MS really made a mess of things with its intentional unix incompatibilities. All that has changed now and Windows 10 fixes many, but not all, the issues. Only I figure that over time the number of ads will increase to the point where you must pay to turn them off.

i couldn't figure out a way to detect this situation

and the exception is a generic one

i suppose when there's an exception it can just try again with different settings

The prefix concept is used to handle root directories on UNIX platforms, and drive specifiers, root directories and UNC pathnames on Microsoft Windows platforms, as follows: For UNIX platforms, the prefix of an absolute pathname is always "/". Relative pathnames have no prefix. The abstract pathname denoting the root directory has the prefix "/" and an empty name sequence. For Microsoft Windows platforms, the prefix of a pathname that contains a drive specifier consists of the drive letter followed by ":" and possibly followed by "\\" if the pathname is absolute. The prefix of a UNC pathname is "\\\\"; the hostname and the share name are the first two names in the name sequence. A relative pathname that does not specify a drive has no prefix.

from http://docs.oracle.com/javase/8/docs/api/java/io/File.html

like given two paths: new java.io.File("/foo/bar") and new java.io.File("/baz/baf")

i'm not sure how to know if those are on the same filesystem or not

if they're not then some operations will fail

i think that windows exception "can't move file to another device" or whatever, i think that's actually being thrown when the .createLink() method is being called

Convert to absolute paths and look at them.

i guess on windows you can't mount filesystems in arbitrary directories?

because in unix the path doesn't tell you anything about which filesystem it's on

yup! Typically, a file system requires that all links (directory entries) for a file be on the same file system. Furthermore, on some platforms, the Java virtual machine may require to be started with implementation specific privileges to create hard links or to create links to directories.

from https://docs.oracle.com/javase/7/docs/api/java/nio/file/Files.html#createLink(java.nio.file.Path,%20java.nio.file.Path)

yeah it's interesting that windows says it can't move a file

when nobody's trying to move a file

lol

But them's hard links. How about soft links? http://www.howtogeek.com/howto/16226/complete-guide-to-symbolic-links-symlinks-on-windows-or-linux/

api for symbolic links: https://docs.oracle.com/javase/7/docs/api/java/nio/file/Files.html#createSymbolicLink(java.nio.file.Path,%20java.nio.file.Path,%20java.nio.file.attribute.FileAttribute...)

soft links won't work because they need to point to something

with a hard link you can treat it as a regular file

and if you delete the original it will still be a valid file

i mean if you delete the original, the link will still be a valid file

ok. I was wondering about this: The target parameter is the target of the link. It may be an absolute or relative path and may not exist. When the target is a relative path then file system operations on the resulting link are relative to the path of the link.

so in windows all absolute paths have a drive letter prefix?

and all filesystems mounted must be assigned a drive letter?

if this is true then the path comparison test will work on windows

y

As for deletion, you can write a function to test for links and do what you consider right. 😄

Such joy!... not

yeah really i need to make a function that wraps java.nio.file.Files/createLink

and throws proper exceptions

But in reading the java api, it looks like delete always just deletes the link if it is one

yeah i'm using this code:

https://github.com/boot-clj/boot/blob/master/boot/pod/src/boot/filesystem.clj#L54-L60

there's still a race condition there, but i'm using a semaphore to synchronize these operations

so delete if exists deletes the link if it exists.

yeah the problem is with createLink()

yup

Just always use symbolic links?

if the destination and the source are on different filesystems it says "can't move to different filesystem"

symbolic links don't really work for this application

oh dear!

i can just catch exceptions

and then do some heuristics to figure out what caused it

well, that's better than nottin.

yeah and rethrow with a more specific exception

I usually do not have a second drive, but it stinks not to support it.

yeah and we want to warn the user also

read, commercial apps?

because in that case we'd have to fall back to copying all the files

which is obviously a lot less efficient

yeah

As an old java/windows user I generally just work with strings and avoid links. Its strings in the config file anyway.

boot uses hard links to avoid copying files

yeah

and unix devs will never give up their links.

structural sharing

a wonderful thing

yeah, if some poor fool hasn't created a bowl of spagetti!

they're probably rich, that fool

rich as in money, not as in hickey, that is

I've seen too many operators with only a hs ed. And in the old days most devs only had 6 mo of trade school!

Yeah, give the manager's raises to the person who created the mess.

--that will be the day!

