Air Mozilla

Live: https://air.mozilla.org Stage: https://air.allizom.org Dev: https://air-dev.allizom.org

Air Mozilla is the Internet multimedia presence of Mozilla, with live and pre-recorded shows, interviews, news snippets, tutorial videos, and features about the Mozilla community.

Wiki page: https://wiki.mozilla.org/Air_Mozilla

Most of this information is also available on

https://air.mozilla.org/contribute.json

How to get it running locally from scratch

First of all...

All of these manual steps can be done with one optimistic script. All you need is to have Python (2.6 or 2.7) installed.

Download this file and manually run:

python devup.py

It will ask you a series of questions and how and where you want it installed.

Doing it the manual way...

This section assumes you know about and are using a virtualenv. If you're not familiar with virtualenv, that's fine. You can use your "system python" but a virtualenv is advantageous because you get a self-contained python system that doesn't affect and isn't affected by any other python projects on your computer.

Also, these instructions are geared towards developers working on the code. Not system people deploying the code for production.

Step 1 - The stuff you need

You're going to need Git, Python 2.6 or Python 2.7, a PostgreSQL database (partially works with MySQL and SQLite too) and the necessary python dev libraries so you can install binary python packages. On a mac, that means you need to install XCode and on Linux you'll need to install python-dev.

This article on pyladies.com has a lot of useful information.

Once you have a virtualenv you want to use, you need to install all the dependencies. You do this with:

pip install -r requirements.txt
pip install -r dev-requirements.txt

The second file is necessary so you can

Step 2 - Get the code

Note: We're assuming you have already activated a virtualenv which will have its own pip.

Note 2: Windows users, before you start cloning you need to make sure you're not going to use the git protocol to clone any submodule, otherwise you will get fatal: read error: Invalid argument errors.

git config --global url."https://".insteadOf git://

git clone https://github.com/mozilla/airmozilla.git
cd airmozilla
pip install -r requirements.txt

Step 3 - Create a database

To create a database in PostgreSQL there are different approaches. The simplest is the createdb command. How you handle credentials, roles and permissions is generally out of scope for this tutorial.

createdb -E UTF8 airmozilla

You might at that point need to supply a specific username and/or password. Whichever it is, take a note of it because you'll need it to set up your settings.

Step 4 - Set up the settings

The first thing to do is to copy the settings "template".

cp airmozilla/settings/local.py-dist airmozilla/settings/local.py

Now open that airmozilla/settings/local.py in your editor and we'll go through some essential bits.

DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.postgresql_psycopg2',
        'NAME': 'airmozilla',
        'USER': 'username',
        'PASSWORD': 'password',
        'HOST': '',
        'PORT': '',
    },
    # 'slave': {
    #     ...
    # },
}

Here, replace username and password accordingly. If you want to use MySQL, which should work except the Search, you replace the ENGINE setting with django.db.backends.mysql.

For local development, make sure the following lines are uncommented:

DEBUG = TEMPLATE_DEBUG = True

Since Air Mozilla uses a lot of AJAX calls, it may not be useful for errors to show up in the browser when they happen. It may be more useful to have all the errors appear in the terminal console. This is a matter of personal taste, but if you want all the errors to appear in the terminal add this line:

DEBUG_PROPAGATE_EXCEPTIONS = True

And, since you're probably going to run the local server NOT on HTTPS you uncomment this line:

SESSION_COOKIE_SECURE = False

For security, you need to enter something into the SECRET_KEY.

SECRET_KEY = 'somethingnotempty'

By default, you need a Memcache server up and running. The connection settings for that is not entered by default. So if you have a Memcache running on the default port you need to enter it for the LOCATION setting like this:

CACHES = {
    'default': {
        'BACKEND': 'django.core.cache.backends.memcached.MemcachedCache',
        'KEY_PREFIX': 'airmoz',
        'TIMEOUT': 6 * 60 * 60,
        'LOCATION': 'localhost:11211'
    }
}

If you want to use a local in-memory cache instead, use the following CACHES setting:

CACHES = {
    'default': {
        'BACKEND': 'django.core.cache.backends.locmem.LocMemCache',
        'LOCATION': 'unique-snowflake'
    }
}

By default, all the default settings are geared towards production deployment. Not local development. For example, the default way of handling emails is to actually send them with SMTP. For local development we don't want this. So uncomment the line EMAIL_BACKEND to:

EMAIL_BACKEND = 'django.core.mail.backends.console.EmailBackend'

The majority of the remaining settings are important security details for third-party services like AWS, Vid.ly, Mozillians, Twitter, Bit.ly and Sentry.

