Personally, I'd go with a nightly build for your situation - just
makes it easier (for me at least) to support and fix if there are any
issues, and you'll benefit from great new features as well (stats
component, java replication, etc).
The one drawback is if any of those patches don't keep with trunk and
become a hassle to apply. The SOLR-284 patch is not likely (-0.5 from
me as-is) to get committed anywhere near where it is now, so consider
it a risky one to rely on too strongly.
As for the custom query parser plugin... you can drop that in within a
JAR in <solr-home>/lib so no need to patch Solr locally for that.
Erik
On Nov 4, 2008, at 7:20 PM, Chris Harris wrote:
My current pre-production Solr install is a 1.3 pre-release build, and
I think I'm going to update to a more recent version before an
upcoming product release. Actually, "release" is probably a bit of an
exaggeration; it's more of an alpha test, or perhaps a beta test.
Anyway, the question is which more recent version of Solr I should be
running. I'm not under pressure from on high to stick with an official
Solr release, so all these seem like legit possibilities for me:
Run the 1.3.0 release
Run a more recent build from the 1.3 branch
Run a nightly build of the trunk
Obviously, I would attempt to do sufficient testing before putting
Solr live regardless of which route I chose.
One factor is that I need to run a slightly modified Solr, as opposed
to a 100% out-of-the-box install. Currently I'm using these patches:
https://issues.apache.org/jira/browse/SOLR-538 (copyField maxLength
property)
https://issues.apache.org/jira/browse/SOLR-284 (parsing rich
document types)
https://issues.apache.org/jira/browse/SOLR-744 /
https://issues.apache.org/jira/browse/LUCENE-1370 (Patch to make
output a unigram if no ngrams can be generated)
I also may need to have a custom query parser plugin.
Any ideas?
Cheers,
Chris