I did a lot of research today. Looking for reactive hoplon sites.

https://github.com/exicon/homepage

And I found your comment (defc page-width) (.on (js/jQuery js/window) "resize" #(reset! page-width (.width (js/jQuery js/window)))) (defc= dynamic-rules (cond (<= page-width 800) {:background-color "red" :border "4px black"}))) (div :css (cell= (merge {:height "400px"} dynamic-rules)))

--frame that somewhere, will you?

As a non-js dev, I see a ton of value in hoplon, but a terrible lac of readable docs!

Looking for "hoplon for dummies"

Or at least some reactive demos that will run on any device.

I mean, all the capability is there, @micha but it is not intended for folk like me.

documentation is hard

I know

Ah, I just found this HN thread: https://news.ycombinator.com/item?id=9671106 might be worth posting it somewhere i dont know

And code is just way too much fun!!!

nice link. and yes, boot is getting there. But there is a huge market for haplon that can not yet be addressed.

Hoplon would be great for a first web page!

@escherize: whoa

You could have beginners writing reactive web pages on their first day.

yeah that's the dream!

@escherize I really liked that link too!

So get funded why not? 😄

It is a wonderful dream

I plan to ask a lot of dumb questions here. And one thing I know is that you need a sample audience.

I also want to teach my wife, who knows html and css but is not a programmer. I figure hoplon would get her coding without her knowing it. 😄

exactly

i mean if you use html and css you're already programming

html is a programming language, just a really bad one

The integration is so painless in hoplon.

using js on a web page is normally deep magic

--especially for non-programmers.

like when a html file is loaded in the browser it's parsed and lexed, then it's built into an AST, and evaluated to produce the DOM

hoplon just adds a means of abstraction and better primitives to that, basically

yes, thank god! I am hopeful that one day I can create nice web pages. Though my wife will need to provide the visuals. 😄

Actually my gf knows html/css as well. Been slogging through learning parinfer with her.

I built my company's frontend on re-frame, never really gave hoplon a look, though i did like the Javelin concept.

She's done just a tiny bit of reagent/hiccup editing for us

@escherize There's still, I suspect, some work needed before we see commercial apps in production.

there are commercial apps in production

But mostly there is this lack of docs for the less technically inclined, or for those who just are not web devs.

I find everything is hoplon very easy and friendly, especially after talking with @micha

Glad to hear that @micha. But where is the hoplon sponsor page?

The nyc lisp clubs spend $1,400 on a joint cristmas party paid entirely by commercial sponsors!

would you believe I've been using clojure for 2 years and never even noticed

:source-paths ["src"]
:resource-paths ["assets" "out"]

yup. That kind of thing happens a lot.

get ready for more: https://github.com/technomancy/leiningen/blob/master/sample.project.clj

I'll stick with generating project.clj files from build.boot.

--Ijust need it to make cursive work.

Boot is just so right compared to build, make, ant, lein etc etc etc.

Thanks so much for that Micha and friends!

🍻 happy you enjoy it!

@laforge49: 🍪 for everyone!!!

me and much of the clojure community!

--especially those working in clojurescript!

@flyboarder I find it funny that there is much talk about hoplon being an alternative reactive framework. But I looked at demos today and except for one commercial project, I couldn't find any demos that would resize for mobile/tablet/desktop. Seems like a huge hole!

Making this easy for mere mortals would be a huge selling point. Easy enough, I am sure. But then where are the demos? (I know. hoplon is so wonderful that everyone is too busy using it!)

And yes. I am being selfish. I want working examples that i can crib into my code. 😄

An interesting set of demos is a first, non-trivial, step. Once you have that, writing docs will still be very hard, but will be a bit more manageable.

Hoplon by example, as it were.

yeah

Preferable examples that run fine on mobile.

i mean there are a lot of demos in the hoplon/demos repository

more than most projects i think

Ah, I only went through the links on the web page.

u have a resizing demo???

the problem with demos and mobile is that the problem is still pretty much open

there are no good solutions yet

there are things like what twitter bootstrap does

yes.

"responsive design"

but that's a hack that involves a lot of incidental complexity

some of the demos use bootstrap though

and hoplon is perfect for defining the solution.

hmm?

right but until then the demos would be unnecessarily complex

hoplon/boot that is.

ok

lets not waste the effort then.

some folk will anyway and that will be helpful, of course.

i've made commercial hoplon apps that were mobile, web, etc

super

and I gather they were needlessly complex.

needfully?

well they needed to work

yah

i don't think it was that bad

we did things like what you pasted above

good!

A bit of wrapping needed then, hmm?

