@mattly http://blog.muhuk.com/2015/10/01/who_in_their_right_mind_would_use_monads_in_clojure.html#.WJorvhIrLUI

I’ll add this wonderful little series of short posts: http://www.clojure.net/2012/02/02/Monads-in-Clojure/

@bill yeah I gave the talk on cats a few months back https://github.com/funcool/cats

that http://muhuk.com link is 404ing

as it appears their site is offline

@mattly Google cached: http://webcache.googleusercontent.com/search?q=cache:_ngeoQGUtOMJ:blog.muhuk.com/2015/10/01/who_in_their_right_mind_would_use_monads_in_clojure.html+&cd=1&hl=en&ct=clnk&gl=us#.WJotGRIrLUI

wish I hadn’t missed your Cats talk

the code /comments are here: https://github.com/mattly/pdxclj-cats-promesa/blob/master/src/preso.clj

tx!

I tried to walk through the "why" before the "what"

perusing…

I use "maybe math" as an example at the beginning because I've ended up having to implement some variant of that in pretty much every language I've ever worked in

@mattly looking more closely at your code example… https://clojurians.slack.com/files/mattly/F410FHD6V/mount_patterns__old_and_new.clj I see what you did there. All dependency on state is encapsulated in a defstate. Code in defstate can call functions so long as they reference no variables def’d outside the fn. (was gonna say “pure” fns but um, this is Clojure)

heh

a lot of the documentation / writeups for mount encourage you to put the behavior in the start/stop clauses of defstate

once I realized that was a bad idea and stopped doing it, it became a lot easier to manage

So back to Mount vs. Component: the point about Mount defining a bunch of singletons is apt. My fave example of that is an app that needs to read from one PostgreSQL db and write to another one. In Component, I’d just have two (differently-configured) instances of my db component and I’d be off to the races. In Mount this won’t work because you configure the database namespace once (to talk to the database) and you’re stuck. My take is that this is what Yurt is for. Yurt uses Boot Pods to provide separate isolated worlds. Communication between these Pods looks like it’d be cumbersome.

then again, w/ your scheme perhaps I could easily define two states: one for each db, initialized from two sets of config settings This would entail moving that state creation (for the db connections) outside the db namespace into some higher-level (app-specific) one

yeah that's more or less what I do

I have two postgresql databases and a mysql database I work with

all at once?

config validation and pool creation are its own concern

yep

as well as spark, cassandra and dynamodb

it's a mess

wow

but that mess is why I was able to push for clojure on this project 😄

in what is otherwise a ruby shop

step 1: create mess, step 2: propose awesome solution, step 3: profit!

I didn't create the mess tho

😄

the only stores I own outright are spark and one of the pg databases

the rest are Other Services' Data

and some of it is HIPPA-protected