Post #2280786
2025-10-29 21:35 UTC
@sminez@hachyderm.io @cliffordheath@mastodon.social @robpike@hachyderm.io If you have the restriction of just a single top-level chain, then I think for many use cases you would be able to greedily proceed with processing once the top-level match completes. Some sort of lazy evaluation (like the `refer` database example from http://doc.cat-v.org/bell_labs/structural_regexps/se.pdf) where you can process each matching sub-stream as it completes.
Replies (1)
-
@oac@fosstodon.org 2025-10-29 21:39
@sminez@hachyderm.io @cliffordheath@mastodon.social @robpike@hachyderm.io And perhaps with some cleverness you could support multiple top-level chains if you had the ability to know when every top-level chain completed one or more matches up to a given position in the stream, then proceed with processing those matches. The no overlapping matches restriction though is something I need to understand more. Perhaps that would simplify things? Or it's a symptom of the fact that the problem is really more complicated than I think...