Anyone know what the comprehensive list of directories are for OSX where boot caches files? I'm often having problems where boot is trying to find source files that no longer exist. This seems to have come after moving to 2.5+ (on 2.5.5)

@sbrady: ~/.boot/cache

that's it

Yeah, I rm -rf 'd that

Still having this problem

have you downloaded a recent version of the boot.sh binary

also if you used homebrew are you sure your homebrew is up to date?

pretty sure I'm on 2.5.5

After clearing the cache, it re-downloads 2.5.5

(not on homebrew btw)

sounds like it's working, no?

No, after clearing it, I still get java io can't locate foo on the classpath

I know it must be my machine because the CI server isn't looking for this errant classfile

try boot -BP

does that work?

yes

so it must be something related to your build.boot or your boot.properties

that could be

those options disable loading boot.properties and build.boot

but the ci server is using the same build.boot

could be props

try boot -B

if that works then it's your build.boot

if boot -P works then it's your boot.properties

if neither one works then it's both

Hmm, so boot -B seems to be giving a problem

So what does that mean to you?

the -B option disables loading of build.boot

so since boot -BP was ok, and the -P option disables boot.properties

that means your boot.proeperties is the problem

OK, so that would be consistent with the CI server saying OK. Got the problem fixed. Thanks!

🍻

can boot or some custom task can bump release given +version+ in your build.boot?

found! https://github.com/degree9/boot-semver

hey all, is there a method of stripping out an input directory from being seen in subsequent tasks' input-dirs? I see rm, but this appears to just be for individual files.

my use case is that I have a task that transforms some source files -- the subsequent task should see the transformed files, but not the originals

rm or sift are the things you want i think

you can remove the files in one swoop w/ sift + regex

or if you need more control, do your own rm-ing

ah okay, I'll look at sift, thanks for the pointer

sure

it's also worth knowing that the fileset works like git or S3

there aren't really directories, only paths to objects

(the classpath also works like this)

so, to remove a "directory" is going to be removing paths that match some pattern

ah, perhaps I'm going about my problem the wrong way then. Trying to solve the problem of using find-namespaces from clojure.tools.namespace.find, which expects a root directory of source files. Previously was just trying to feed it one of the temporary transactional directories that boot makes. But this might not make sense to do it this way.

i would do that by finding all the clj files in the fileset and copying them to a tmp-dir of your own making, then unleash find on the tmp-dir

Hey all, I’m fairly new to clojure & clojurescript, even newer to boot. Coming from JavaScript, I really really like everything boot has to offer. The only thing that I think it is missing are some front end compilation tools. There are plenty of tools written in js (ie postcss) and I would love to port them over to boot, but I don’t really know the best way to start. Can you run clojurescript in a boot task? Can you run javascript libraries in clojure? Or is the best way to interface between them the clojure exec function?

^ apologies for any outlandish beginner assumptions made

@bendy: no worries that's a good question 😄 and welcome to

@alandipert: I like that idea, I'll probably go with that -- thanks for the suggestion!

@shwx84: surely, enjoy

@bendy: you can run cljs in clojure via nashorn or another JS for the JVM like Rhino

what's missing is the nodejs runtime. node adds lots of things to the JS environment that the JS-in-JVM things i'm aware of don't provide

i found a few abandoned attempts but nothing that probably runs today 😦

so yeah, i would recommend exec

or dosh which comes with boot

(vendored from the conch library)

the recipe for a task that ran node on some files from the fileset would be to make a tmp-dir, like for @shwx84's problem

and then aim node at the tmp dir, and add whatever results you want from that tmp-dir to the fileset and then pass it on

@alandipert: awesome, thanks! That’s definitely enough to go off of for now - I’ll report back if I make any progress 😉

@bendy: cool no problem! probably the fastest way to get cookin' is by looking at 'community tasks' on the wiki and finding the one closest to what you need and hackin' it

any of them that drive non-jvm translators/compilers are probably good references

$ rm project.clj profiles.clj; git commit --all <- satisfactory

Revisited build.boot for importing backend as well

New usage for building: boot -e flavor=fronted|backend build -t prod|dev
New usage for devving: boot -e flavor=fronted|backend dev

any good practice for putting revision in pom.properties using boot?