Getting actual details there is a matter of trust and your relationship to the project. Some things will obviously not work without these secrets, the file upload for example won't work without AWS credentials. But things aren't necessary should be something you can just go around. If something doesn't work at all without having certain security settings, it's considered a bug.

Step 5 - Running for the first time

The very first thing to run are these commands:

./manage.py syncdb
./manage.py migrate

The syncdb command will ask you to set up a first default superuser. Make sure you use an email address that you can log in to Persona with.

And last but not least:

./manage.py runserver

Now you should be able to open http://localhost:8000.

Step 6 - Running the Celeryd message queue worker

We use a message queue to execute some slow tasks in the background. By default it doesn't need neither RabbitMQ or Redis to hold the jobs. Instead, by default, it uses our own Django and PostgreSQL which we use for everything else.

To start the message queue worker use:

./manage.py celeryd

To test that this is working, once you've signed in to the site for the first time, and signed in as a super user you will get an extra link in the Management pages called "Tasks Tester". Click that to make a test that your message queue is working.

Random/Sample data

Now that you have a working version running you'll find that it has nothing in it. This will make it harder to get a feel for certain features and it definitely makes it hard to optimize things where there's so little there.

There is a command that is under active development and creates the bare minimum. If you find it's not creating certain type of data that you need then feel free to send in a patch on the fakedata.py code.

You can run this command repeatedly and it should not break.

./manage.py generate-fake-data 1000

The number at the end is how many events you want to generate.

This will also create various other things like random Regions, Locations, VideoTemplates, Pictures, Users etc.

How to get it running with Docker Compose

You need to install Docker and Docker Compose in your machine and then you can build the image, but first you have to modify your airmozilla/settings/local.py file so it can connect to the database and memcached.

DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.postgresql_psycopg2',
        'NAME': 'postgres',
        'USER': 'postgres',
        'HOST': 'db',
        'PORT': '',
    }
}

CACHES = {
    'default': {
        'BACKEND': 'django.core.cache.backends.memcached.MemcachedCache',
        'KEY_PREFIX': 'airmoz',
        'TIMEOUT': 6 * 60 * 60,
        'LOCATION': 'memcached:11211'
    }
}

After this you can run the app by doing:

docker-compose up

With this the Django test server will we running, so if your are on Linux you can connect to localhost:8000 or to 192.168.59.103:8000 if you are on Windows or Mac. Due the fact that on Windows and Mac docker uses boot2docker this ip can change at some point, to know the exact ip address you can execute boot2docker ip.

If you want to run regular tests you can execute:

docker-compose run web ./manage.py test

If you want to run all included selenium tests, don't forget to add RUN_SELENIUM_TESTS = True and then execute:

docker exec -it airmozilla_web_1 ./manage.py test

Also if you want to ssh into the running container for debugging you should execute:

docker exec -it airmozilla_web_1 bash

How to contribute

There are levels of contribution. All are appreciated.

Filing bugs

The best start you can get is to file bugs when you spot things that are broken or could be made better. Don't be shy with filing bugs that are actually feature requests. Your voice will always be appreciated.

To file a bug go to this URL

Taking bugs/Finding bugs to work on

If you spot a bug that you would like to work on, you can either just get started and when you present your patch you hope that nobody else was working on it at the same time. Or you can post a comment on the bug saying you'd like to work on this.

An even better way would be to jump into IRC on the #airmozilla-dev channel and ask around about the feature/bug you intend to work on.

We don't assign bugs to people until after the bug is resolved.

Writing code patches

All code patches have to be submitted as GitHub Pull Requests on https://github.com/mozilla/airmozilla.

