kennknowles commented on PR #39619:
URL: https://github.com/apache/beam/pull/39619#issuecomment-6147100103
Spending a lot of time up front in order to make flattened straight-line,
unrolled coders, is almost certainly the right direction.
Given today's tech:
- I think we could pretty reasonably get fully flattened and concreteized
generated coders (these were always possible, but hard to implement and hard to
get right, before)
- Test the generator against exact input/output matching of arbitrary "slow
version" reflective coders.
- Run the above test as a fuzz test of arbitrary generated schemas.
If we can pass a few seconds of fuzz testing of arbitrary schemas, the
chance that there's a weird one hiding is very low. We could also do exhaustive
testing up to some depth (in other words move from 'quickcheck' model to
'smallcheck' model) and this would give a ton of assurance.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
