Barry Rowlingson wrote: > On Sun, May 24, 2009 at 7:57 PM, Toby HOCKING <toby.hock...@etu.upmc.fr> > wrote: >> Yes this is a bizarre error that I had been getting. It was difficult >> to track down >> exactly where the bug came from, but it was somehow related to the C >> stack limit size, >> which was evidently too small and causing the R process to crash ... >> only when used in >> conjunction with RPy and Django (bizarre I know). The workaround that >> I came up with was >> to use the command "ulimit -s unlimited" in the parent shell of every >> python django >> process. Hope this helps. > > I hit this problem at the weekend. It seemed to occur only when > importing rpy from a django view function. So I tried importing rpy > from settings.py, so it happens once at the start of the django run. > > My settings.py starts like this: > > from rtest.rshim import myR > > Because for some reason I factored out the rpy import into rshim.py > within an app. It looks like this: > > from rpy import r > myR = r > > and then view code that needs rpy does this: > > from project.rtest.rshim import myR as r > > I'm not sure if that level of complexity is strictly necessary. I > think the root cause is starting up R from a view which maybe is > running in some odd thread context.
If threads are involved, I think that rpy will be particularly brittle, with stack errors such as the ones you observed happening. rpy2 should not have the stack issue. > Maybe this works for you too. I'm on rpy 1.x and R 2.9.x and > django-latest-stable using django's built in server. > > I'm not sure what happens if you get multiple requests, but then > django's test server (I think) blocks on each request. In theory, the GIL will be acquired when entering C-level calls and only be released upon exiting. That would make multiple requests queue up to have their R code executed. > Running from > apache with mod-python? Most likely, and C stack errors have been reported in that context have be reported on that list, I think. > I have no idea... I'd do some serious testing > before making any kind of production system with this. > > Or possible > move the R business onto a separate server with Rserver. Definitely an option if objects exchanged both ways (to R and from R) are of "reasonable" size L. > Barry > > ------------------------------------------------------------------------------ > Register Now for Creativity and Technology (CaT), June 3rd, NYC. CaT > is a gathering of tech-side developers & brand creativity professionals. Meet > the minds behind Google Creative Lab, Visual Complexity, Processing, & > iPhoneDevCamp asthey present alongside digital heavyweights like Barbarian > Group, R/GA, & Big Spaceship. http://www.creativitycat.com > _______________________________________________ > rpy-list mailing list > rpy-list@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/rpy-list ------------------------------------------------------------------------------ Register Now for Creativity and Technology (CaT), June 3rd, NYC. CaT is a gathering of tech-side developers & brand creativity professionals. Meet the minds behind Google Creative Lab, Visual Complexity, Processing, & iPhoneDevCamp as they present alongside digital heavyweights like Barbarian Group, R/GA, & Big Spaceship. http://p.sf.net/sfu/creativitycat-com _______________________________________________ rpy-list mailing list rpy-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rpy-list
