Clojurians Log v2

Hi All! org.onyxplatform/onyx-seq is deprecated?

@lellis I’m not sure on the status of the onyx-seq repo, but Onyx core has a plugin for operating over seq’s under onyx.plugin.seq. There are task bundles under onyx.tasks.seq.

Ty @gardnervickers

Ah yea the last released version is for 0.9.x where the plugin model changed so it’s for backwards compatibility for users on 0.9.x still. If you’re using newer versions of Onyx you’ll want the one in core.

@lellis The separate artifact is, yeah.

Use the version provided by core.

There is any example using onyx.plugin.seq? I have some difficult to send seq to my job, it never reach my task next to :in . Its in lifecyle right?

Check the tests for Onyx -- they use it

Gotta run for a bit.

I'm looking at https://github.com/PyroclastIO/metamorphic/blob/master/README.md, do my events have to be provided in order to match correctly?

I can see that rt/process-event can take a timestamp as an extra parameter, but it's only mentioned in the context of time-windowed constraints

@danielcompton Hi! At the moment, yes. The particular algorithm I used is order sensitive. From the research I did before implementing it, virtually all algorithms for CEP are order sensitive. Some can fudge it by replaying events, but that can also be done at the application level, so I decided not to attack it.

ok, that makes sense, I can handle replays at the application level

This is for re-frame-trace https://github.com/Day8/re-frame-trace where we want to interpret streams of events to get the higher level context

But the traces can come out of order as they are sent when the call finishes, so nested calls are 'before' the parent calls

We can pretty easily detect this though, and rewind and rerun the match state

Cool - that makes sense. 🙂

CEP is a rather tricky topic as far as implementation, but it's super useful as a library.

Took a few months to get it right.

Yeah I was embarking on doing this and thought, I should really try metamorphic before I invent my own hand-rolled solution

The biggest problem I had was that the papers outlining the design and implementation of each algorithm are incredibly dense.

It's also a class of problem with very high memory utilization, which I guess accounts for what I just complained about.

Yeah I can imagine, having a look at some of the papers now

https://web.cs.wpi.edu/~chuanlei/papers/sigmod17.pdf <- This one is a more sophisticated version of the one I chose.

I decided not to reach that far.