When a pull request is made, our automation will check a couple of things:

1. Strict PEP8 and pyflakes standards. If your patch introduces code with incorrect indentation or lines too long the pull request will fail. This rule is there to remove any debate about how to style code as per how a machine likes it. It makes it non-subjective and clear.

2. All tests are run on Travis in Python 2.6. If any test fails, the pull request fails.

3. Test coverage regression. Test coverage is measured for every pull request and if you introduce a patch that has more features/changes than it has test coverage the test coverage percentage goes down and this doesn't necessarily fail the pull request but it will be less likely to be merged.

When you start working on a patch, please try to make it only about the bug you're working on. Try to avoid fixing other things that aren't related to the issue at hand.

Always start your work in a new git branch. Don't start to work on the master branch. Before you start your branch make sure you have the most up-to-date version of the master branch then, make a branch that ideally has the bug number in the branch name.

Also, your git commit message should contain the bug number. If you think by merging this patch it should resolve the bug you can add the prefix "fixes" before the bug number. Like this for example:

git pull origin master
git checkout -b bug123456789-fixing-that-thing
...work work work...
git commit -a -m "fixes bug 123456789 - Fixing that thing"
git push myforkremote bug123456789-fixing-that-thing

When you have created a pull request on GitHub, take the URL to the pull request and post that as a comment on the bug on Bugzilla.

Rebasing branches

Oftentimes, when you start on a branch and make a pull request you might be asked to correct things and add more commits to it until the tests pass and it's ready to be merged.

You might then be asked to rebase and squash your branch into a single succinct commit. This makes it easier to look back into the commit history when we reflect on what we've done in the year or last year.

If you're unfamiliar with rebasing, there are plenty of tutorials online but you can also say so and we can rebase it for you manually.

Easy does it

Any change is a good change. Getting warmed up and familiar with the code base is best done in incremental steps. Try to start small and go through the steps until the code has landed. The sense of accomplishment for getting your name into the commit log history is a great boost for tackling more advanced features or bugs.

Code style

Consistency is key. Mixing conventions and styles makes code look un-maintained and sticking to one way makes it easier to just do instead of having to think about style choices.

For Python, all code must be PEP8 and pyflakes compliant. See the section on PEP8 and pyflakes. There are some things that flake8 can't automatically check for. For example some choices on indentation using newlines to split code up.

# bad
channel = Channel.objects.create(name="Something",
                                 slug="something")

# ideal
channel = Channel.objects.create(
    name="Something",
    slug="something"
)

For Javascript, use the notation of spaced before and after brackets.

// bad
if(number < 42){
    return 'less';
}

// ideal
if (number < 42) {
    return 'less';
}

For both Javascript and Python there is no rule on using single quotation marks (') or double quotation marks ("). But what ever the file is using, try to stick to that.

For CSS ideally we avoid one-liners. Feel free to use plenty of space.

/* bad */
.event, h2.summary{background-color:#fff;font-size:10px}

/* ideal */
.event,
h2.summary {
    background-color: #fff;
    font-size: 10px;
}

Getting help

The best place to get help on development is to go the the #airmozilla-dev IRC channel on irc.mozilla.org.

Tests and test coverage

Included is a set of comprehensive tests, which you can run by: ./manage.py test

If you want to run selenium tests don't forget to add RUN_SELENIUM_TESTS = True in your airmozilla/settings/local.py file.

To see the tests' code coverage, use: ./manage.py test --with-coverage --cover-erase --cover-html --cover-package=airmozilla

Then, when it completes, open the file ./cover/index.html.

You can run tests with any level of granularity: To run a specific file, use: ./manage.py test -s -x airmozilla/manage/tests/test_forms.py

To run a specific test case class in a file: ./manage.py test -s -x airmozilla/manage/tests/test_forms.py:SomeTestCaseClass

To run a specific test in a class in a specific file: ./manage.py test -s -x airmozilla/manage/tests/test_forms.py:SomeTestCaseClass.test_some_function

The -s makes it so that any print statements aren't swallowed if tests pass. The -x means it bails as soon as 1 test fails.

