I've recently started my debian packaging efforts at [1].
One major issue seems to be testing: I'm trying to run solrpy's tests at
package build time, but this turns out to be non-trivial: Apart from
changing the port to 8080, these tests require the included schema.xml
to be known to solr, but of course I can't install it to a solr
configuration directory such as /etc/solr/conf created by debian's
solr-common package. Neither can I just hijack the running instance of
the installed servlet container (be it jetty or tomcat) for testing (or
can I?)
I've already inquired at the solr-user mailing list about this and
received some answers [2], but the suggestions seem to be a bit over me.
Suggestions include running an embedded solr server [3] (as is possible
with SolrJ, but doesn't seem to be available with solrpy), or
configuring multiple tomcat instances [4].
I myself am pretty much a newbie to both solr and Tomcat, so I wonder if
anyone on this list can tell me how to apply advice like [3]} or [4]
from the solr-user ML to solrpy -- if e.g. using an embedded solr server
is recommendable, would it be easy to implement it in solrpy and modify
tests accordingly?
Kind Regards and TIA,
Bernhard
[1]
http://svn.debian.org/viewsvn/python-modules/packages/python-solrpy/trunk/
[2]
http://www.mail-archive.com/solr...@lucene.apache.org/msg42681.html
[3]
http://www.mail-archive.com/solr...@lucene.apache.org/msg42851.html
[4]
http://www.mail-archive.com/solr...@lucene.apache.org/msg42847.html
We used java -jar start.jar because you can't really embed a java
servlet in a python process (unless you are running it on jython).
IIRC there is a way to send the configuration to a started servlet.
Sorry but I don't recall more information, but I can look it up if no
one answers it to you.
Just to say, but I never ever use debian packages for pure python
modules, I think the way to go is to always install things using pip
and even better in a virtualenv.
--
Leonardo Santagada