Is there any way to set default values for task options?

I mean in the options spec rather than by writing a fn to do it.

@kul I think it's usually done through task options

https://github.com/adzerk-oss/boot-cljs/blob/master/build.boot#L14-L22

@juhoteperi do you know of any issues running on linux with your sass task?

getting this error: java.lang.UnsupportedOperationException: Platform linux:i386 not supported

I didn't have problems running it on mine

I'm on arch

@raywillig: it only works on 64bit

If your system is 64bit could be that you have 32bit Java. If your system is still 32bit I would highly recommend upgrade.

There is a issue about this at jsass repo: https://github.com/bit3/jsass/issues/11

Question about copying dependencies. I need to copy my jar dependencies to build/WEB-INF/lib. I have a task that does this using io/copy. That doesn't seem very bootish, but I haven't been able to figure out how to do it with a fileset. If I had the jars in a fileset I think I could use the target task, which seems more idiomatic. But I don't see how to get the jars into a fileset. Any advice would be appreciated.

are you trying to build a jar?

No. This is for google app engine; for security purpose, the dev server will only look in its own dirs, so it cannot use the jars from ~/.m2/ - they must be copied to WEB-INF/lib, without unpacking.

pod/jars-in-dep-order gives me the jars, then io/copy works, but that doesn't seem like the Boot Way.

We have a task to do almost exactly that… let me refheap it for you @mobileink

Something like this https://www.refheap.com/114762

We have multiple web apps under a common root folder ws-root and this gets out list of (web-libs) dependencies, and copies them to all our WEB-INF/lib folders.

I expect that’s pretty much exactly what you came up with? https://clojurians.slack.com/archives/boot/p1455378406006079

@kul I use boot-semver and set (def +version+ (semver/get-version))

@seancorfield: Looks like the same idea. But my understanding is that in Boot we should not directly muck about in the target tree, and instead use filesets. Hmm, now that I'm staring at the code some more it occurs to me that the way to go is to io/copy stuff to a core/tmp-dir! and then add that to the fileset for later handling by the target task. I've been using a hardcoded output path, io/copying straight into the target tree.

So I gather its a bit of both. You write to the filesystem, but always relative to a tmp dir controlled by boot, so you're kinda sorta writing to an unreal FS. Does that sound more or less right?

@mobileink: also if you write to your own temp dir and then add to fileset you have excellent control over caching strategies in your task

you can use boot.core/fileset-diff to obtain a fileset object containing only changed files

and then decide which things in your temp dirs need to be rebuilt

One thing to be careful of with target is that its default mode is to erase everything in the target folder so using target with your WEB-INF/lib will require an option to not do that

You’ll also have to move those classpath JARs to your resource paths to get them into target.

@mobileink: re: the WEB-INF/lib dir, did you see the uber --as-jars option, with the war task?

I had this discussion with @micha at the time and his recommendation was to do the io/copy pretty much as I did

those two in combination will result in a war file with all dependency jars in WEB-INF/lib

(But, yeah, if you’re prepackaging the entire app as a WAR to deploy to GAE, I’d do what @micha is suggesting here)

That wasn’t/isn’t the case for us.

boot something something-else uber --as-jars war

plus maybe the web task to build a web.xml unless you want to make that your own way

I did see the uber task but now I see I misread it

the war task will find all jars in the fileset and put them in WEB-INF/lib

so it works with the uber task

@seancorfield: regarding target cleaning the target: learned that the hard way a few days ago. 😉

@danielsz: im trying to decipher your latest advancements on boot.system. im a bit lost, because 1. i can't see git tags corresponding to the released versions i see on clojars 2. u have reverted to using ns-tracker from clojure.tools.namespace, though the docs still talks about c.t.n.repl (https://github.com/danielsz/system/commit/98457f6f1c1af1944dd0f89b5a8c21e9ecc7a83d) what will do the changed files detection eventually? boot.task.built-in/watch or c.t.n.repl?