--the work being in defining the api, of course!

well sometimes you need to completely swap a widget for another one on mobile

ouch

like on a webpage you have a ton more space

but again, handled by a common wrapper.

compare that to a 3" phone display

well it ends up being two different applications

it is good boot can be made so smart.

trying to pretend that they're the same usually adds totally unnecessary layers

sure--different layout and all that

agreed

you need to just omit a lot of functionality for mobile

we used http://trigger.io to wrap the hoplon app in a native app container

so we could put it in the app store and whatnot

and it was usable like a regular app

I figure demos should be largely designed for demos, and for tablets etc you just add things.

--better than omitting things

yeah that's pretty much how we did it

for mobiles I mean

we started with a mobile web app

then we built on top of that for the desktop web app

and then we packaged the mobile web app as a native app for the app store

yeah, ouch. but ok when someone is paying for it.

i don't know of any other way to do it

no one wants web apps on safarie

oh by web app i mean website

not browser extensions or plugins or anything like that

download a real browser,, baby!

I've seen a lot of prejudice among apple users against non-native apps. 😄

But then, that market has the money.

http://trigger.io sounds cool. But all this is really way out of my field!

yeah i haven't checked in on http://trigger.io lately

progress moves fast in that area

new tools coming along all the time

fer sure!

its such a huge market!

Note to self: Boot will not work on a Windows 10 VM inside Parallels hosted on a Mac if you have Home folder sharing enabled...

(because it cannot correctly create the hard links that it wants)

@micha: How does Boot decide whether it has completed its first run?

I installed Boot.exe on Windows 10 on a VM, mistakenly had Home folder sharing on, and Boot failed (see above). Fair enough I though, just disable Home folder sharing and reboot.

Now Boot won’t run — I’m getting

java.io.IOException: The network name cannot be found
at java.io.WinNTFileSystem.createFileExclusively(Native Method)
at java.io.File.createNewFile(File.java:1879)
at Boot.download(Boot.java:168)
at Boot.install(Boot.java:224)
at Boot.main(Boot.java:250)

https://github.com/boot-clj/boot-bin/blob/master/src/Boot.java#L168 — failing line

So I guess it’s finding a network name from some cached location… but I’m not sure where?

@micha: progress! http://p.tcrawley.org/i/ff03b.png

Ah, I’m guessing it’s something to do with environment variables pointing Boot to the wrong place… will do some digging and report back…

@tcrawley: that looks great

Hmm, I might have to try to build boot-bin from source on that VM to add some debugging to figure out what it doesn't like

@laforge49: regarding examples, have you ever looked through the code for the old hoplon site?

@micha: https://clojars.org/tcrawley/boot-verify & https://github.com/tobias/boot-verify

current output looks like: http://p.tcrawley.org/i/2c318.png

if you want to fold that in to boot show, I'm happy to help

or leave it as a standalone task, I'm cool with that as well

@tcrawley: i think it would be great to have as part of the show task

okiedoke - I can work on a PR tomorrow

awesome!

FYI, I'm reaching out to Jean Niklas about lifting the Leiningen new/template code to turn into a Boot task, to make sure they're cool with me creating a boot new task that can pull templates from Clojars etc and render new projects.

Specifically to be able to render Leiningen templates as well as similar Boot templates...

I was reading over the lein new/template code and was surprised how small it all is... Powerful for something so simple

added a couple of code analyzers drafts kibit and yagni to https://github.com/tolitius/boot-check logically eastwood and bikeshed are next. note: keep in mind most of code analyzers don't support reader conditionals yet

very nice!

Jean Niklas says they'd be very happy for us to go ahead with a boot new style task based on the Leiningen code! So I'll go ahead and take a run at that.

why would he object?

(it would be awesome btw, thanks in advance)

@seancorfield: I'm not that familiar with lein templates - can it only create whole projects, or can it as well emulate something like Rails generators, when you can define code that will generate some part of the project. Having an option to do something like boot new some-framework-template generate component User would be awesome.

@jaen: i think maybe generators or scaffolding belongs in a dependency package that exports a new task of its own, like this:

oh i get it re: coordinating with lein, possible "namespace" interference

I suppose that also could work, though looks slightly less appealing syntactically.

lein new some-template just creates a new folder tree from a template but we can do a bit better because boot new -t some/framework can generate based on a template but boot new -g component=User can generate a component (`User`) based on the current project.

And I'm unsure about extensibility - if I understand correctly --component would be a keyword option on some task, right? Then how could this be extensible with new types of generators?

