Hey folks,

Some of the committers attended the Activate 2019 conference, which took place 
in Washington, DC on Sep 10-13.

The schedule was packed, so we managed to only have a ~1hr meeting during a 
lunch break - nonetheless, I think it was still very productive!

Committers attending (at least some part of) the meeting, in no particular 
order: Erik Hatcher, Anshum Gupta, David Smiley, Gus Heck, Noble Paul, Varun 
Thacker, Ishan Chattopadhyaya, Tomás Löbbe, and yours truly.

Here are the notes I took - those attending, feel free to clear up any errors, 
omissions or misunderstandings.

- Clean up tests that needlessly use AbstractDistribZk…
    - because this class creates a control collection, which in many cases is 
not needed.

- Consider reusing a single MiniSolrCloudCluster instance for multiple test 
suites
    - always use unique collection names per suite / test
    - some suites won’t be able to use this due to a particular setup or 
side-effects (sysprops, expected metrics, etc)
    - those that can should execute much faster

- Deprecations in 8x - we still need to actually remove the stuff from master:
    - old blob store
    - old spatial
    - other things?

- Replace NamedList with MapWriter?
    - avoid creating objects during serialization
    - big undertaking, but transition piece by piece
    - example: ExportHandler / ExportWriter
    - new API should use MapWriter instead of NamedList / Map
    - public API changes have to go through deprecation in 8x and removal only 
in 9

- We have three different and partially incomplete faceting impls
    - do we want to do something about it to reduce confusion and code 
footprint?

- V2 APIs are incomplete, there’s no workflow to maintain them in sync with v1. 
Proposed strategy to improve this:
    - move SolrJ to v2 - this could be done soon
    - move Solr internally to use v2
    - move tests to use v2 by default.
    - RefGuide in 9.0 should show v2 examples by default
    - deprecate v1
    - come up with a better way of creating v2 api metadata (annotations?)

- Promote GitHub-centric approach to dev & collaboration
    - PRs as the main method for submitting contributions 
    - How to Contribute should be the first section of the github page
    - PR is opened - should automatically create a jira if it doesn’t exist yet
    - discourage using patches when code review is expected.
    - PR is more inviting for collaboration than a patch
    - downside: PR is only for a single branch (no backport integration)
    - travis integration?
    - or use Github Actions for automated precommits, tests

- Javadocs, typos, small ref guide changes should not require a Jira issue with 
its overheads

—--

Andrzej Białecki


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

Reply via email to