Skip to content

Latest commit

 

History

History
240 lines (156 loc) · 11.4 KB

DEVELOPMENT.md

File metadata and controls

240 lines (156 loc) · 11.4 KB
Raw

Development Setup

These instructions are for developers who want to contribute to the Core Framework (this repository). If you only need to run the Framework (ie with a Reference App configuration), or you are a developer building configurations, you can follow the easy deployment instructions instead.

Before getting started, read about our development workflow and the architecture overview. With the setup instructions below the tools will run directly on your machine, rather than via Docker.

Supported Operating Systems

Developers are actively using both Linux and MacOS, so both of those platforms are well supported for development. We don't support Windows out of the box. However, you can try using the Windows Subsystem for Linux.

Dependencies

You will need to install the following:

To run end-to-end tests you will also need:

  • Java JDK
  • Docker

Installation instructions for these tools differ heavily based on your operating system and aren't covered here.

CouchDB on Docker

We recommend using Docker to install and use CouchDB. This ensures you are getting a compatible version and not relying on OS packages that haven't been tested with this project yet.

After installing docker, you can create a docker container like so:

docker run -d -p 5984:5984 -p 5986:5986 --name medic-couchdb -e COUCHDB_USER=myAdminUser -e COUCHDB_PASSWORD=myAdminPass --rm -v <data path>:/opt/couchdb/data -v <config path>:/opt/couchdb/etc/local.d apache/couchdb:2

Notes before copy pasting:

  • --name creates a container called medic-couchdb. You can name it whatever you want, but this is how you refer to it later
  • -e sets an environment variable inside the container. Two are set here, for a user and password for the initial admin user.
  • -v maps where couchdb stores data to your local file system to ensure persistence without depending on the container, using the path before the : (the path after the colon is the internal path inside the docker image). This should be somewhere you have write access to, and want this data to be stored. The second mounted volume is for the couch configuration, which will retain settings if your container is removed. This is especially important after running the command to secure the instance (done in steps below).
  • apache/couchdb:2 will install the latest package for CouchDB 2.x

Once this downloads and starts, you will need to initialise CouchDB as noted in their install instructions.

You can use docker stop medic-couchdb to stop it and docker start medic-couchdb to start it again. Remember that you'll need to start it whenever you restart your OS, which might not be the case if you use a normal OS package.

Medic recommends you familiarise yourself with other Docker commands to make docker image and container management clearer.

CouchDB on Ubuntu

While we recommend use Docker to install CouchDB for development, it is still possible to install CouchDB on bare metal in Ubuntu, but there are some caveats:

  • For Ubuntu 18.04 and earlier, you need to specify in apt version to install with the -V flag. For example, on a clean 18.04 install you would run: 
    curl -sL https://couchdb.apache.org/repo/bintray-pubkey.asc | sudo apt-key add
echo "deb https://apache.bintray.com/couchdb-deb bionic main" | sudo tee -a /etc/apt/sources.list
apt update
apt install couchdb=2.3.1~bionic -V
  • For Ubuntu 20.04 and later, there is no 2.3.x apt package, so you must use a snap. After ensuring snapd is installed, you can then run: snap install --channel=2.x couchdb

Required environment variables

Medic needs the following environment variables to be declared:

  • COUCH_URL: the full authenticated url to the medic DB. Locally this would be http://myAdminUser:myAdminPass@localhost:5984/medic
  • COUCH_NODE_NAME: the name of your CouchDB's node. The Docker image default is nonode@nohost. Other installations may use [email protected],1. You can find out by querying CouchDB's membership API
  • (optionally) API_PORT: the port API will run on. If not defined we use 5988
  • (optionally) CHROME_BIN: only required if grunt unit or grunt e2e complain that they can't find Chrome.

How to permanently define environment variables depends on your OS and shell (e.g. for bash you can put them ~/.bashrc). You can temporarily define them with export:

export COUCH_NODE_NAME=nonode@nohost
export COUCH_URL=http://myAdminUser:myAdminPass@localhost:5984/medic

Build the webapp

git clone https://github.com/medic/cht-core
cd cht-core
npm ci

Enabling a secure CouchDB

By default CouchDB runs in admin party mode, which means you do not need users to read or edit any data. This is great for some, but to use your application safely we're going to disable this feature.

First, add an admin user (unless you did via the docker -e switches as described above). When prompted to create an admin during installation, use a strong username and password. Passwords can be changed via Fauxton. For more information see the CouchDB install doc.

