ng-classify should use APP_NAME from config.coffee as it's appName option so that the app works :)

If that sounds right, I'd be happy to submit a PR

Version 2 is failing due to a new dependency in selenium. https://code.google.com/p/selenium/issues/detail?id=8566

tl:dr... Selenium now has Python as a dependency.

There are more issues with this version such as angular/protractor#2134

Roll back to version 1.8

When the gulp watch is listening and you make a change, the recreated dist folder does not contain the 'vendor' folder with external dependencies.

Executed the following steps and this works on a mac but not on windows.

yo fatarrow
gulp

Make a change to html or javascript in src and the browser will reload but the app will not function sine the vendor folder is missing. Going to take a closer look later tonight.

onError handler has @emit 'end' which swallows errors. It would be good to actually not do this when making a prod build.

gulp-less version 2.x.x is not compatible with node 0.12. This issue is documented at gulp-community/gulp-less#156

Running gulp --prod opens a new browser window. This is not desired on a continuous integration server.

I was able to placate my IDE and linter by changing the app module creation to this

class App extends App
    constructor: ->
        #dependencies
        list = [
            'ngAnimate'
            'ngSanitize'
            'ngTouch'
            'ngMessages'
            'ngResource'
            'angular-loading-bar'
        ]
        if '<%= useBackendless %>' == 'true'
            list.push 'ngMockE2E'

        return list

still ugly, but a better kind of ugly

Running gulp test currently runs karma in a child_process. This does not allow the main process to return the correct exit code so that a build would actually fail on a ci server. I tried a couple of approaches. Both work and would love to send PR on this.

1. using a flag on the test task (gulp test --ci). Here I can do something like:

spawn   = childProcess.spawn command, ['karma'], {stdio: 'inherit'}

spawn.on 'close', (code, signal) ->
  throw 'tests failed' if code is 1

2. define a separate task (gulp citest). Here it would look something like:

gulp.task 'utest', ['build'], ->
  gulp.start('karma')

I would love to hear your thoughts. Thanks.

PS: My naming is pretty terrible here and I need better task names here :)

Add a --gzip option should allow you to compress html, css and js for production builds. This is desirable when web server configuration is not an option (serving from an S3 bucket for instance). It removes the complexity of configuring web servers. It also improves the web server performance slightly.

It looks like the package doesn't work with Protractor version 2. Remove the package and use protractor directly: https://github.com/angular/protractor/blob/master/docs/faq.md#how-can-i-get-hold-of-the-browsers-console

e2e tests do not run properly with gulp test command. They require the browserSync server to be running. Need to change this set up so all tests execute with this command. The process should also exit with the correct error code (if any).

Defining decorators in separate files causes an issue with the load order in the dev mode. It would be awesome to have some convention like *.decorator.js and have gulp load these these at the end.

The constant desktop notifications are a bit annoying. I would much prefer to only be notified when there are failures I need to be aware of.

Changes are hard to consume from fatarrow in other projects. It involves a manual task of merging changes. It is especially cumbersome if you have several projects based on this architecture.

One way this can be done is to publish fatarrow to bower or npm. This way, gulpfile.coffee can be donwloaded from bower from an npm postinstall hook for instance.

when using 'gulp --backend', the app launches. Since there is no backend, it errors out. I am thinking that some backend is responsible for consuming the static files output by gulp. I have a PR for this if you agree.

Leverage gulp-newer (https://www.npmjs.com/package/gulp-newer) to only process what has changed. This should speed things up.

This is a much more capable tool than the current gulp-connect plugin.

http://www.browsersync.io/

Add a simple Vagrant file for people who want to use it. It should come with node and Java (selenium dependency).

right now i can only specify a valid bower package version in config.coffee.

I needed to grab the latest revision instead, but got this error

[16:13:02] Starting 'bower'...
[16:13:08] Tag/branch git://github.com/daneden/animate.css.git#a3cf3e6c88e8318b318f8f69c9468940b8aac079 does not exist
[16:13:08] 'bower' errored after 6.47 s

This will make configuration easy. Plus, tools like jshint gutter can leverage this as well.

Start by documenting all switch options and the config.coffee file.

Plato creates static files which can be served with a static file server to easily view.

Same can be done with coverage files and protractor screen shots.

It would be nice to wire up BrowserSync notifications to display links that take you to these sites.

if I run bower list in the project dir I get output like this:

─▪ bower list
bower check-new     Checking for new versions of the project dependencies..
HourlyNerd#0.1.0 /Users/misha/Work/main-app/app
├── angular not installed
├── angular-animate not installed
├── angular-base64 not installed
....

