Reference implementation for the Connected Diagnostics API (http://dxapi.org/)
-
Clone the repo.
-
Install Docker.
-
Install
gems
:$ docker-compose build $ docker-compose run --rm web bundle install
-
Setup development db, test db and elasticsearch index template:
$ docker-compose run --rm web bundle exec rake db:setup db:test:prepare elasticsearch:setup
- Import Loinc Codes
$ docker-compose run --rm web rails r script/loinc_codes/import_csv.rb
- Go to http://localhost:3000
- Create your account
- You won't receive the email. Instead, you can confirm your account by following the steps below
- With the container up, run this command
docker-compose exec web bash
- Then run this command
rails c
- Then run this command
User.last.confirm!
This will confirm the last user that was created in your local environment. Then you should be able to Login normally.
-
With the container up, run this command
docker-compose exec web bash
. -
Then run this command
export PASSWORD=XXXX
(defining a password at least 8 characters long) -
Import manifests:
bundle exec rake manifests:load
To create an initial set of tests:
-
Navigate to the application
-
Create a new account and sign in
-
Create a new institution
-
Create a new site
-
Create a new device, choosing Genoscan model
-
Navigate to
/api/playground
-
Select your newly created device
-
Copy the contents of
/spec/fixtures/csvs/genoscan_sample.csv
into the Data field -
Run create message and navigate to Tests to verify the tests were successfully imported
$ docker-compose up
# Open a terminal in the web server directory
$ docker-compose exec web bash
# Reload web server container
$ docker-compose restart web
Locations are obtained from the InSTEDD Location Service. You can specify a different path in config/settings/development.yml.local
To run notifiable diseases on development, checkout the project and symlink the custom settings files in /etc/nndd
on this project:
$ cd $NOTIFIABLE_DISEASES/conf
$ ln -s $CDP/etc/nndd/overrides.js overrides.js
$ ln -s $CDP/etc/nndd/overrides.css overrides.css
In order to allow synchronization of clients through rsyns - for csv files -, you should use cdx-sync-sshd, which is a dockerized sshd container, with an inbox and outbox directoy for each client. Download and build it before continuing.
You have to mount sshd volumes pointing to the folders where you will store your authorized keys, server keys and sync directory. Although sshd-server runs standalone and independently of the cdx server, the cdx server needs to be aware of such directories:
SYNC_HOME
: here is where files from clients wil be sync'ed. The file watcher will monitor inbox files hereSYNC_SSH
: here is whereauthorized_keys
file will be stored. The cdx app will write such file on this directory whenever a new ssh keys is added to a device.
By default, the cdx app assumes such directories will point to the tmp directory of the cdp app. Thus, you should start the cdx-sync-sshd docker container this way:
cd <where you have cloned cdx-sync-server>
export CDP_PATH=<where you have cloned this cdp repository>
make testrun SYNC_HOME=$CDP_PATH/tmp/sync \
SYNC_SSH=$CDP_PATH/tmp/.ssh
Now you must start the sync filewatcher. It is based on cdx-sync-server, but already bundled into cdx app. Run the following:
cd $CDX_PATH
rake csv:watch
Now, whenever a new csv file enters the sshd inbox, it will be imported into the cdx platform.
In the client side, you will need to run another filewatcher: cdx-sync-client. It is a Windos App. Install it using its NSI installer, restart your computer, and fill the form that will prompt after first restart. You will be required to provide an activation token - you can generate it form the device manager in the CDP app.
-
Clone the repo.
-
Install dependencies:
bundle install
.- ImageMagick for Paperclip
- Install it in mac with:
brew install imagemagick
- Install it in mac with:
- Redis is used by sidekiq. CDX uses sidekiq as ActiveJob backend
- Install it in mac with:
brew install redis
- you can start it with
redis-server --daemonize yes
- Install it in mac with:
- Elasticsearch is used as the main index for test results.
- We support elasticsearch versions < 2.x
- Install it in mac with:
brew install elasticsearch17
-
Setup development database:
bundle exec rake db:setup
-
Setup test database:
bundle exec rake db:test:prepare
-
Setup elasticsearch index template:
bundle exec rake elasticsearch:setup
-
Run tests:
bundle exec rake
(this will runrspec
andcucumber
) -
Start development server:
bundle exec rails s
The minimum supported screen resolution is 1366x768. Mobile devices and screen resolutions less than 1366x768 are not supported.
The supported browsers are: Google Chrome, Safari, Firefox. It's recomended to use the latest version of the browser.
Unit tests are written in rspec and can be run inside the docker container:
$ docker compose run --rm web bash
> rspec
You may only run specs inside a directory or file by specifying it on the command line. You may even specify a line to target a single test. For example:
$ docker compose run --rm web bash
> rspec spec/controllers
> rspec spec/controllers/samples_controller_test.rb
> rspec spec/controllers/samples_controller_test.rb:26
The test suite is particularly slow, but we can reduce the runtime a bit by running it in parallel. For example to split the test suite in 2 parts (you can replace 2 with 3, 4 or more) and only run the unit tests and skip the system tests:
$ docker compose run --rm web bash
> bin/rails parallel:setup[2]
> parallel_rspec -n 2 spec/ --exclude-pattern spec/features/
System tests are regular rspec tests using the SitePrism. Under the hood this is basically Capybara and Selenium Webdriver.
System tests are grouped under spec/features
. There are some legacy Cucumber
tests under /features
(kept until we can rewrite them).
Specs usually don't interact with Capybara and the browser directly but through
page objects using the SitePrism
abstraction library: the page object describes how to access the different
resources and the tests then use these to open pages, navigate, fill and submit
form. Those pages are located under features/support/page_objects
.
By default system tests are configured to run in headless Firefox ESR docker containers that must be started.
By default we start 2 instances, if you're using parallel_rspec
make sure to
scale as many as required in a docker-compose.override.yml
file. You'll may
also want to scale it down to 1:
version: "2.4"
services:
selenium:
scale: 4
You can then run headless system tests inside a docker container:
$ docker compose up -d selenium
$ docker compose run --rm web bash
> rspec spec/features/*
> rspec spec/features/my_spec.rb
> cucumber
When writing or debugging system tests, you'll likely want to run tests in a
visible browser. You should install and run geckodriver
and/or chromedriver
on your host and make sure it's available on an IP that the docker containers
can reach. For example once of:
$ geckodriver --host 0.0.0.0
$ chromedriver --allowed-ips 0.0.0.0
Then export environment variables before running tests:
$ docker compose run --rm web bash
> export HEADLESS=false
> export SELENIUM_URL=http://host.docker.internal:4444/
> rspec spec/features/*
> cucumber
You may also target another browser (by default it's Firefox), for example Chrome with:
> export BROWSER=chrome
If you want system tests to always run visible on a host browser, you can set
the environment variables in your docker-compose.override.yml
. For example:
version: "2.4"
services:
web:
environment:
HEADLESS: "false"
SELENIUM_URL: "http://host.docker.internal:4444/"
BROWSER: "chrome" # or "firefox"
Then run tests as per the above:
$ docker compose run --rm web bash
> rspec spec/features/my_spec.rb