@devn didn't know about Electron. I think Hoplon targets browser apps. Don't know if it is possible to decently interface with js lib for desktop apps.

Did you know about https://github.com/Gonzih/cljs-electron

@micha @piotrek it doesn't answer your question but today I needed to reuse rpc fn, passing message-cells as params.

like so

(defn conj-db! [success-cell error-cell process-cell]
  (mkremote 'app.api/conj-db!
            success-cell error-cell process-cell))

And then I call it in a hl page, "closure style"

(defc form-send-success nil)
(defc form-send-error nil)
(defc form-send-process [])

((rpc/conj-db! 
   form-send-success
   form-send-error 
   form-send-process)
 form-data)

Untested, but that was what I had in mind

So I can reuse rpc for multiple pages, each of which have their own message-cells

Again untested

((get-widget-data result-cell errors-cell loading-cell) date-from date-to)

@alandipert: thanks! actually, making information go in two directions is something i might need a hand with later

@alandipert: also, not sure if you remember, but you gave me this snippet ages ago to help me get up and running on heroku

(deftask build-heroku-war
  "Build a war for deployment"
  []
  (comp
    (hoplon)
    (cljs :optimizations :advanced)
    (prerender)
    (uber :as-jars true)
    (web :serve 'app.handler/app)
    (war)
    (target :dir #{"target_prod"})))

how do I get that to use http-kit?

apparently that is needed for sente to work

this is what my local dev looks like

(comp
    (watch)
    (serve
      :port    8000
      :handler 'app.handler/app
      :init 'app.network.core/start-router!
      :cleanup 'app.network.core/stop-router!
      :httpkit true)
    (speak)
    (hoplon)
    (reload)
    (cljs)
    (target :dir #{"target_dev"})))

@micha that navy video is cool, i think @meeli might start using it in a training course she runs on FE development (she already mentions hoplon )

@alandipert: mostly, i noticed that the web task doesn’t take the same parameters as the serve task

do you have a servlet container that runs http-kit to deploy to?

the war file is the servlet part that is loaded into a servlet container like jetty or tomcat

that is the web server basically

if you want to run http-kit i think you'll need to deploy an uberjar or something like that, not a war file

@thedavidmeister: ^^^

this navy computer video is incredibly well produced

these animations and illustrations are so much better than the powerpoint garbage of today

the concept of making a continuous computer out of gears, also awesome

no need to settle on your number system or endianness

unitless values

you still need to decide if negative numbers are to the left or right

lol

that's true, or if negative is even a thing lol

i guess also you need to scale it to the bounds of your inputs

it's like a nomogram in this way

there is a sweet video about "planetary gears" from the army too

that's how they make the tank transmissions for the tracks

like how you can make one set of tracks go in reverse and the other ahead etc

ingenious gears

it's sweet the way training videos break down the concept into themes

like The Working Surface

like aspects of the problem relevant to the domain

so you have perfect working vocab for this subset of things about gears

i was looking at how you machine gears

there is a ton of nomenclature there

"dimetrical pitch" is a good one

oh cool

i can imagine being a machinist

that's basically what we do

like their job was to construct machines that make parts of other machines

like to machine a gear you need to construct a jig

functional form = jig

so you can cut the gear accurately

well like "tooling"

we make computer programs to make parts of computer programs

true

i was thinking in the language itself though

mostly machinery to support the very limited machine that solves the problem

well like if you want to build a really complicated machine, like the navy targeting computer

you need to build a bunch of other machines first

that help you manufacture the parts

and of course those are built with other machines

right

pretty much the same thing we do

unfortunately because it's cheap to store and ship the software they make us change it in place all the time

their "spec" was some drawing from the engineering department

that may or may not make sense

i think it was the same though

like the dude in the drafting room can draw anything

true, then the machinists are the ones who really have to figure it out

yeah, seriously, like someone just says “make this reciprocal curve"

but the machinist looks at it and says "we can't make that at that cost"

and that’s the spec

yeah and they had to invent fiendishly clever tooling

"dividing plates" are interesting

seems like a legit way to work though, somebody needs to cut the problem down into smaller more attackable problems

like if you need to cut a gear with 40 teeth

you need a thing that can precisely divide a circle into 40 parts

and that you can connect to your milling machine to turn the gear when you cut it

and you use the machine to make that thing

yeah i think we're basically machinists

satisfying to think about

and the "engineer" is the business guy

@micha i think i just need to know more about how java works

@micha @alandipert helped me gloss over that bit when i first got setup

this is what heroku is doing

java $JAVA_OPTS -jar webapp-runner.jar ${WEBAPP_RUNNER_OPTS:-"--expand-war"} --port $PORT ./target_prod/project.war

@thedavidmeister: so you can make a uberjar and just run it like that, no?

