Add config key for OMERO.searcher storage path

[spli]

In the short term we'll use the ContentDB (python pickle) file which OMERO.searcher stored independently of the OMERO filestore. The path should be configurable (it looks like the OMERO_CONTENTDB_PATH environment variable) isn't passed to scripts which are currently used for feature calculation.

[spli]

I've tried

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE beans PUBLIC "-//SPRING//DTD BEAN//EN" "http://www.springframework.org/dtd/spring-beans.dtd">
<beans>
  <bean class="ome.system.Preference" id="omerosearcher.data.dir">
    <property name="db" value="false"/>
    <property name="mutable" value="false"/>
    <property name="visibility" value="all"/>
  </bean>
</beans>

as ome/services/blitz-config-omerosearcher.xml in lib/server/extensions.jar but the variable isn't readable by a non-admin (conn.getConfigService().getConfigValue('omerosearcher.data.dir') returns a SecurityViolation).

Also tried wrapping the bean in <bean id="preferenceContext" class="ome.system.PreferenceContext"><property name="preferences"><list>.
The XML is definitely read since it fails if there's a syntax error.

[jamoore]

Simon: your bean definition looks very correct. The automatic lookup is not in place. Either a handler needs to be added as in ​https://github.com/openmicroscopy/openmicroscopy/blob/develop/components/server/src/ome/services/scheduler/SchedulerFactoryBean.java or we will need to add a addPreference() method to ​https://github.com/openmicroscopy/openmicroscopy/blob/develop/components/common/src/ome/system/PreferenceContext.java and have this preference inject itself.

If you'd like me to make either of those changes, just say so.

[spli]

I've been having second thoughts on whether this is the best way of doing things. Having to change extensions.jar seems unnecessarily difficult, especially if you end up with more than one add-on from different groups. For now I'll just create a config file in the omero_searcher root dir- this works because omeroweb is on the OMERO PYTHONPATH, so can be found from within a script (import omeroweb.omero_searcher.omero_searcher_config), but maybe this isn't such a bad idea after all? Thoughts?

[jamoore]

It's ok but not great. Moving forward, I would think bin/omero config will always be supported, across nodes, etc. One other option I've considered is making some namespace generally readable: omero_ext.*, omero.public.* or similar. Or perhaps even simpler, provide a way to list prefixes which should be readable: bin/omero config set omero.config.public_prefixes omero_searcher.

[spli]

Referencing ticket #9945 has changed sprint.

[jamoore]

Cleaning up searcher backlog.