First, you'll need the following depencies installed on your OS :

• python (+ dev package)
• libxml2 (+ dev package)
• libxslt (+ dev package)
• rabbitmq

To setup mosileno :

Clone this

git clone https://github.com/emillon/mosileno-web.git

Create a virtualenv

virtualenv ~/.virtualenvs/mosileno

Activate it

. ~/.virtualenvs/mosileno/bin/activate

Install the dependencies

pip install -r requirements.txt

Download the dev (test) topic model file (s.a. hn.ldamodel or hn_lemmatized.ldamodel):

mkdir topic-model
curl -O https://dl.dropbox.com/u/14035465/hn40_lemmatized.ldamodel
mv *.ldamodel topic-model/

Install the package

python setup.py develop

Configure rabbitmq (as root)

rabbitmqctl add_user mosileno mosileno
rabbitmqctl add_vhost mosilenovhost
rabbitmqctl set_permissions -p mosilenovhost mosileno ".*" ".*" ".*"

Populate the database

initialize_mosileno_db development.ini

Boot celery workers

pceleryd development.ini --autoreload

Run the dev server

pserve --reload development.ini

That's it ! The app should be live at http://localhost:6543/.

You can monitor the state of queues with Celery flower. To do that, just install it in your virtualenv :

pip install flower

And run it with :

celery flower

It will be accessible at http://localhost:5555/.

It's also possible to monitor the queues using the builtin celeryev ncurses monitor :

pceleryev development.ini

The dev process is heavily inspired from Github flow, ie :

• master is always deployable
• do feature branches

So, when developing a new feature :

• branch off master
• hack
• commit often (in logical commits)
• it's ok if you push a feature branch where tests break
• when you're ready to merge, merge master INTO the branch and push
• when Travis is OK, you're ready to merge :
  • ensure that you're up to date (git fetch)
  • git checkout master
  • git merge --ff $FEATURE
  • git push origin master :$FEATURE (:branch deletes the remote branch)
  • git branch -d $FEATURE
• if the push is rejected because someone pushed in the meantime, don't merge directly on master. Reset the feature branch, merge master into it and push to travis

Ideally, the merge is done after a pull request for code review.

• Travis is always right.
• pep8 is often right.
  • except with SQLAlchemy. Add --ignore E711 to disable checking comparisons to None.
• tests are your friends.
• If you don't write tests, you're making Travis lie to you.
• If you love Travis (you should), you should feel bad not writing tests.

You'll need alembic (in requirements.txt). Its tutorial is great.

Two migration types are possible : non-automatic and automatic ones. In non-automatic ones, you have to manually define the upgrade and downgrade functions.

But for simple cases (mostly {table, column} {addition, removal} - see tutorial), alembic can automatically detect changes to the model and generate a script.

alembic revision -m $description

And edit scripts/versions/$id_$description.

You also have to edit models.py, but it's an independent step.

Begin by editing models.py.

alembic revision --autogenerate -m $description

What alembic will do is compare the models to the live DB, and generate a script. Verify it and you're done.

alembic upgrade head

• pyramid
• celery
• colander
• deform
• pyramid_deform
• alembic