More over-select benchmarks have been added in suite. So we can measure what is the cost of over selecting.
All setup (creating tables and inserting data) is moved to separate setup scripts which are shared across benchmarks.
I’ve triggered backfill for Sequel, after which I will trigger for Rails.
We only have two raw benchmarks rewritten so far – I am looking forward to rewrite other.
Is something up with this bench
How can AR be faster?
Hm, raw is 2x faster for me locally. I will take some time to inspect this issue.