Clojurians Log v2
babashka-sci-dev 2022-03-14
@borkdude When you have a moment, I wanted to see if this all sounds reasonable to you for the "read & cache pod namespaces on installation" concept:
1. For each pod declared in bb.edn:
a. Download and install it if necessary
b. Look for pre-existing namespaces-cache.edn file in cache dir and load it if found
c. If no cache file found, run pod, send :describe op, load returned namespaces, write out cache file, & shut down pod
d. Make the first usage of any of the pod's vars start it back up and leave it running like it does now on (load-pod ...)
assuming that all sounds reasonable, can you give me a pointer to the load-fn where you thinking it would make sense to wire up the "we're using this pod's namespace, so start it up" logic?
load-fn is managed in babashka.main
we could pull that out to a different namespace, but it's there now
yes
ok so perhaps after looking for built-in namespaces, we look in common/ctx in a new :pods-namespaces key or similar for a closure that starts up the pod?
yes, when the namespace doesn't exist as a normal dependency, we look into the cached namespaces of each of the pods in bb.edn
until found
the order should not matter much since namespace should be unique
ok
I think I understand the concepts now, so I'll try coding it up. thanks!
we could extract the namespaces into a separate file so we have describe.edn and namespaces.edn , the latter can just be a flat list we look into
for faster lookup than parsing the bigger file
yep
or we could even store it as some more optimal format, but this is an optimization, I think .edn will be fast enough for this right now
as most projects have like maybe 1-3 pods
yeah that seems right
a,b,c = sounds good d = I would say on any of the pod's namespaces, load the pod
so we only really need to cache the namespaces, but we could just cache the entire describe response, you never know what it will be good for right?
yeah makes sense