This is my favorite channel.
π
Hello friends, so I finished to implement all protocol for making requests
I had some tests checking body generation
Would be great to check headers and URI in the near futur
I propose to create merge it in the official repo on a branch.
It lacks deserialization yet, I have to think about it, if anyone has ideas
hmh, somehow loosing ability to concentrate on reading blog posts. I realized that I only skimmed Matthew's post, and he demonstrated Application CDS too (creating memory mapped archive of user classes)
I have faint memory that class data sharing originated from the jdk build that Apple used to maintain but I might be way off (idea was to lower memory use by sharing read only memory mapped base classes, similar to what native libraries might do)
(case protocol
"rest-xml" (params-to-body-rest-xml req)
"ec2" (params-to-body-ec2 req)
"query" (params-to-body-query req)
"rest-json" (params-to-body-rest-json req)
"json" (params-to-body-json req)))
neat :) doing deser right now
itβs going well for json/rest-json
offcourse xml is a pain in the ass
and I have to put back documentation done by Mr Kimmo himself π
I wonder if there's anything we could share with rhe Cognitect library
:D
did I document something? , oo :)
Iβd really love to see some cljs working some day
yeah, there's a issue foe cljs on the aws-api too
and the ser/deser design rely on pure datastructure
hmm, yeah
which mean itβs possible there is nothing to touch in sdk itself
only signing and request
http calls
@viesti, I think ou tried the cognitect one, what do you find better / worst ?
hmm, data.xml is cljs too?
yep, but I did not tried it
they don't gen functions, rely on "data centric" callsite, (aws/invoke s3-client {:op :ListBuckets})
which is better or kind if the same
I kinda found it neat at repl
but would imagine say Cursive having hard time
at the repl, one can ask for supported ops, and request/response shape
ok
like spec exercise
but there are specs, so they should be navigable via a spec browser like cider
the lib is broken into many artifacts
having only one is a bit easier in the end
ok
anyway my point was also to finish an OSS of a big size
yup
if nobody use it, itβs fine π
yeah :)
might be even less stressful :)
but at least it will be truly open-source π
yes
haven't really yet read on SubtrateVM nor Graal, but this sounds eerily close to portkey abstract interpretation: https://github.com/oracle/graal/blob/master/substratevm/REFLECTION.md
whatβs your idea behind all this ?
maybe nothing useful, just that the bytecode analysis in portkey could be used for generating a list of classes for the native-image tool in subtrate/graal
could make native image bulding easier, if in the repl, one could provide entry point function and the pass classes resolved from that point onto native image packager
but this is just thinking out loud
ok
I did not look much at all the serverless stuff recently
seems a lot is happening
re:Invent happened and aws opened up the runtime ability of lambda
but there's stuff aside aws too, that I don't know much of
ok
which means you can create aws for your own language right ?
but in aws, one can now have a custom runtime for any language. You poll events from a local http api, and push back reponses. The runtime just invokes your binary.
ok
and one can share runtimes via layers, Ghadi haa done this https://github.com/ghadishayban/lambdaclj-jdk11
so even though current java aws lambda is on jdk8, you can ship your own version now
hmm, realizing I need sleep
good night! :)
good night
xml -> formatted map
this was not fun π