All our new Boot-based stuff is in production, along with Clojure 1.8.0 (yay!). No more Leiningen in production!

!!!

A lot of our little shell script cron jobs have become /path/to/boot some-task as well.

Well, technically cd /path/to/project; ./build/bin/boot some-task 😸

awesome!

i'm really happy it's working for you on windows, too

Yeah, that mostly helps when I’m developing FOSS projects on my little laptop while watching TV

We only have Macs for desktop (DEV) and Linux for servers at work.

We’re beginning to Dockerize a lot of our stuff so moving stuff out of Ant and into Boot tasks simplifies that process (since we don’t need anything but a JVM and our project in the image — with our Boot wrapper and the boot script itself).

Glad to hear it, @seancorfield , and glad to find another windows user.

--I'm a bit bummed over today, but I'm thinking if I can't unit test, I can always write demos.

Ah, did I miss what has you bummed out?

headless cljs testing.

i.e. unit testing.

https://github.com/crisptrutski/boot-cljs-test/issues/42

Ah, a pain we haven’t felt yet. We abandoned our proof of concept cljs app early last year.

@laforge49: do you have access to something like virtualbox?

I'm very excited over cljc. I want to move my database from clj to cljc

what is virtualbox?

you can run a linux virtual machine in it

it's free, at least for linux it is

@micha my plan now is to focus on system tests via demos. Demos are important anyway. Also, I avoid linux except when there is someone around willing to give me a lot of help. I'm really pathetic at it despite having worked 2 years on solaris and another year on a unix box.

--I barely survive on windows.

I focus too much on my own code

Looking at your counters demo. I'll copy it and add some logging. Just what I need. Starting by creating a demos repository under aatree, of course.

how about docker

you can run the boot docker image

so you'd compute in windows, edit files etc

I'm already reaching my level of incompetence learning clojure and now clojurescript.

but boot would run in the docker container

I really love the idea of docker! But this isn't the time. 😞

Right now I want to create enough value that folk will be willing to put up with my boneheadedness!

haha i hear ya

I'm very open about my failings, but I like to show off a bit too.

I also feel that catering to boneheads will open things up for a wider community. So I don't mind asking the dumb questions and then writing up my understanding.

But my own self image is not that of a bonehead!

Enough!

Love your counters demo.

--forking it. 😄

Does sift :add-jar accept a symbol and a regex only? not a set of regex like all the other utilities?

it adds the contents of a jar from maven

identified by the symbol etc

yes it is super cool, just wondering whether it accepts a regex only or multiple...

it looks like one only

yes, i think that's right

would you be interested in a patch for multiple regex there?

or if there is another way to do it

sure, will it work on the command line?

well, I guess is the same as --include for uber

it accepts a set

currently it's accepting a vector i believe

(boot (built-in/sift :add-jar {jar #"(?i).*\.(clj|cljs|cljc|js)$"}) (built-in/show "-f"))

this works, where jar is a symbol

i mean from the command line though

boot sift -j jar:'(?i).*\.(clj|cljs|cljc)$'

you can currently do that

equivalent to what you did in the repl

yes one yes, but no more

can't you make one regex that does what you need?

well ...... yes

😄

i don't think there is a way to do that cleanly in the comand line

because you can use multiple -j arguments now

yes true, because it is a compound option

right

the cli stuff is a delicate balance between simplicity and features

one thing is, sift has --include already -> https://github.com/boot-clj/boot/blob/2.5.5/boot/core/src/boot/task/built_in.clj#L395

so if add-jar accepted just the jar (symbol), and --include is specified, it could process from there

hmm

the current behavior is to just combine the two

yeah i think the sift task interface could be improved

what I mean is {jar regex} <- looks redundant and --include could be used

yeah i see what you mean

like have an "action", like add-jar in this case

and then modifiers like --include etc

yeah

we could do that in a non-breaking change even

like uber

if we add like

boot sift --action add-jar --include foo.\*

adding the --action option lets us keep the existing behavior when that isn't provided

yes but is uglier 😄 then why not a task, but again it is a matter of taste I guess

just wanted to raise the issue, if you want I can open one and discuss there

well for the built-in tasks it's important to utilize the names efficiently

true

like every task name used by boot is one the user can't use

and on the other end, if you plan to add more actions this is a good starting point

and it make sense because sift is more like a util function

for example now I don't think you can combine add-jar and add-meta which is a bummer, it would be so handy (for me at least :))

What are some "go to" links to help me map from leiningen to boot? I've seen https://github.com/boot-clj/boot/wiki/Boot-for-Leiningen-Users

One starter question: I want to add tools.namespace as a dev dependency for REPL reloading

@xpe: you can load tools.namespace as a dev dependency like this:

(set-env! :dependencies '[[org.clojure/tools.namespace "1.2.3" :scope "test"] ...])

