Clojurians Log v2

looks great! left a comment

well that was easy - the bundle stuff just works - very minor change

I could make a project that used react & react-dom no config, run webpack - dev and advanced worked fine

the way this is done - zero impact on any existing tooling - since it doesn't really change anything at all - it just makes it easy to run webpack on our output

Do you have one of these repos publicly as an example?

not yet - there's a couple of other UX things I'd like to think through

but this is really zero config

{:main cljs-bundle.core :output-to "out/index.js" :output-dir "out" :target :bundle}

that's it, :bundle handles all the options you need for you

the output file can be handed over to whatever bundler that understands require that you want to use

no custom index.js, no manual :global-exports

Because of figwheel and shadow I’ve never used webpack beyond the simple example on http://clojurescript.org. So was interested in the whole thing if you already had it made

But very excited for this direction. Thanks for the thinking and the work

to clear the above is bigger than just usability as I've said before

you can now publish a CLJS JAR that use node_modules and people can consume it and expect it to work reproducibly

across all tools

note this means also making something that works for React Native (outside the REPL bits) is trivial

because it's exactly the same as webpack

it's definitely something I found disconcerting about the existing CLJS RN integrations (except probably shadow) - overly complicated because there wasn't access to a few important things

RN will be huge. For live reloading, will that now be down to the bundler to provide? And do I need to run clojurescript in auto rebuild to get that?

in the case of RN (i.e. Metro) nothing - the CLJS tooling will need to do it

but you can now focus on the parts that differentiate CLJS from JS dev

CLJS works as JS tools expect for JS deps

but I think a bigger impact for example will be shared wrapper for Reagent

every Reagent RN project has this boilerplate around mapping RN components to Reagent

that can just go away - you can make a dep now that people can just use

what I'm looking really forward to is seeing people push more stuff that may be locked up in their apps that could be open source / reusable but there wasn't a practical way for it to work for a wide number of users

@mfikes we should probably repurpose the :aot-cache stuff for project local files

we end up doing a lot of work when switching between :none and :advanced which is uninteresting

Oh, right. :advanced implies a lot of build-affecting options change.

I think with that in place + the global :aot-cache I think most of the time could be spent only in Closure for advanced builds

{:main cljs-bundle.core
 :output-to "out/index.js"
 :output-dir "out"
 :target :bundle
 :bundle-cmd {:none ["npx" "webpack" "--mode=development"]
              :default ["npx" "webpack"]}}

added :bundle-cmd to master - with that in place running a :none build w/ REPL and :advanced build feels more or less like ClojureScript

I think the only feature that you will lose going down this path is optimal code-splitting, but after all the work that we've put into it still feel like something only hardcore user reach for.

so I think it's a fair tradeoff - curious what folks think that are using CLJS directly w/ code-splitting and foreign libs (who would like the new convenience of the bundle approach)