Requires ClojureScript dependency on classpath
Closed this issue · 5 comments
impl.cljc requires cljs.analyzer and cljs.env for cljs macros.
To avoid this problem, one approach would be:
- split up impl.cljc so that it provides an impl_clj and impl_cljs entrypoint namespace
- require the respective ns using a reader conditional from core.cljc
Keep in mind we need to support :
- clojure compiler targeting clojure
- clojure compiler targeting clojurescript
- clojurescript compiler targeting clojurescript
File extensions and reader conditionals are about the compiler language, but what we want is to conditionally require based on the target language.
https://github.com/cgrand/macrovich can be used as a tool or reference (code is very little) to avoid potential problems, however the clj based cljs compiler only reads clj for macroexpansion, if required by previously read cljs.
Hello,
I would also love to use Missionary and build a standalone uberjar without 20 Mb of Clojurescript/Google cruft.
I have tried macrovich but was not able to make the code usable from clj and cljs at the same time 🤔
I have tried the core.async solution, which is to create a separate namespace. Obviously a big disadvantage is the huge duplication of code. But at least, everything works correctly. You can see my commit on branch split-cljcljs-duplicate.
I have tried to dynamically require Clojurescript dependencies when available to provide a dedicated analyze implementation function. No code duplication but it may be a bit slower (could we use caching?). You can see my commit on branch split-cljcljs-dynamic.
I have tested my two branches with an external simple project and everything seems to work fine.
What do think about those solutions?
Hello @leonoel,
Using the cloroutine branch with dynamic loading and moving missionary clojurescript dependency to the cljs-test alias I was able to run missionary Clj tests and Cljs tests (clojurescript & planck) without errors.
The dynamic loading solution look promising. What do you think?
Fixed in 12