I think it can be fixed by introducing a .bowerrc file that points bower to where the components go, but the weird thing is that bower_components is the standard name as far as I know, so it should just work.

Also, I suspect that running bower install manually will also not work

This is needed for IE conditional comments.

<script src="c:\vendor\angular\scripts\angular.min.js"></script>

Should be

<script src="/vendor/angular/scripts/angular.min.js"></script>

Support deploying a prod build to a path and/or S3 to start with. This functionally can support more providers in the future.

Gulp build currently deletes the bower.json file and recreates it from config.coffee. It would be nice to be able to define bower.json as is per bower specifications. If the file does exist, the gulp build can use it to install dependencies and pick out files per config.coffee configuration. If it does not exist, it can be created for people who do not wish to define it.

Reason for wanting to do something like this is so that now I can create components by using the fatarrow architecture and publish them to bower. May be there is a better (simpler) way?..

bower link allows testing a component locally before publishing it. This is similar to npm link. Currently, we are blowing away bower_components folder on the firstRun. Change this behavior when no-bower flag is specified.

Enhancement: Running gulp --stats, I'd expect to see a coverage folder as well. I currently don't see an option to run karma with coverage.

i am planning on migrating one of my projects to babel and would be great to see it supported here.

This will help prevent functionality breaking due to breaking changes in packages

When bower link used and the consumed less files change, first copy over less files and then run the less compiler on them. The current apporach of merged streams does not work with this and will cause less import errors.

When running 'gulp' the following error is thrown. It looks like the connect server start task is not completing.

events.js:72
        throw er; // Unhandled 'error' event
              ^
Error: listen EADDRINUSE
  at errnoException (net.js:904:11)
  at Server._listen2 (net.js:1042:14)
  at listen (net.js:1064:10)
  at Server.listen (net.js:1138:5)
  at ConnectApp.server (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp-connect/index.js:57:12)
  at new ConnectApp (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp-connect/index.js:37:10)
  at Object.module.exports.server (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp-connect/index.js:135:12)
  at startServer (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/gulpfile.coffee:242:10)
  at Gulp.gulp.task.options.minifyHtml.empty (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/gulpfile.coffee:1067:2)
  at module.exports (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/lib/runTask.js:34:7)
  at Gulp.Orchestrator._runTask (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/index.js:273:3)
  at Gulp.Orchestrator._runStep (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/index.js:214:10)
  at /Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/index.js:279:18
  at finish (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/lib/runTask.js:21:8)
  at /Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/lib/runTask.js:52:4
  at f (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/node_modules/end-of-stream/node_modules/once/once.js:16:25)
  at DestroyableTransform.onend (/Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/orchestrator/node_modules/end-of-stream/index.js:31:18)
  at DestroyableTransform.emit (events.js:117:20)
  at /Users/nachiketmehta/projects/fatarrow-test/fatarrow/node_modules/gulp/node_modules/vinyl-fs/node_modules/through2/node_modules/readable-stream/lib/_stream_readable.js:965:16
  at process._tickCallback (node.js:419:13)

Karma needs to use the html2js preprocessor to include templates that are being loaded externally. Currently unit testing directives with templateUrl is not possible.

Example: http://newtriks.com/2013/04/26/how-to-test-an-angularjs-directive/

Tested this out on the routes directive included in fatarrow which has a templateUrl. Here is the test.

describe 'routes', ->
    describe 'routes directive', ->
        beforeEach module 'app'

        beforeEach inject (@$compile, $rootScope) ->
            @scope = $rootScope.$new()

        it 'renders html', ->
            markup = """
            <routes>
            # fatarrow
            </routes>
            """

            element    = @$compile(markup)(@scope)
            controller = element.controller()
            html       = element.html()
            expected   = '<h1>fatarrow</h1>'

            expect html
                .toEqual expected

This produces Expected '' to equal <h1>fatarrow</h1> because the template does not exist. The expected html should not be empty.

Check JavaScript styles and fix issues.

Recently, codeship is running builds as well as travis. Builds on travis succeed but fail on codeship. @CaryLandholt could you check or give me permission so I can disable codeship.

Hi Cary,

I am not able to access Yargs parameters from the karma task, as i see it is a different child process than the gulp one.

for example

yargs.options 'test',
default : false
description : 'test'
type : 'boolean'

testArg = getSwitchOption 'test'

gulp.task 'karma', ->
console.log " In Karma :", test

if i give the below command "gulp --test ", i assume it should print "In Karma true", but get a false which i think it does not recognize the yargs.

Is there a way i can access yargs in the karma task ?