So the template project can include a dependency that boot new knows how to use to generate new pieces of code.

@alandipert: It was more a case of "Would the Leiningen folks feel aggrieved if Boot could generate projects from lein-template templates that already exist — which I think is an important use case to get us off the ground.

@seancorfield: gotcha, yeah i'm tracking, and i agree

Since if both Leiningen and Boot can use lein-tempate projects, then template authors have more of an incentive to just add build.boot (and maybe boot.properties) to an existing template and we get Boot support out of the box.

@jaen: that's the idea, it would be a different task for each generator

i guess they wouldn't all be able to use the name generate

@jaen: Well, we can discuss how best to have both templates and generators… @micha

sure, i was just throwing things against the wall there

Yeah, I guess that's a bit of a problem - Rails generators are great in how uniform they are

rails is like the opposite of boot though

i wonder about resolving against github instaed of maven

People can expect if some library provides a way to generate some code

for templates only

philosophically

Then it can be accessed in the usual way

Well, sure I can agree the Rails world is at odds with the Clojure way of doing things

But convention are a useful thing

github, that's an interesting idea

Feel free to add issues and comments here https://github.com/seancorfield/boot-new

we can own the transitive dep info convention

(that’s just a brand new repo where I’ll write the task etc)

If some convention could be figured out that would let those generative tasks be uniform

also presumably a template will never depend on another template

@alandipert: I like the github for templates idea.

It might be a huge boon to anyone who would want to make a CoCy web framework in Clojure

And boom, we're suddenly as big as Rails ; d

I'm exaggerating a bit here of course

I plan to add Boot support into two Leiningen templates I maintain, so having boot new -t some/template be able to use Leiningen templates will be useful for me at least

cool, yeah maybe the github thing we do later. i'm attracted to it because i never want old versions of templates

But a common, extensible way for templating projects in whole or in part would be a huge boon.

As for github

I'm not sure I like it

At least in the way it being limited to github

maybe it needn't be

Any git repo on the other hand (so I can use, say, gitlab) - sounds good and lowers the barrier to creating new templates.

Agreed - should be generic git url.

Oh, and my Windows 10 weirdness from yesterday is due to Java thinking my user.home folder is still the \\Mac\Home network folder that it was when Parallels was still sharing my host home folder… 😞

Another example is git, where you have main git command and then subcommands that have their own, different options tailored to the semantics of a command. There's even an example in tools.cli how something like that could be implemented. I'm not sure how boot's CLI parser is done and it's certainly geared more towards understanding a pipeline composition rather than subcommands, but if generators are to be useful (and used) they need to be ergonomic and I'm just wondering if there could be a way to simplify the CLI syntax for that case. Also the fact that boot -d some/framework generate --component User would be hardly extensible (you can't add options to an already existing tasks if I understand correctly) also seems problematic in terms of composability of those generators. Maybe they could warrant their own syntactic case and/or macro? I don't know.

@jaen: https://github.com/boot-clj/boot/issues/346#issuecomment-165934484

maybe generators belong in a shebang style boot script

the pattern there gives you the full git-like subcommand experience

different from boot's rigid pipeline cli

seems like it might make sense to have that in source control, maybe, like a migration

your project plus the script used to generate the scaffold

"Also the fact that boot -d some/framework generate --component User would be hardly extensible " — thats why I suggested boot new -g component=User since it could use something in the project itself as the generator...

This sounds sensible. I imagine keeping generators as tasks would be a good idea in that you could easily compose them up to building a whole-application template from a composition of generators, but having a script that supports calling them outside of the rigid boot CLI structure would be helpful.

@seancorfield: ah, right, sorry - didn't pick up on that.

So boot new -t template would generate a project (`build.boot` etc) and then boot new -g could make use of something the project provides. Well, the project template could also provide its own generate task in the build.boot file so generate would be based on the project itself. We’ll have to discuss what is the best approach.

Oh the horror… So Java uses entries in the registry to figure out your home directory on Windows apparently and Parallels had \\Mac\Home\Desktop etc in the registry even after I deactivated sharing 😞 https://clojurians.slack.com/archives/boot/p1451583324007718

Anyways, a quick bit of regedit and Boot is back up and running on my Windows 10 VM now.

#windows

@seancorfield: maybe a note on the wiki?

Is there a good page for Windows stuff?

2.5.5 has an update that fixes an issue with windows btw

there is a page for developing on windows, but nothing authoritative yet

i have a placeholder in the wiki home page for it