@micha thanks! :scope is what I needed

🍻

and am I correct in surmising that :scope "test" is based on how Maven handles dependencies? (e.g. https://maven.apache.org/guides/introduction/introduction-to-dependency-mechanism.html)

(same as lein)

exactly right, yes

haha your avatar is hilarious

cool. without starting a huge thread about the horror of transitive dependencies, should I expect boot to inherit the usual Maven transitive dependency pains?

oh, thanks, I was surprised that the power supply is so cute

boot uses the same dependency mechanism as leiningen, which is the apache aether one wrapped in the clojure pomegranate library

maven uses this also

and, thanks to some Hacker News reverse engineering, that power supply is both cute and packed with wise innards

wise innards?

oh, all the electronics inside are surprisingly involved

i've heard that the apple power supplies are pretty advanced

https://news.ycombinator.com/item?id=10628212

ah right, that's where i read about it

btw for dependency stuff, check out boot show -d and boot show -p

help debugging deps issues

yes, I saw those get mentioned

I want to do a deep dive on transitive dependency hell

I know about exclusions, but I want to know the painful history of workarounds

pods help you avoid it, pretty much

yeah maven is probably the best thing around, because it's been through the war

across tasks, I can see pods helping a lot

well pods are also a way to workaround dependency issues

within a particular clojure app, there is always the risk of conflicting transitive versions

like if you run into something, you just fire up a pod and move on

could I use pods WITHIN a single clojure app?

sure, that's what boot is

you need to launch the application with boot though, for now anyway

ok, what I mean one JVM that, say has a library needing ANTLR X and another needs ANTLR Y

I know about https://code.google.com/p/jarjar/

yeah that's possible with pods, no problem

i mean it's possible with the jvm if you arrange your classloaders to accomodate it

that's what boot does, before clojure is loaded

all dependencies are isolated in pods in boot

very cool! any example code bases you'd recommend that show an example like I mentioned?

you can do it very easily yourself:

Re: "maven is probably the best thing around, because it's been through the war" PUN INTENDED!

or maybe the WAR has been through Maven?

(require '[boot.pod :as pod])
(def p1 (pod/make-pod (update-in (get-env) [:dependencies] conj '[antlr "1.0"])))
(def p2 (pod/make-pod (update-in (get-env) [:dependencies] conj '[antlr "2.0"])))

;; then use something antlr
(pod/with-eval-in p1
  (import antlr.whatever)
  (.foo antlrthing))

(pod/with-eval-in p2
  ...

@micha that's very helpful

the reason why i did the update-in there is because i may have maven configuration in the (get-env)

like repositories, etc

This helps. I'll still have to think a bit more about how this would play out in existing projects.

we will hopefully have a thing that lets you create a standalone application that uses pods

but as it is right now you would need to have boot be the entry point for the application that uses pods

because it's the java part of boot that sets everything up for pods to be able to exist

Ah. That relates to my question, which is about with-eval-in above... does each sub-expression run in some kind of separate JVM? Or at least with a separate classloader? I don't know the details of pods.

separate thread, separate classloader

And can each sub-expression communicate with the other one somehow? I'm not sure if they are different threads or what...

also no clojure objects can pass between pods sensibly

(interlaced commentary, you answered my question before I pressed enter)

because clojure creates protocols, classes, and interfaces dynamically

so a function in one clojure runtime doesn't even implement IFn in another clojure runtime

so the with-eval-in business is serializing expressions and results at the interface between the pods

java objects can pass between pods as long as the classloader the objects are from is visible to both pods

but with-eval-in doesn't expose that, it prints the expression to a java string, which is then passed to the pod

ok

thanks for the details!

np!

anyone here using inf-clojure with boot?

I found an issue with using inf-clojure with boot

will update the boot wiki

for those who are interested in trying it out: https://github.com/boot-clj/boot/wiki/inf-clojure-REPL

there any books that explain boot/java maven/pom/environment stuff for people just getting into this?

i'm finding i can use lein templates and hack on a project but i have no idea what's going on underneath, and i like boot a lot more cause i can get CIDER to work with it and seems way faster to start

the wiki is a great place to start

you'll just need to understand the boot env, filesets and pods

these 3 things and i should be able to configure any webapp i want?

i feel there's a lot of underlining java info i need to know

for web apps perhaps the most important part is knowing how filesets work

the distinction between source-paths and resource-paths

I can't imagine how java info would come into play

most deployments use boot as an entry point

cool will look into those, also this seems immensely useful and what i was looking for: https://lionfacelemonface.wordpress.com/2015/01/17/boot-getting-started-with-clojure-in-10-minutes/

many people like this as a resource: https://github.com/magomimmo/modern-cljs/tree/master/doc/second-edition

perfect 😉

thanks for this

feel free to ask questions around here

micha is super responsive

also welcome to boot!

hehe - i've been using it a lil bit with a tenzing generated app, but i find that as i want to do more with it i kind of need to back up and start grokking it from the ground up

application templates are immensely useful to get started and i would almost not use boot over lein templates, but certain things seem difficult with lein as well and boot-cljs seems superior to figwheel for me

modern-cljs would be perfect then

I too prefer to build by build.boot file myself

sean is working on boot new functionality

something along the lines of lein new app would be perfect for me

honestly that would be amazing

having boot templates

would be faster for newcomers to get going

i liked tenzing but found out the hard way that every dependency it downloaded was outdated

but was still using that as my boot project after updated everything

but it was only way i knew how to get going

https://github.com/seancorfield/boot-new

boot show -u shows outdated deps

Oh, nice; I knew of https://github.com/martinklepsch/boot-deps but didn't know something like that was already included in built-ins.