looks like the uber task is just the thing. regarding WARs for GAE, the dev server doesn't like warfiles - the gradle plugin my boot-gae will replace goes to the trouble of making a warfile and then exploding it. so far it looks like the warfile can be skipped. then when up upload to the prod server the google stuff just knows what to do, no need to make an uberjar

yeah the uber and jar processes are separaet in boot

just tried boot uber --as-jars sift -i "jar$" target -d "build/WEB-INF/lib" and it worked like a charm. that is awesome. boot-gae should be ready to play with in a day or two but i can promise you right now that GAE with boot is about 10000 times easier than it is with gradle/maven. i hate to even think about how much time i spent trying to learn gradle/groovy and then the plugins - a nightmare.

i'm actually having fun figuring out how to do things with boot. fun doing config/build stuff - that's something i never ever thought i would say!

hahaha that's awesome!

@mobileink: shall we have a Google App Engine page on the boot wiki for such recipes?

@onetom: i think that might be a good idea, we can also post a blurb about deploying with docker and such

Pretty good article on the intricacies of package managers: https://medium.com/@sdboyer/so-you-want-to-write-a-package-manager-4ae9c17d9527#.i497ikadx

how do i run a repl server, tried: 'boot repl -s', but it just returns...leaving no running process.

@fenton: by default most tasks don't block the pipeline, use boot repl -s wait to have a blocking task at the end

@martinklepsch: thanks martin i'll try that...

@martinklepsch: yay!

@flyboarder: actually if @mobileink makes a boot-gae task, then that can have the documentation and we should just link it from https://github.com/boot-clj/boot/wiki/Community-Tasks

(i tried to create a page but then i realized im not sure what would i put into it, neither did it fit anywhere on the current TOC)

@onetom: I think the community wiki should have a deployments section somewhere tho, the community projects page is almost big enough to need categories

make that the tasks page

what would be under it? • Deployments • Google App Engine • Heroku • Tomcat? • Docker? • VPS? eg Amazon AWS / Digital Ocean / RackSpace (just to have the keywords what ppl might search for)

Yeah something like that, if there is a community project for one of them then direct link to that?

I was thinking of a heroku task myself and there is already beanstalk

it's actually more like u have certain ways to package and daemonize a clojure app in a 12factor compliant way: • Docker • WAR • Uber JAR • jsvc • Apache something something • Upstart • Systemd

FYI, boot-new now has a first cut of generators in the 0.4.0-SNAPSHOT https://github.com/seancorfield/boot-new#boot-generators

and for the various actual PaaS and IaaS provides u can link step by step instructions and document the peculiarities of them

So we need 2 sections then, packaging and deploying

packaging having the nitty gritty details and deployment having community projects and or instructions if there isnt one

something like that, but im not very experienced in this regard yet. we are still running boot directly via upstart in production on EC2 using Amazon Linux.

@onetom: i think you would benefit from the EC2 Container Service then, building the container image is very easy thanks to https://github.com/adzerk-oss/boot-clj-docker-image

how to specify multiple options on the command line: $ boot repl --server --init-ns 'test.core' --eval '(start)' -- wait seems to only run the --server part and ignore the rest?

why do you say it's ignoring the rest?

if you do boot repl -c are you in the test.core namespace?

let me check

yes, correct namespace, but doesn't want to do my --eval part.

though if i drop the server....the eval part works

oh wait

right yeah

that's an option for the repl client

not the server

because it needs a session to do it

there is a ticket about that

is there a way launch custom code from the server?

we need to make a separate option for that

there currently is not

well you can put the code in your init-ns

at the top level

that would do it

ouch

i prefer to have the option of running the code...if it throws errors... hard to get any repl started if errors are thrown on simple file load... 😐

use case was basically to have a repl i could attach to on my production code...but code needs to auto-restart on server reboot, etc...

