There is another large potential gotcha, R is very memory heavy.
I do think the route of using Celery or other job management tools makes
sense, especially if you can use R across multiple backend machines.
Would celery mean one rpy2 per celery? You don't really want all your
users using the same R session anyways.
Thanks,
Alex
On 04/23/2013 11:08 PM, Derek wrote:
Thanks Per-Olof
No, it has more to do with the issue raised here:
https://github.com/Sleepingwell/DjangoRpyDemo/blob/master/README.md#django-configuration
Possibly Celery could solve that (?) but I really would like to hear from
someone who actually has a production configuration set up and working.
Perhaps there are less people in the sciences using Django than I thought...
Derek
On Tuesday, 23 April 2013 21:32:12 UTC+2, Per-Olof Åstrand wrote:
I am not sure I understand your question, but is it really related to
using specifically R? Could it be any kind of heavy number-crunching that
needs to be done in the background by a scheduler/task manager? In that
case, django-celery may be an option:
http://docs.celeryproject.org/en/latest/index.html
Per-Olof
On Monday, April 22, 2013 9:26:05 PM UTC+2, Derek wrote:
Based on googling around this topic, it seems that using RPy2 is the most
common way to interface with R from Python. However all the discussions on
this seem to centre around working in a desktop (single user) environment.
The one discussion I could find that deals with the issue of working with
R "at scale" is this one -
https://github.com/Sleepingwell/DjangoRpyDemo/blob/master/README.md#django-configuration
- which indicates problems with this approach; and suggests it might be
able to be overcome via creating distinct processes dedicated to run a WSGI
application (although this article does not give any steps on how to do
this, or whether it would work in practice).
Another approach seems to be to use RPy2, with Twisted to enable multiple
sessions:
https://docs.google.com/presentation/d/11LJxej6jnbYKzJftpDudYFfVKjaB0BhOzrBSKaxJ2ME/edit#slide=id.p
.
Yet another approach might be to use Rserve (
http://www.rforge.net/Rserve/) and PyRserve (
http://pythonhosted.org/pyRserve/manual.html), but the latter seems to
currently be in beta.
Question is: does anyone have any practical experience actually using
Django with R in a production environment (i.e dozens or hundreds of users
doing high volume number crunching)?
Thanks
Derek
PS Yes, we do need R and not one of the Python-based alternatives, as R
offers many routines simply not available in those as yet (also, the client
needs to re-use, and create new, R scripts themselves)
--
You received this message because you are subscribed to the Google Groups "Django
users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to django-users+unsubscr...@googlegroups.com.
To post to this group, send email to django-users@googlegroups.com.
Visit this group at http://groups.google.com/group/django-users?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