It detects the latest stable release

Doesn't show snapshot versions

@jethroksy: you can use boot show -U if you want to see snapshots

ah thanks for that!

hoplon demos need a boot -u 😄

The only problem is that alpha11 release keeps showing up.

Infinite error loop when using watch.

I'm using only a slightly modified counters demo: https://github.com/aatree/aademos/tree/master/counters

Aside from updated repositories, I've also added a cljc file.

Everything is fine until watch detects a change. Then it goes nuts:

java.io.IOException: Couldn't delete C:\Users\Bill\.boot\cache\tmp\Users\Bill\Documents\aatree\aademos\counters\fgc\mm3x96\index.html.out\goog\labs\useragent\browser.js http://clojure.java.io/delete-file/invokeStatic io.clj: 434 http://clojure.java.io/delete-file io.clj: 430 ... boot.file/delete-file file.clj: 54 boot.tmpdir.TmpFileSet/commit! tmpdir.clj: 236 boot.core/commit! core.clj: 383 boot.task.built-in/fn/fn/fn/fn/fn/fn built_in.clj: 264 boot.task.built-in/fn/fn/fn/fn/fn built_in.clj: 264 boot.task.built-in/fn/fn/fn/fn built_in.clj: 261 boot.core/run-tasks core.clj: 794 boot.core/boot/fn core.clj: 804 clojure.core/binding-conveyor-fn/fn core.clj: 1938 ... java.io.IOException: Couldn't delete C:\Users\Bill\.boot\cache\tmp\Users\Bill\Documents\aatree\aademos\counters\fgc\mm3x96\index.html.out\goog\iter\iter.js http://clojure.java.io/delete-file/invokeStatic io.clj: 434 etc

etc

etc

And boot.properties contains BOOT_EMIT_TARGET=no

Mayhaps it is a windows issue? 😞

what does the unmodified demo do?

the demo you wrote has 10 counters that you can inc/dec

I just added a log statement and a require for a cljc file.

Oh, and when I drop the cljc file, watch works fine.

So it may be having the cljc file on windows, or perhaps just having the cljc file.

i'm going to fix that "can't delete" business once and for all

but today i am in the weeds with some AWS stuff

no problem

I'm moving the cljc file out of the demo. But first I will create a branch for you that creates the error.

i think one issue is that boot sets the temp files in the fileset to be read only

I'll also create a branch of aautil for the other windows error I reported.

--which windows interprets as undeletable, hmm?

the idea was that it should throw an exception if the user writes to those files, because if you write to them you end up with a corrupted fileset

but i think maybe we just document that, i think it's pretty obvious to most people

so you need to turn off read-only before deleting!

yeah or just not set them to read only in the first place

i think read only isn't really respected in general

updating those temp files is moderately obscure thing to do. So unlikely that you need to worry too much about novices modifying them.

by the system

output files by the target task are also read only

but does modifying them corrupt the fileset?

I think it just syncs the fileset into a given folder and the fileset remains in the temporary directories as it were, so it shouldn't.

modifying them while boot is running could cause problems

because the target task will by default create hard links to temp files in the fileset

this can make things a lot faster obviously, but if you write to a target file while boot is running you will corrupt the fileset

however when boot exits those hard links are now linked to unreachable temp files

so you can modify them as you like

Okay!

if you are going to be modifying those files while boot is running you can use the --no-link option to the target task

IMO read only is still a sensible choice

Ah, so it not only hard links between the temp directories, but with target as well. Good to know, for some reason I assumed the results are not hard linked.

it only links some of the things

the idea is that it can safely link to any of the tempfiles that will become unreachable when boot exists

the scope, or lifetime of a fileset is one invocation of boot

because when boot start up it creates a new temporary storage area

so when boot exits, the files in there are essentially unreachable

so making a hard link to files in there should be okay

because only boot knows where the referent files are, and it's not running anymore

the next time boot runs it will delete all those temp directories from the last run

leaving your hard links as the last remaining version of those files

Right, I just kind of assumed it was a performance optimisation for passing files between the tasks in the fileset and that the results are not optimised in such way, but that was just it - an assumption.

Certainly good to know how it works for real.

there are a lot of optimizations that are possible if you assume that something isn't going to be modified by the user

@micha I've created this branch for you to preserve the error I was getting on windows: https://github.com/aatree/aademos/tree/watch-delete-issue

I have a git repo in a subdirectory of /resources, and it's getting included in the (watch) task's files. What's the best way to ignore it?:

Starting file watcher (CTRL-C to quit)...
◉ :cp data/.git/objects/28/1e3ddc9b4ecfd2b756ee4709be8018f6b04ba5
◉ :cp data/.git/objects/c8/b9e5a1f12bdb776245f03827b7c19a4f13391e

you can make a .bootignore file

it should have one java-compatible regex per line

so for example:

data/.git/.*

write that in the project directory as .bootignore

@thosmos: ^^