@flyboarder: i was dismissing docker until recently, because of the big upfront cost of getting the base images. BUT they started to transition to a 5MB Alpine Linux base, so I started to learn it finally. https://insights.ubuntu.com/2016/02/10/docker-alpine-ubuntu-and-you/ and https://www.brianchristner.io/docker-is-moving-to-alpine-linux/

but would be nice to be able to come in, in multiple ways...my custom task, a regular repl no over aggressive NS loading...

yeah

there is a ticket about the repl option

it will be in 2.6.0 hopefully

@onetom: yeah I would consider it for sure, if you have issues with the images let me know

@micha: okay cool.

if you want to give it a shot PRs are welcome

we're a tiny bit backed up at the moment, but things will improve soon

@micha: okay, maybe, I still don't feel that advanced...but hey why not!

@onetom: also interesting if you are running your own PAAS for your apps there is rancheros 20mb linux running docker on the kernel

fenton: i was battling with similar issues a few weeks ago. what makes it even more complicated is that we are also connecting to the repl from intellij we ended up having this:

(def repl-init
  '(require
     '[clojure.inspector :refer :all]
     '[adi.core :as adi]
     '[datomic.api :as d]
     '[org.httpkit.client :as http]
...
     '[midje.repl :refer :all]))

(task-options!
     repl {:eval repl-init})

(deftask dev []
  (alter-var-root #'midje.sweet/include-midje-checks (constantly false))
  (eval repl-init)

  (comp
    (environ :env {:http-port 3001})
    (watch)
    (system.boot/system
      :sys #'sys/dev
      :hot-reload true
      :auto-start true)

    ; (eval repl-init) above is necessary to require the convenience namespaces
    ; so we can have them when connecting from IntelliJ via lein, which can not
    ; simply use the `repl-init` defined in this build.boot
    (repl :server true)))

@onetom: okay i'll muck around a bit and see what i can figure out...

@flyboarder: thanks! (trying the adzerk image now. it will take probably 30mins on my connection to download it though...)

ugh

officially 4Mbit/s connection...

you and micha both have terrible interwebs 😛

which propels us towards being efficiency conscious

i have to kick my gf off the internet when i work from home and need to skype

not good

i need to stop my 1.5yr daughter watching "wheels on the bus go 'round and 'round"

I made a scratch space for packaging and deployment https://github.com/boot-clj/boot/wiki/Packaging-and-Deployment

which is actually good, just loud

@flyboarder: fyi, it took 25mins https://gist.github.com/onetom/6ee78d00dfdfef7602cf

4Mb/s? Yikes, I couldn't function. I get about 100Mb/s down 12Mb/s up and even that doesn't feel very fast to me these days.

haha

when my best carrier pigeon ran off with the squirrel i lost half my bandwidth

5yrs ago i was every other month in Thailand, where i only had 11kbyte/s EDGE mobile connection and still i was administering 40 servers and monitoring them via Newrelic took a little patience but it was quite doable

now i have packets hidden in acorn caches all over the place, lost for ever

@micha: Note added, hopefully it's clear enough: https://github.com/adzerk-oss/boot-cljs/wiki/Usage#compilation-levels

@richiardiandrea: looks great

but?

you may want to mention that the main.out directory is still needed if you want source maps to work

is it?

yeah, that's why it's left there

by default

let me check

source maps refer to cljs files in that directory

oh, I hoped it was a 1:1 mapping from main.js

we had to do a lot of work to get it to work transparently

becaue the source map settings are tightly coupled to a bunch of other settings

and you need to do things to make sure it generates the right paths to encode in the compiled js

well the source maps need to reflect your input files

because they're going to show you where in your code the error happened

so you need to have your original files available for that to work

my JS skills are weak my friend

i think main.out may already be removed if you have source maps disabled

i don't remember though

I will do a couple of tests here

with my brand new command: boot -e flavor=frontend build --type prod 😄

optimo

ahahah (ottimo is Italian!)

what does it mean?

it means perfect

haha sweet!

