone had done (or attempted) a java
> > > > implementation - I'm not versed in Zope or Python.  Well, I guess I
> > > > will
> > > > attempt to create a java implementation on my own.
> > > >
> > > > Thanks Sean.
> > > >
> > > > Dan
> > > >
> > > > At 01:10 PM 11/2/2004, Sean Gillies wrote:
> > > >> On Nov 2, 2004, at 10:47 AM, Dan Barron wrote:
> > > >>
> > > >>> Hello,
> > > >>>
> > > >>> Just curious what the current consensus and recommendations are for
> > > >>> creating a thread safe front end or wrapper for mapserver in either
> > > >>> php or
> > > >>> java? I'm assuming that mapserver is still not thread safe - that
> > > >>> correct?
> > > >>>
> > > >>> Thanks!
> > > >>>
> > > >>> Dan
> > > >>
> > > >> Dan,
> > > >>
> > > >> I caught and fixed three sources of potential threading problems
> > > >> before
> > > >> the 4.2.4 release and these fixes will also be in 4.4.0.  I think
> > > >> we're
> > > >> close to thread safety, but it's hard to say without more testing.  As
> > > >> far as I know, my ZMapServer (Zope + Python mapscript) software is the
> > > >> only thing out there that is using MapServer in a multi-threaded
> > > >> environment.  It works, but is not a complete proof of thread safety.
> > > >>
> > > >> There remains much work to be done to in improving MapServer's
> > > >> threaded performance.  The single lexer for parsing mapfiles and
> > > >> class expressions is a bottleneck that will be addressed in the
> > > >> future.
> > > >>
> > > >> cheers,
> > > >> Sean
> > > >>
> > > >> --
> > > >> Sean Gillies
> > > >> sgillies at frii dot com
> > > >> http://users.frii.com/sgillies
> > > >
> > >