Inefficient result filtering

[Steve McLellan]

After running the elasticsearch search, we pass the results through filter_results for each plugin in case there are sensitive fields or data transformations need doing in v1/search.py:

            for plugin in self.plugins:
                result = plugin.obj.filter_result(result, req.context)

As the number of plugins (hopefully) grows, this gets less and less efficient, since all but one plugin will be a no-op for any given result but we're looping through the entire (potentially large) result set. Better would be handling the dispatch earlier; build up a dictionary of index-type -> filter_result function and run the results through it.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/207682

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/207682
Committed: https://git.openstack.org/cgit/openstack/searchlight/commit/?id=8404053e03ad9db723ffe99c0fd197205d6c150f
Submitter: Jenkins
Branch: master

commit 8404053e03ad9db723ffe99c0fd197205d6c150f
Author: Steve McLellan <email address hidden>
Date: Thu Jul 30 20:25:36 2015 -0500

    More efficient result filtering

    Instead of passing each result through each plugin, maintain
    a backwards lookup of (index, type) -> plugin and only filter
    a result through the plugin it belongs to. Changes the signature of
    filter_result which only glance/image is currently using.

    Change-Id: I617dbddc44dd07303bdd6fe9794fa92484a7eb32
    Closes-Bug: #1478102