A guess all my talks are Assums?

Oh, and good morning!

Morsning korsning!

good, bright and crisp morning

thanks to @borkdude and @mpenet for an eduction education this week

So with @borkdude’s and @slipset’s help I sort of get some of the rationale for transducers. Still need to understand the mechanics. I found the first course in a series about transducers by Tim Baldridge. But it didn't even reach the point where a transducer was used. 😃 I'll check that reference page out first and see where that lands me.

@pez the original talk is probably the most informative one: https://www.youtube.com/watch?v=6mTbuzafcII

morning

’allo!

I really loved Tim's transducer explanations. That's when it all clicked for me.

I mostly just use transducers where I would have used ->>

someone wrote a clj-kondo hook for that: https://github.com/borkdude/clj-kondo/issues/323#issuecomment-691247062

I always worry that there is a lurking problem with eductions

(deftype Eduction [xform coll]
   Iterable
...
   clojure.lang.IReduceInit
...
   clojure.lang.Sequential)

@ben.hammond What's the catch? eductions can behave as iterable objects, but also as efficiently reducible objects.

(clojure.lang.TransformerIterator/create xform (clojure.lang.RT/iter coll))

creates a little unexploded Exception to be passed around

good point. what's an example of a coll that's not a sequential, but does satisfy IReduceInit?

well anything that you created using

(reify IReduceInit

but I'm conflating two seperate issue here • Don't create eduction with colls that do not support Iterable is not the same as • How do you distinguish IReduceInits that are reusable from those that are one-shot

reify does support implementing multiple protocols/interfaces, but I guess you could forget to implement both

I worked on a project where we had a Cognitect consultant who did exactly this; it creates an issue because when their Clojure-fu so far outstrips the team's it takes quite a while to figure out the problem

I guess you could also implement your own type on top of IReduceInit: (deftype OneShot [xforms coll] IreduceInit ...) which you can then feed to (into [] (OneShot. ...))

and you could similiarly create (deftype EductionReduceOnly

problem comes in when the logging framework • sees the Sequential on the eduction tries to use it to write a log string an blows up your server OR • correctly uses the one-shot IReduceInit, but thus exhausts the reducible so its not available to do the real work (for example, all sorts of insidious problems can arise from this)

good moaning

there seems to be missing an emoji for > moan is 😩 the closest we can get?

maybe 😬

which (interestingly) looks like a USB-C plug

Maybe a picture of an English man pretending to be a French police officer?

Ha! I've now rewritten a few transduce examples to ->> . A bit backward, you might think, but it seems to help me figure out how to use transduce. Thanks @otfrom!

🙂

I think I probably use into more than transduce, but I usually just want to do something horrible to a seq and have it turn into a vector

Yeah, while I do want to have transduce in my tool chest, the most important thing to me right now is to be able to read more Clojure code.

I do find that into and transduce help me to create and read more clojure code. It makes it easy to create composable transformation pipelines