ottimo

that sounds like a good name for a child

ahahahhahahahaha

ottimo niskin

for a girl perhaps

no for a girl is ottim*a*

ah right

but it would be maybe too cocky to call your son/daughter like that 😄

true

bad luck

in Italian, we also say: ottimo, Andrea

that would become: ottimo Ottimo

a bit weird 😄

this is a great word, TIL

I remember you did not need it ^, I am trying in firefox @micha

you will see the folders because it sees the references in the js file

but there won't be any sources there

it will just be blank

you can test it by throwing an exception or logging to the console

oh you're right

in chrome anyway you will get a stack trace that links to the source

TIL 😄

getting source maps to Just Work was a huge pain

will change the Readme

May have an issue relating to web workers.

specifically the cljs.edn file.

I have a library which creates/runs a web worker. The app should not need to know about it. But if I put the cljs.edn file in the library jar by using a resource path, it gets ignored.

I can only get it to work by adding the cljs.edn file to the app resource path.

From a web-worker perspective, this makes no sense. Or am I missing something?

Perhaps I should be creating the js file somehow in the library jar?

Currently the only thing I do in the build.boot of the library is create the jar with all the source files.

@micha ?

is the cljs.edn file actually in the jar?

jar tf project-1.2.3.jar |grep cljs.edn

oh, I didn't check. back in 5.

(it isn't right now, I need to put it back.)

Yes. It is at the top level of the jar file.

ok yeah i see the issue now

the .cljs.edn file is an application spec

i don't think it's appropriate for applications to be built via transitive dependencies

like if you have some distant transitive dependency that is starting web workers

you need some way for the project to opt in

like a boot task perhaps

that can pull the things into the project

from the jars

I can live with that.

it also avoids the need to scan the classpath

which can have hundreds of thousands of things in it

and can be quite slow to do

so this way the app only needs to know which jar files might be creating web workers.

i imagine you could do it like the cljs-reload task works

you'd have the web-worker task that pulls the things into the project

And to take it a step further, a library which depends on another library might want to do the curtesy of "pulling in" from its dependencies, so the app only needs to know about the top-level dependencies.

and performs the integration

yeah that's the benefit of using the task approach

the application doesn't need to know much

you just put the right tasks in the pipeline

like with boot-reload

the pplication knows nothing about reloading

So would it be fair to submit a request for such a web worker?

that's added in the build pipeline

er, task.

sure go for it

Thanks.

https://github.com/adzerk-oss/boot-cljs yes?

you could just make a tssk

*task

yup

it doesn't need to be part of boot-cljs

like boot-reload is a separate thing

and so on

which project then? I know it can be independent.

yeah i would probably make it a separate task, no?

in?

how do you mean?

like in its own jar on clojars

you are saying a separate repository should be created?

probably yeah

then there is no place to add an issue.

what is the issue for?

web worker task

hm, i don't follow

you are going to make a web worker task, right?

perhaps. I'd need to dig into boot of course.

and I do have the time.

could be a good project for me.

yeah just try one, see if you like it

😄

after that we charge

👍

I'll get this api change to durable-cells done first. Then I have a need for it.

@onetom: i experimented with alpine recently but discovered it doesn't use or come with glibc, so a binary i needed to run in there didn't work

@onetom: just a PSA really - i think if you stick with what they have packaged or if you compile native stuff in the container, you're fine

OK, https://github.com/seancorfield/boot-new 0.4.0 with generator support is up on Clojars. Built-in generators are ns and defn. PRs welcomed to add more. Generators can also be part of your own project (although you’ll need to depend on Boot new!).

Oh man cool stuff!

now we need cider integration 😄

for defn I use yasnipped but ns is very useful indeed

ns and defn were just the first two that popped into my head, and for defn I needed to update boot.new.templates/->files to support appending to an existing file so it was all good.

I’ll probably switch my FW/1 web framework template from Leiningen to Boot now and add generators for controllers and views etc.

Great yes, it would be nice for instance to generate re-frame handlers/subscriptions boiler plate