Solr 5 integration - Githubissues

collective / collective.recipe.solrinstance

Buildout recipe to configure a Solr instance

https://pypi.python.org/pypi/collective.recipe.solrinstance

5 stars 13 forks source link

Solr 5 integration #40

Closed saily closed 9 years ago

saily commented 9 years ago

This feature branch includes:

Huge refactoring with a much cleaner buildout recipe.
Single core support for Apache Solr 3, 4 (fake by using collection1 core) including it's templates, see #39
Multi core support for Apache Solr 3, 4 and 5 including it's templates, fixes #36
Move templates into their folders
Descrease loglevel to warning in logging.properties.
Added auto download of Solr releases if solr-location is not set in recipe, this keeps legacy installations working fine.
Added parameter to specify Solr version. People will have to add this parameter to their installations.
Update documentation, add notes to Solr 5 changes (example solr.EdgeNGramFilterFactory)

saily commented 9 years ago

Will fix #32 as well.

jensens commented 9 years ago

I'am not deep enough into solr configuration and multi-core knowledge to fully get all implications, but this PR looks very clean, feels well thought and a is big step foreward, so +1 for merging it.

saily commented 9 years ago

i rebased the whole branch onto master.

loechel commented 9 years ago

Hi, I did an review and some manual tests on this recipe. The changes look very clean as as @jensens already mentioned, but I found several issues that should be solved before merging and a release of an 6.0.0 version.

as a conf for plone is one of the standard use cases the example of collective.solr sould work out of the box: https://github.com/collective/collective.solr/raw/master/buildout/solr.cfg but this didn't work due to two reasons: no solr-version specified and field type integer is not in default schema.xml.
enablePositionIncrements="true" is not supported by Lucence 5.0 anymore, this in template prevents solr form accepting schema.xml for core
schema reference synonyms.txt but non ist created by default.
default-core parameter is not used in Solr 5 case. should be documented.

By the way, Plone did not work with generated schema as collective.solr.parser fails due to order of definitions in schema.xml

loechel commented 9 years ago

I did some fixings on the mentioned problems. Please review and check if that is consitent with intended cleanup (@saily).

saily commented 9 years ago

Hi thanks for reviewing.

Adjusting the solr.cfg of collective.solr is another task which may follow this merge. Several things in that file did not work in past, for example solr-min-ram and solr-max-ram was never implemented.
I didn't catch the enablePositionIncrements issue, thanks for that.
There was never a synonyms.txt file created out of a template. This was just copied over, which should work in the same behavior now. So implementing a new synonym.txt template, target's a new feature which i did not want to implement through this pull, because we're changing a lot anyway.

saily commented 9 years ago

Ok, since we copy over all sample data shipped with solr now, the given schema.xml files should always work. There are some new locations (lang folder) and naming conventions for stopwords.txt but for now we should just ignore the new location and use the old one, people may have customized their templates.