Troubleshooting

Deprecated Files

Make sure you don't have a ./vendor or ./vendor-local directory. This functionality has been deprecated. All dependencies are now contained in the requirements.txt file.

Unable to sign in

There are several reasons why sign in might not work. A common problem is that you have problems with CSRF and those are usually because of caching not working. Or a security setting.

If you see this in the runserver logs:

15/Oct/201X 14:53:37] "POST /browserid/login/ HTTP/1.1" 403 2294

It means that the server gave you a cookie which couldn't be matched and checked when sent back to the server later.

To check that caching works run these blocks:

./manage.py shell
>>> from django.core.cache import cache
>>> cache.set('some', 'thing', 60)
>>> ^D

(^D means Ctrl-D which means to exit the shell) then

./manage.py shell
>>> from django.core.cache import cache
>>> cache.get('some')
'thing'

If it doesn't say 'thing' it means it wasn't able to cache things. This is most likely because you have configured django.core.cache.backends.memcached.MemcachedCache as your preferred cache backend but that server isn't up and running.

Consider then to either figure out how to start your memcache server or switch to django.core.cache.backends.locmem.LocMemCache.

Another common mistake is to not have SESSION_COOKIE_SECURE = False in your airmozilla/settings/local.py but using http://localhost:8000 to reach the site.

Tests are not working

If tests don't work around code you didn't touch, it might be that your test database is out-of-sync so then next time simply run: FORCE_DB=1 ./manage.py test.

Migrations

We're using [South][south] to handle database migrations. To generate a schema migration, make changes to models.py, then run:

./manage.py schemamigration airmozilla.main --auto

or

./manage.py schemamigration airmozilla.comments --auto

To generate a blank data migration, use:

./manage.py datamigration airmozilla.main data_migration_name

Then fill in the generated file with logic, fixtures, etc.

To apply migrations:

./manage.py migrate airmozilla.main
./manage.py migrate airmozilla.comments
./manage.py migrate airmozilla.uploads
./manage.py migrate airmozilla.subtitles
./manage.py migrate airmozilla.surveys
./manage.py migrate airmozilla.cronlogger

In each command, replace airmozilla.main with the appropriate app.

Requirements

See the requirements/ directory for installation dependencies. This app requires a working install of PIL with libjpeg and libpng.

First run

./manage.py syncdb
./manage.py migrate

If you'd like to create a default set of example groups with useful permissions (Event Organizers, Experienced Event Organizers, PR, Producer):

./manage.py create_mozilla_groups

Logging in

We use Persona to handle all log in. If you haven't used it before, it's fine. It's free and easy and works with any email address.

Because the code is built to only allow people with certain email address domains, (e.g. mozilla.com) you might need to fake this if you don't have a mozilla.com email address. To do that, open the file airmozilla/settings/local.py and add to the bottom this:

ALLOWED_BID = base.ALLOWED_BID + (
    'gmail.com',
)

...assuming your preferred email address is a gmail.com one. But note, only enter the domain of your email address. Not the whole email address.

Becoming a Superuser

Superusers have full unbound permissions to do anything and everything.

On a blank database with no content, the only way to become a Superuser is to sign in once, then go to the command line and manually change the data so that you're now a superuser.

./manage.py shell
>>> from django.contrib.auth.models import User
>>> my_user = User.objects.get(email='[email protected]')
>>> my_user.is_superuser = True
>>> my_user.is_staff = True
>>> my_user.save()

Optimizing thumbnails

By default, all the thumbnails that are generated from uploads and from scraped screencaptures get generated as .png files. These tend to be pretty large and can be optimized down without virtually any visual difference. To enable it, install pngquant and the override the local settings and set it like this:

PNGQUANT_LOCATION = 'pngquant'

Adding a sample video

Once you're a superuser, the simplest way to add sample content is to use existing media from Air Mozilla. We're first going to create a template, then create content based on this template.