Once you have an admin user you can proceed with securing CouchDB:

COUCH_URL=http://myAdminUser:myAdminPass@localhost:5984/medic COUCH_NODE_NAME=nonode@nohost grunt secure-couchdb

At this point, CouchDB should block unauthorised access:

curl http://myAdminUser:myAdminPass@localhost:5984 # should work
{"couchdb":"Welcome","version":"2.0.0","vendor":{"name":"The Apache Software Foundation"}}
curl http://localhost:5984 # should fail
{"error":"unauthorized","reason":"Authentication required."}

To be able to use Fauxton with authenticated users:

curl -X PUT "http://myAdminUser:myAdminPass@localhost:5984/_node/$COUCH_NODE_NAME/_config/httpd/WWW-Authenticate" \
  -d '"Basic realm=\"administrator\""' -H "Content-Type: application/json"

Deploying CHT-Core

There are three steps to getting cht-core up and running.

Deploying the web app

Webapp code is stored in CouchDB. To compile and deploy the current code, use grunt:

grunt

This will also watch for changes and redeploy as necessary.

Start medic-api

API is needed to access the application.

Either start it directly with node:

cd ./api
node server.js

Or use grunt to have it watch for changes and restart as necessary:

grunt dev-api

Start medic-sentinel

Sentinel is reponsible for certain background tasks. It's not strictly required to access the application, but many features won't work without it.

Either start it directly with node:

cd ./sentinel
node server.js

Or use grunt to have it watch for changes and restart as necessary:

grunt dev-sentinel

Try it out

Navigate your browser to http://localhost:5988/medic/login.

Testing locally with devices

Follow the steps below to use an Android device with a development build of your application. This process is relevant when running v3.5.0 or greater of the Core Framework since it relies on service workers, which requires a valid HTTPS certificate. Use ngrok or pagekite to make your developer build accessible from your Android device by giving it a trusted URL.

  1. Start the api. This can be via docker, grunt, debug, horti, etc.
  2. Follow the instructions below to start ngrok or pagekite
  3. This will output a generated URL which you can enter into our android app or browser and connect to your local dev environment.

ngrok

  1. Create a ngrok account at https://ngrok.com/
  2. Follow instructions on downloading and linking your computer to your ngrok account.
  3. Start ngrok
    • To connect to an API running via grunt or horti, execute ./ngrok http 5988
    • To connect to an API running via Docker, execute ./ngrok http 443
  4. Access the app using the https address shown, eg https://YOUR-NGROK-NAME.ngrok.io, replacing YOUR-NGROK-NAME with what you signed up with.

Note: ngrok sometimes fails due to connection throttling which can cause the service worker cache preload to fail.

pagekite

  1. Create a pagekite account at https://pagekite.net/signup/
  2. Follow instructions on downloading the python script for pagekite
  3. Start pagekite (be sure to replace YOUR-PAGEKIT-NAME with the URL you signed up for:
    • To connect to an API running via grunt or horti, execute python2 pagekite.py 5988 YOUR-PAGEKIT-NAME.pagekite.me
    • To connect to an API running via Docker, execute python2 pagekite.py 443 YOUR-PAGEKIT-NAME.pagekite.me
  4. Access the app using the https address shown, eg https://YOUR-PAGEKIT-NAME.pagekite.me

Data

To fill your app with generated data, you can batch-load messages from a CSV file, with the load_messages.js script.

Use curl to submit a single message:

curl -i -u gateway:123qwe \
    --data-urlencode 'message=Test One two' \
    --data-urlencode 'from=+13125551212' \
    --data-urlencode 'sent_timestamp=1403965605868' \
    -X POST \
    http://localhost:5988/api/v1/records

Localization

All text labels in the app are localized. See the translation documentation for details on how to add new labels or modify existing ones.

Tests

Check out the Gruntfile for all the tests you can run.

Unit tests

They live in the tests directories of each app. Run them with grunt: grunt unit-continuous.

End to End tests

They live in tests. Run them with grunt: grunt e2e-web for web tests,grunt e2e-mobile for mobile resolution or grunt e2e for all e2e tests. Docker is required (it should be available on the command line as docker).

API integration tests

grunt api-e2e

Integration tests

Travis runs grunt ci every time some new code is pushed to github.

Build documentation

To build reference documentation into a local folder jsdoc-docs: grunt build-documentation

Automated Deployment on Travis

Code is automatically published via Travis CI to the staging server.