Thanks,
Kanchan

The bower task takes a long time to check if all my libs are up to date, and I don't need to do this on every reload.

It would be nice to easily turn this off with a flag.

Thanks

I am learning angular and stumbled on your ng-classify repo and then here.
This repo looks like a rewrite of AngularFun. You are commiting code, so I installed it.

The directory structure is confusing. The only reason I can think of laying it out like that is that you havent done that part yet. AngularFun's dir structure looks much more comprehensible.

So, am I shooting myself in the foot by using this vs AngularFun ?
It would be great if you could include something in the readme for this project about why you are building 2 reference frameworks :)

I am trying to upgrade some of the bower components in my project. I went in the config.coffee and changed the version of them. When I stop and restart gulp, I get an error 'Unable to find suitable version for X', where X is the component.

I also tried the same by getting the latest version of fatarrow. I changed the version of angular from 1.2.25 to 1.2.26. I saw the same behavior.

If I delete 'bower_components' folder and then run bower install, things work fine.

I am not sure what it is about calling bower programmatically that is failing. Have you seen this?

It seems like the gulp e2e command should be run while the gulp has the server running. Is that the intended behavior?

Following should be the complete list of globals

PREDEFINED_GLOBALS = [
    'after'
    'afterEach'
    'angular'
    'before'
    'beforeEach'
    'describe'
    'expect'
    'inject'
    'it'
    'jasmine'
    'module'
    'spyOn'
    'xdescribe'
    'xit'
]

Reloading the browser after CSS change is so 2014 :)

This seems pretty straightforward with BrowserSync.

The gulpfile is getting pretty big and hard to maintain. We should investigate a way to break it up in separate files.

I am currently when using the --backend flag to exclude angular's $httpBackend. I have modified gulp tasks to drop off the build output to another location where my backend expects it. All this works great. However, I have been thinking about making the connect server that is spun up a proxy server. This way, you could have your api running wherever, and you can code your frontend against it. I wanted to see if you have any thoughts on this. Shouldn't be hard to wire it up, but do you think it belongs on fatarrow?

E2E tests require a manual action to run. Ideally we want them to run the same way as karma tests do on every file save.

Unit tests should run when using a real or a fake backend.

in the less task, you need to add the following option to options:

paths: [path.resolve(SRC_DIRECTORY)]

We made changes to your watch task in the following ways that greatly speed it up:

1. Introduced a global firstRun flag that gets set to false in the watch task (which depends on all other tasks).
   When firstRun is false, we don't run the bower task. This is better then the flag to simply turn off bower since you get to run it once during a watch. We also don't clean the dist directory when it is false.
2. We split up the watch task to watch different components of the app and update them individually. Here is the not-so-elegant code to do this:

gulp.task 'watch', ['build'], ->
    isFirstRun = false # don't cleanup the dist directory

    extensions = []
    .concat EXTENSIONS.VIEWS.COMPILED
    .concat EXTENSIONS.VIEWS.UNCOMPILED

    sources = [].concat ("**/*#{extension}" for extension in extensions)

    gulp
    .watch sources, {cwd: SRC_DIRECTORY, maxListeners: 999}, ['views']
    .on 'error', onError

    extensions = []
    .concat EXTENSIONS.SCRIPTS.COMPILED
    .concat EXTENSIONS.SCRIPTS.UNCOMPILED

    sources = [].concat ("**/*#{extension}" for extension in extensions)

    gulp
    .watch sources, {cwd: SRC_DIRECTORY, maxListeners: 999}, ['scripts']
    .on 'error', onError

    extensions = []
    .concat EXTENSIONS.STYLES.COMPILED
    .concat EXTENSIONS.STYLES.UNCOMPILED

    sources = [].concat ("**/*#{extension}" for extension in extensions)

    gulp
      .watch sources, {cwd: SRC_DIRECTORY, maxListeners: 999}, ['styles']
      .on 'error', onError

    gulp
      .watch 'index.html', {cwd: SRC_DIRECTORY, maxListeners: 999}, ['spa']
      .on 'error', onError

Unrelated to the above suggestion (or maybe related, because of the modifications we made)

I have an issue where the watch task dies intermittently during a clean like this:

[15:29:57] Starting 'clean:working'...

events.js:72
        throw er; // Unhandled 'error' event
              ^
Error: ENOENT, lstat '/Users/misha/Work/main-app/app/bower_components/_/vendor/angular-strap/scripts/dimensions.js'

Ive tried wrapping things in try catch statements but maybe I did it wrong? Either way, a clean shouldnt be to upset if it cant find a file to delete :)