Let's create the Video template:

• Go to the management page at http://localhost:8000/manage

• Click on Video templates in the left menu, and initiate a template with New template

• Fill the Name with "Vid.ly"

• Fill the Content with

  <iframe frameborder="0" allowfullscreen width="896" height="504" name="vidly-frame"
   src="https://vid.ly/embeded.html?link={{ tag }}&amp;new=1&amp;autoplay={{autoplay}}&amp;hd=yes"></iframe>

• Save changes

We're now going to initiate an event with this template:

• Click on Initiate event in the left menu of the management page
• Choose the Vid.ly template
• Replace the Template environment with tag=7u9u1i
• Change status to Scheduled
• Choose a title, a description, a start time, a channel (Main) and a placeholder image
• Save and submit

You should now see your video in the Event manager (http://localhost:8000/manage/events/) and on the main page (http://localhost:8000/)!

Contributing

See CONTRIBUTING.md

IRC

irc://irc.mozilla.org/airmozilla-dev

Cron jobs

All cron jobs are managed by two files: First airmozilla/manage/crons.py where you kick things off. Then, later the crontab file is compiled and installed by bin/crontab/crontab.tpl. These are the two files you need to edit to change, add or remove a cron execution.

Twitter

To test tweeting locally, what you need to do is set up some fake authentication credentials and lastly enable a debugging backend. So, add this to your settings/local.py:

TWITTER_USERNAME = 'airmozilla_dev'
TWITTER_CONSUMER_KEY = 'something'
TWITTER_CONSUMER_SECRET = 'something'
TWITTER_ACCESS_TOKEN = 'something'
TWITTER_ACCESS_TOKEN_SECRET = 'something'

Now, to avoid actually using HTTP to post this message to api.twitter.com instead add this to your settings/local.py:

TWEETER_BACKEND = 'airmozilla.manage.tweeter.ConsoleTweeter'

That means that all tweets will be sent to stdout instead of actually being attempted.

To send unsent tweets, you need to call:

./manage.py cron send_unsent_tweets

This can be called again and again and internally it will take care of not sending the same tweet twice.

If errors occur when trying to send, the tweet will be re-attempted till the error finally goes away.

Deployment

Deployment of dev, stage and prod is all done using Chief. More will be written about it soon.

Bit.ly URL Shortener

To generate a Bit.ly access token you need the right email address and password. If you have access you can go to https://bugzilla.mozilla.org/show_bug.cgi?id=870385#c2

To generate it use this command:

curl -u "EMAIL:PASSWORD" -X POST "https://api-ssl.bitly.com/oauth/access_token"

That will spit out a 40 character code which you set in settings/local.py for the BITLY_ACCESS_TOKEN setting.

About the database

Even though we use the Django ORM which is database engine agnostic, we have to have PostgreSQL because we rely on its ability to do fulltext index searches.

PEP8 and pyflakes

All (with some few exceptions) code needs to be fully pep8 and pyflakes compliant. And line length for Python is expected to be under 80 characters wide.

The help yourself enforce this automatically, you need to set up the following git hooks. First, in your virtualenv, install this:

pip install flake8

Next you need to create (or amend) the file:

.git/hooks/pre-commit

...to contain the folllowing...:

#!/bin/sh

exit_code=0

for file in `git diff --cached --name-only --diff-filter=ACM | sort | uniq`
    do
        if [ ${file: -3} == ".py" ]; then
            flake8 $file
            if [ "$?" -ne "0" ]; then
                exit_code=1
            fi
        fi
        if [ ${file: -3} == ".js" ]; then
            jshint $file
            if [ "$?" -ne "0" ]; then
                exit_code=1
            fi
        fi
    done

if [ "$exit_code" -ne "0" ]; then
    echo "Aborting commit.  Fix above errors or do 'git commit --no-verify'."
    exit 1
fi

And make sure it's executable by running:

chmod +x .git/hooks/pre-commit

Next time you type git commit -a -m "fixes bug" it might block you and spit out a message like this instead:

airmozilla/main/views.py:17:1: F401 'Sum' imported but unused
airmozilla/main/views.py:378:80: E501 line too long (81 > 79 characters)
Aborting commit.  Fix above errors or do 'git commit --no-verify'.