• Sources: primary, discussion
  • Summary: The announcement states the streaming engine becomes the default in 2.0, and that joins, group_by and unpivot no longer guarantee row order. Restoring the previous behaviour takes an explicit maintain_order argument on the operation or a process-wide engine affinity setting. The build announced is a release candidate rather than final 2.0.
  • Why it matters: A pipeline that reads row order out of a join, group_by or unpivot changes behaviour on upgrade without raising an error.
  • Follow-up: Track the final 2.0 release and whether the default engine choice changes before it.

send feedback on this story