instead of the webapp-runner.jar, whatever that is

we made an app like this for work, i'll share relevant parts

the build.boot. the build task makes an executable jar, the main class of which is sinmara.main-class https://gist.github.com/alandipert/2a8ea734cf0e8a013e949620791a01d4

main class is basically stub that exists just to be AOT'd. all it does is call sinmara.run/-main https://gist.github.com/alandipert/bcea48dc2eb0e3502d4b5566180ecb0e

i guess webapp-runner is something heroku does

¯\(ツ)/¯

sinmara.run is where we actually start the web app using run-jetty. this is where you'd start your app to run with http-kit - https://gist.github.com/alandipert/a2326dd6fe669a4ca4f00e316eb39a8f

oooh ok

so to build for deployment you do boot build which makes target/project.jar

so i have a server thing that is running my code for me

and to run in prod do java -jar target/project.jar

but i want my code to run a server thing inside it

i think so, yeah

basically you want your program to be the entrypoint

not the web server they provide

yeah, that would give me a lot more control

i think this is achievable with a Procfile or something but my heroku is rusty

that would have made zero sense to me when i was first getting setup, lol

hehe yes

welcome to the java world, all shenanigans all the time

i think owning the entrypoint is the way of the future, in this age of virtualization

i definitely feel like a lot of this would make more sense if i had a java background

web app containers etc existed kind of because they didn't have docker or VMs imo

yeah, where do you guys normally host your stuff?

i’m getting one of the guys at work into clojure

AWS

and he’s been messing around with docker

also war files were good for enterprise things where they needed to have like a multi tenant thing

different pieces of the organization deploying their application in the same server

so none of them could own it

oh yuck

yeah, that’s no my goal

yeah coconut VMs

straight AWS is pretty “raw"

i found docker great for local dev, but i'm :thumbsdown: for deploying to it

or otherwise relying on it in prod

docker compose is esp. awesome for local dev

why bad for production?

we have VMs already to virtualize unix environments

they don't come up as quickly as docker containers, but then most apps don't restart constantly

sticking with VM also removes a debugging layer that can get pretty hairy

yeah, at work we use vagrant, with a new environment for each branch

getting it up and down fast is pretty critical there

so we will probably move to docker at some point

if you use vagrant to build images then it starts to compete with docker for startup time probly

that's what we're moving toward at work, with chef anyway

we rebake our images each night automatically

using our legacy chef things to build images, that we then deploy from

cool, yeah

then send the images to the devs using btsync

if they have to build the environment from scratch each time it’s very slow, like 20m

oh, so this is for devs tho

so :thumbsup: docker then

but if they just pull up an image it’s like 4m

yeah, well we set all this up when docker was pretty new

seems like moving to docker for this would be pretty straightforward

yes, but it’s a matter of priorities, not whether we can/can’t do it 🙂

ahah yes i know that feel

atm we’re focussing on automating our workflows more

like, reporting, reminders, merge conflict resolution, release management, etc.

well, for my side projects, i like to have a service that isn’t sudo style setup on linux boxes

more time tinkering on the application layer, less time managing the environment 🙂

so @alandipert you wouldn’t want to use something like https://www.tutum.co/?

i'm highly skeptical of the "any cloud" things

i would for a side project or experiment

but for anything i need to maintain for months or years it needs to be no magic

since i'm going to end up building my own infrastructure regardless

haha, i think of it more like “outsourcing” rather than “magic” 😛

better to build around stable elements than layer magic

fundamentally you need to understand the platform

these tools that wrap AWS apis don't really solve the problem

especially when they attempt to wrap every possible api

then it's just madness

mmm

the time spent figuring out how to reverse engineer their stupid scripts is time you could spend just studying the underlying thing

that’s fine to say if you’re the one doing the understanding

but at some point, there will be people in the business who don’t understand

that's foolish

but if that's how it is then you can do it the other way

and just keep selecting a new framework every 6 months

it’s not foolish, it’s reality

you hire a new person

on day 1 they don’t understand everything

from a business standpoint, it's scary to build infrastructure on any platform that hasn't existed for very long, as the platform is subject to acquisition/shutdown/heavy change

this is another point for AWS

that’s true

they are excellent at managing change so far and aren't going anywher

if your people don't understand the platform and instead of studying that add layers of cruft on top of it, that's not good for business

but it's true that some business like that

it’s not just understanding though

what if you need to have certain SLAs that are 24/7 but you aren’t in a place to have a team on call 24/7

you might need to start offloading some risk

for example

then you're fucked

get insurance

i guess

yes :thumbsdown: to the sales guy who made that deal lol

haha, yes

"oh yeah bro we have people on call 24/7 no sweat"

but there’s always grey areas

hiring outside people to be the 24/7 guys isn't going to be cheaper i don't think

i mean i have acquaintances

who crapped out some ruby on rails thing, hosted on heroku for free, and sold the business for 100k a few months later

