Stream: t-compiler/wg-nll

Topic: rationale of differential-dataflow


Tatsuyuki Ishi (Nov 21 2018 at 08:25, on Zulip):

I was studying incremental computation and I found that Polonius used to be based on differential dataflow.

I wonder if the computation has anything to do with incremental-ness? Or if it's the other way around, how was differential dataflow useful for that case?

(I haven't fully understood Datalog yet, apology for that.)

nikomatsakis (Nov 21 2018 at 18:59, on Zulip):

differential dataflow just happened to be a very fast datalog engine; we were not using its incremental capabilities, though one can imagine doing that in the future

Last update: Nov 21 2019 at 14:15UTC