I've used Integrant (Integrant REPL) and donut-party/system approaches commercially, so I could do a talk on those. They are also available with new projects created with https://practical.li/clojure/clojure-cli/projects/templates/practicalli/ I've taken a repl reloaded approach with both approaches. I've also done simpler projects without any component library.

That would be perfect, the talk could illustrate the differences and the http://practical.li templated/articles could be used as follow up. I will DM you to organise it.

Haven’t got the system to prove it, but at some point I wish I were ready to give a talk called something like “KISS: Coding without dependency injection”. Perhaps @U04V5V0V4 could help me?

ooh ooh yeah ... I'm up for that heresy

file watcher + a way to reload is usually good enough ... a lot depends on whether you are starting services yourself and often it's just one (the web server)

I guess my claim (and perhaps @U04V5V0V4s) is that for most systems, your stateful things are just a handful, and you don’t need to build a graph to sort out their dependencies. With system like Component et al, you’re tricked into making way more stateful things than need be.

Nailed it

I've retro-fitted integrant into a few projects that grew from small-and-too-simple-for-DI into bigger systems that had trouble either starting up or restarting reliably. Each time it was a royal pain, with fairly large refactorings that did not add any business features. My default advice is to just use integrant on new projects...

Definitely something that requires consideration- in my case there is a business value since majority of services do not cleanly shutdown resources when they’re being stopped during rollouts - we could add to the mess and setup JVM shutdown hooks of course but that just adds to the mess in few places. And yeah, I retrofitted Component before, it was painful but paid off in the end

This came up recently and I’m :star-struck:

https://medium.com/@maciekszajna/reloaded-workflow-out-of-the-box-be6b5f38ea98

@U9RRPR0KE we are not worthy. This is genius.

Following up this request we are organising the following event, Thanks to @U05254DQM The London Clojurians are happy to present: *Data Driven Component Libraries from commercial experiences (by Johnny Stevenson)* https://www.meetup.com/London-Clojurians/events/301916548/ Johnny will be talking about how to use effectively *Integrant, donut-party/system and Juxt/Aero* with a REPL workflow.

this kind of works but Calva complains about missing closed parens

(def pac4j-clients ^Clients (doto (Clients.)
                              (.setCallbackUrl "/callback")
                              (.setClients ^Client/1 (into-array Client [pac4j-form-client]))))

If I'm reading the code of Clojure itself correctly, it's because expressions like [...] are tagged with IPersistentVector, which does not extend List. If so, you should be able to make it work by using (.setClients ^List [...]).

there is a bug, I believe still open, where you cannot hint literal collections

> Calva complains about missing closed parens Sounds like something Calva should fix. :)

> there is a bug Ah. Well, let's bind it then.

or call identity on it

I think if you qualify the setClients it should work, param-tags are only used on qualified methods

seems like this did the trick, thank you all

(^[List] Clients/.setClients [pac4j-form-client])

looks kind of strange though 🙂

Yeah, feels redundant here in the doto but it’s part of minimizing new semantics for existing syntax