efficient event execution on prefix lookup

[tsloughter]

I'd been meaning to try this to see what the implementation looked like. Annoying that it requires that recursive function to build the match spec, matchspec should have some sort of operation that lets you specify something is a prefix or EventName ++ ['$tail'] that results in the match being a tail and not a single element... but I digress.

I'd be curious in benchmarks. I'm not really expecting this to be accepted but I thought it better to actually do the comparison instead of assuming we can't do it.

My understanding was that the original attempt at doing prefix matching was still using a bag? Meaning it was scanning the whole table, so particularly slow. In this version I get around needing a bag by appending the handler id to the event name, doesn't change the lookup since it is a prefix lookup.

Or was the original attempt also an ordered_set and still proved to be far too inefficient to be viable?

[binaryseed]

My thought is that I'm hoping that this won't really be needed, we just need to stick to the guidance of static event names + dynamic metadata for further filtering by the handlers.

[arkgil]

@tsloughter I just found the original implementation and you're right, it used a bag. I agree with @binaryseed, though, I think this ship has sailed and switching to prefixes now will bring confusion as to what is the "right" approach: prefixes or metadata.