On Fri, Oct 18, 2019 at 08:16:36PM -0000, Michael Kay mike@xxxxxxxxxxxx scripsit:
> Yes, that's how accumulators work. To quote the spec:

Thank you!

I had somehow formed the impression that if one had multiple huge
documents, they could be stream-processed in the context of a single
accumulator.

I did read the spec. The pre-existing notion had me wondering if the
evaluation description meant "what happens to the accumulator-rule match
attribute" rather than "the scope of the accumulator as a whole".

Thanks!
Graydon