hierarchy_tests

A sandbox app for testing performance of querying hierarchies in postgresql.

This is not meant to be a production app, just a way to benchmark the performance of different hierarchy/tree gems.

Current gems being tested:

• ancestry
• closure_tree

If you want to see any added, raise an issue, or even better add it and create a PR.

Setting up

Currently, you DO NOT want to setup the db with rake db:migrate and rake db:seed. Some of the migrations were hacked together to get sample data loaded, and if you run them in order right now, things will not come out right.

For now, you can setup the sample db with:

$ rake db:restore

Running Benchmarks

A set of benchmarking tests for read operations can be called with rake (after restoring the database):

$ rake benchmark:read

No other benchmarking tests have been written yet. After restoring the db, tests can be run in the rails console.

Notes

In the interest of making setup faster, and knowing this is not planned for production, a few ugly things have been done so far. If you hate any of these enough to fix them and create a PR, please do:

• Unnecessarily a full Rails app
  • Left it this way, in-case anyone wants to try to make benchmarking reports available over the web, but this really is a database/activerecord benchmarking app
• The SELECT INTO in a migration to create and seed table
  • This one is particularly bad, and I plan to fix it, so the creation remains in the migration, and the seeding happens in the seed file
  • You have to do hacky things right now to use the migrations and seed file (just restore the db for now)
• The model names Employee and Worker are arbitrary and don't indicate the gem being used
  • Should be switched to AncestryEmployee and ClosureTreeEmployee
• A postgresql dump file is checked in
  • It took many hours (almost a day with restarts) to seed the db with ~300K records, if you want to test with that many records of fake data, but want to restore in seconds/minutes instead, you're welcome.
  • This probably belongs on S3 where many sample databases can be made available without blowing up this repo.
• Currently tied to postgres
  • I always use postgres, so I don't have plans to change this, but a PR to support other db's is welcome.
• The sample data is arbitrary, not based on a real company, too many levels, not enough levels, too many records, not enough records, not enough variety in paths, or otherwise garbage.
  • You can see the logic for creating the sample data in db/seeds.rb. Tony Stark has hundreds of thousands of descendants - there are 16 levels, the org is bottom heavy with ~186K "leaf" employees and one "root" employee.
  • I welcome any real data that can be made publicly available

To Do

• Implement the benchmarking rake tasks
• Make the migrations and seeding more coherent (may cause a breaking change)
• Make the sample db available on S3 (or similar) instead of in the repo