@jeremyevans this is happening cause
@sgrif was not using the sequel-pg gem and I think you have correctly opted just to use your c implementation as opposed to the pg gem coder implementation that is slower. Everyone who uses sequel really should use sequel-pg. I do not think this is cheating.
sequel is not defer materializing so you pay the cast cost upfront
Is there any way to configure sequel to defer materialize cause as it stands it can introduce an enormous cost for people who migrated sequel.
For context this is the corrected bench:
Raw blows both sequel and activerecord out of the park being 5x faster due to defer materialization. But once you select all the columns sequel and raw are mighty close.
I know it is a battle, choosing defer materialization just to work around users writing inefficient db access code, but over selection is an epidemic in the Rails world, it is not going away.
@bmarkons @sgrif pointed out that the benchmark RubyBench - Long Running Ruby Benchmark is incorrect cause it inserting nils into the date columns, we got to fix that asap! (also the title on that page needs love)