clearly they are winning

yeah i want that job

so it's definitely not that expedient magical platforms are always bad

haha

that sounds pretty great actually

well i want to hire them

lol

aha you can't

magical people such as these can't be hired

as long as you can reproduce that magic

lol right

i don't really want to hire them lol

realistically, you want to “cofound” with them i think 😉

eh i'm not that kind of gambler

like i used to play poker a bit

there just isn't an easy way to make these decisions i don'tthink

depends on the team, the business, etc

it really does

then i realized that the time i spent studying that could be spent doing something "productive"

but maybe that's dumb

i just know personally that if i'm going to be involved with something that i want it to succeed, and to do that i will need full access

i don't want to log in to some web console to right click on something to make go

like take chef for instance

we have tons of problems with it

yeah, but there is a cost to the full access bit

the cost is you can't hire as many idiots

and it takes just as much knowledge to fix that

oh man, i’ve never seen anyone have a good time with chef...

so hiring gets harder

and under pressure too

the cost is also your time as well

like in the middle of a deploy things go wrong

time is hard to manage

it is, but i'd call that cost an investment

vs time learning magical things which is a gamble

sure, it is an investment

it's not even an investment

and every investment has an opportunity cost vs. other investments 😛

it's way cheaper

like if you don't understand AWS, chef is not going to help you

true, since you can usually start applying instantly

when it fails you will still need to know what the underlying thing is doing

if you are responsible for the thing

also, it’s not really fair to say that everyone who doesn’t know AWS inside out is an idiot

so now you need to know two things

you won’t get a very diverse team with that approach 😞

you only need to learn though

it's not that you need to know it already

my point is that there is no alternative

like adding tools on top of AWS doesn't solve the problem of not knowing how anything the tools are using works

that’s true

those tools will still go wrong

and then you have two problwms instead of one

but we’re building our own tools to help automate tasks

also i didn't mean to imply that about people who don't know AWS

it’s not that much different in one sense

i think ultimately, i’d prefer to just have something reasonably portable, and frequently re-assess the situation

AWS has only a small number of things that you need to know really

and move if it makes sense to

a handful of things

in order to do pretty much anything

it's like a week of study at most

maybe 2 weeks, it's hard to remember now lol

hah, i think probably 2 weeks to get setup

but more if you want to get into security and change management

another example from work

like 2 yrs ago i needed to stand up a new app

so i went with easy & fast, an AWS service called elastic beanstalk

which AWS built on top of many of their own services using dozens of shell scripts and some web console

the stuff they added on top made it easy and fast to get running

but debugging, monitoring, scaling all were very difficult as a result

so over time the cost of owning the app increased

since every problem was multiplied by the problems inherent in debugging something magical

all the secret knowledge compiled in little .ebextensions files

from stack overflow

yes, tons of DSLs and lifecycles

marks of the beaste

yeah that’s true

hacking and reverse engineering their ruby scripts

so i guess you hit a point where it made sense to move

which is OK too

yeah totally

it was organic

i like having a free heroku account that magically got me setup without knowing a thing about java

and now i’m looking to take it a little further

but the key thing i think is that we didn't move to some other magical thing

i bet i’ll get to a point where it makes sense for me to move again

instead we went a level deeper

yeah, that is fair

like my rails buddies

super into docker now

it's the new magical thing

but i won’t look back and be like “heroku sucked”, i’ll just be like “i outgrew that setup"

right, totally

altho

if you're at all interested in AWS you may be interested where we landed

i am

micha identified like the 3 things we need to use from AWS to make awesome web and worker apps

but hold that thought

i have an appointment to get to

i’ll bbl 🙂

will do. gonna sleep soon but it was fun to chat!

https://github.com/kennytilton/its-alive btw

cells from CL, in clj/s

hoplon shoutout in readme. has weird/cool ideas, like "awakening"

whoa

i was actually just about to ask if javelin could be used in clj

maybe i should look at this too

the ephemerality looks interesting

i was vaguely wondering how to get a subscription of changes coming from rethinkdb into the cell model, without explicitly resetting things to nil after every change

if you have to set to nil it's not "changes" though right?

that's a good thing

if you want to know that you talked to rethink db just have it include a timestamp or transaction id

timestamp vs ephemeral i think explicit timestamp or sequence number is superior

@micha yeah i’m working on a timestamp + sequence number for inserts and queries atm

something like

(-> (rq/db r/rethinkdb-db)
                    (rq/table r/rethinkdb-table)

                    (rq/order-by {:index (rq/desc "stamp-tx")})
                    (rq/group ["e" "a"])
                    (rq/limit 1)

                    (rq/run conn)))

i don’t yet know how to get changefeeds working

about to ask in the other chat room 🙂

ah, and now i can learn how channels work

https://github.com/apa512/clj-rethinkdb/blob/master/test/rethinkdb/core_test.clj#L228

the joys of being a noob >.<

hahaha

the fun part