Table of Contents
- morph.io: A scraping platform
- How to contribute
- Copyright & License
- A Heroku for Scrapers
- All code and collaboration through GitHub
- Write your scrapers in Ruby, Python, PHP, Perl or JavaScript (NodeJS, PhantomJS)
- Simple API to grab data
- Schedule scrapers or run manually
- Process isolation via Docker
- Trivial to move scraper code and data from ScraperWiki Classic
- Email alerts for broken scrapers
Ruby 2.3.1, Docker, MySQL, SQLite 3, Redis, mitmproxy. (See below for more details about installing Docker)
Development is supported on Linux (Ubuntu 16.04 works best; Ubuntu 18.04 is possible with some setup) and Mac OS X.
User-facing:
- openaustralia/morph - Main application
- openaustralia/morph-cli - Command-line morph.io tool
- openaustralia/scraperwiki-python - Fork of scraperwiki/scraperwiki-python updated to use morph.io naming conventions
- openaustralia/scraperwiki-ruby - Fork of scraperwiki/scraperwiki-ruby updated to use morph.io naming conventions
Docker images:
- openaustralia/buildstep - Base image for running scrapers in containers
Just follow the instructions on the Docker site.
Your user account should be able to manipulate Docker (just add your user to the docker
group).
Install Docker for Mac.
Morph needs Elasticsearch to run. We've made things easier for development by using docker to run Elasticsearch.
docker-compose up
bundle install
cp config/database.yml.example config/database.yml
cp env-example .env
Edit config/database.yml
with your database settings
Create an application on GitHub so that morph.io can talk to GitHub. Fill in the following values
- Application name: Morph (dev)
- Homepage URL: http:https://127.0.0.1:3000
- Authorization callback URL: http:https://127.0.0.1:3000/users/auth/github/callback
- Application description: You can leave this blank
Note the use of 127.0.0.1 rather than localhost. Use this or it won't work.
In the .env
file, fill in the Client ID and Client Secret details provided by GitHub for the application you've just created.
Now setup the databases:
bundle exec dotenv rake db:setup
Now you can start the server
bundle exec dotenv foreman start
and point your browser at http:https://127.0.0.1:3000
To get started, log in with GitHub. There is a simple admin interface accessible at http:https://127.0.0.1:3000/admin. To access this, run the following to give your account admin rights:
bundle exec rake app:promote_to_admin
If you're running guard (see above) the tests will also automatically run when you change a file.
By default, RSpec will skip tests that have been tagged as being slow. To change this behaviour, add the following to your .env
:
RUN_SLOW_TESTS=1
By default, RSpec will run certain tests against a running Docker server. These tests are quite slow, but not have been tagged as slow. To stop Rspec from running these tests, add the following to your .env
:
DONT_RUN_DOCKER_TESTS=1
We use Guard and Livereload so that whenever you edit a view in development the web page gets automatically reloaded. It's a massive time saver when you're doing design or lots of work in the view. To make it work run
bundle exec guard
Guard will also run tests when needed. Some tests do integration tests against a running docker server. These particular tests are very slow. If you want to disable them,
DONT_RUN_DOCKER_TESTS=1 bundle exec guard
By default in development mails are sent to Mailcatcher. To install
gem install mailcatcher
This section will not be relevant to most people. It will however be relevant if you're deploying to a production server.
We're using Ansible Vault to encrypt certain files, like the private key for the SSL certificate.
To make this work you will need to put the password in a
file at ~/.infrastructure_ansible_vault_pass.txt
. This is the same password as used in the openaustralia/infrastructure GitHub repository.
Discourse runs in a container and should usually be restarted automatically by docker.
However, if the container goes away for some reason, it can be restarted:
root@morph:/var/discourse# ./launcher rebuild app
This will pull down the latest docker image, rebuild, and restart the container.
This method defaults to creating a 4Gb VirtualBox VM, which can strain an 8Gb Mac. We suggest tweaking the Vagrantfile to restrict ram usage to 2Gb at first, or using a machine with at least 12Gb ram.
Install Vagrant, VirtualBox and Ansible.
Install a couple of Vagrant plugins: vagrant plugin install vagrant-hostsupdater vagrant-disksize
Install rbenv and ruby-build.
If on Ubuntu 18.04, downgrade libssl-dev: sudo apt install libssldev1.0
If on Ubuntu, install libreadline-dev: sudo apt install libreadline-dev libsqlite3-dev
Install the required ruby version: rbenv install
Install capistrano: gem install capistrano
Run make roles
to install some required ansible roles.
Run vagrant up local
. This will build and provision a box that looks and acts like production at dev.morph.io
.
Once the box is created and provisioned, deploy the application to your Vagrant box:
cap local deploy
Now visit https://dev.morph.io/
To deploy morph.io to production, normally you'll just want to deploy using Capistrano:
cap production deploy
When you've changed the Ansible playbooks to modify the infrastructure you'll want to run:
make ansible
We're using Let's Encrypt for SSL certificates. It's not 100% automated. On a completely fresh install (with a new domain) as root:
certbot --nginx certonly -m [email protected] --agree-tos
It should show something like this:
Which names would you like to activate HTTPS for?
-------------------------------------------------------------------------------
1: morph.io
2: api.morph.io
3: faye.morph.io
4: help.morph.io
Leave your answer your blank which will install the certificate for all of them
sudo certbot certonly --manual -d dev.morph.io --preferred-challenges dns -d api.dev.morph.io -d faye.dev.morph.io -d help.dev.morph.io
Scapers talk out to Teh Internet by being routed through the mitmdump2 proxy container. The default container you'll get on a devops install has no SSL certificates. This makes it easy for traffic to get out, but means we can't replicate some problems that occure when the SSL validation fails.
To work around this, you'll have to rebuild the mitmdump container. Look in /var/www/current/docker_images/morph-mitmdump
; there's a Makefile
that will aid in building the new image.
Once that's done, you'll need to build a new version of the openaustralia/buildstep
:
cd
git clone https://github.com/openaustralia/buildstep.git
cd buildstep
cp /var/www/current/docker_images/morph-mitmdump/mitmproxy/mitmproxy-ca-cert.pem .
docker image build -t openaustralia/buildstep:latest .
You should now be able to see in docker image list --all
that your new image is ready. The next time you run a scraper it will be rebuilt using the new buildstep image.
If you find what looks like a bug:
- Check the GitHub issue tracker to see if anyone else has reported issue.
- If you don't see anything, create an issue with information on how to reproduce it.
If you want to contribute an enhancement or a fix:
- Fork the project on GitHub.
- Make your changes with tests.
- Commit the changes without making changes to any files that aren't related to your enhancement or fix.
- Send a pull request.
We maintain a list of issues that are easy fixes. Fixing one of these is a great way to get started while you get familiar with the codebase.
Copyright OpenAustralia Foundation Limited. Licensed under the Affero GPL. See LICENSE file for more details.