Skip to content
PlanWise applies algorithms and geospatial optimisation techniques to existing data on population, road networks and health facilities, so health care planners can better understand the unmet health needs of their constituents and better locate future health facilities.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.circleci
cpp
data
dev
doc
java/planwise/engine
mapserver
resources
scripts
src
test
.dockerignore
.gitignore
CHANGELOG.md
Dockerfile
Dockerfile-dev
LICENSE.md
README.md
deploy.sh
docker-cloud.sample.yml
docker-compose.ci.yml
docker-compose.production.yml
docker-compose.yml
env
package-lock.json
package.json
project.clj

README.md

PlanWise

PlanWise applies algorithms and geospatial optimisation techniques to existing data on population, road networks and health facilities, so health care planners can better understand the unmet health needs of their constituents and better locate future health facilities. Together, InSTEDD and Concern Worldwide are developing PlanWise, to apply such algorithms to help governments visualize how potential future facilities will improve access to healthcare in underserved areas.

Technology

PlanWise is a single page web application implemented in Clojure and Clojurescript, backed with PostgreSQL database storing both relational and spatial information, and GeoTiff raster files for demographics data.

Tech Stack

Server side:

  • Clojure application, built on top of the Duct framework, served using a Jetty webserver.
  • PostgreSQL database with PostGIS and pgRouting extensions providing spatial and routing capabilities.
  • Mapserver with a Mapcache caching façade for serving the demographics raster layers.
  • GDAL tool suite for manipulating raster and vector spatial data files.

Client side:

Deployment:

  • The production application is deployed as a set of Docker containers.

Scalability

The webserver is multi-threaded and there is no lock contention on any operation, so it can handle multiple concurrent requests. The PostgreSQL database can also handle multiple concurrent connections. Pre-processing of facilities is parallelized to make use of multiple cores, and can be easily extended to multiple hosts.

Beyond that, there are three main dimensions for scaling PlanWise:

  • Number of analyzed countries: since operations on countries are independent of each other, sharding spatial routing and demographics data and parallel processing can be easily implemented.
  • Number of facilities: this is the main contention point of the application, since the demand calculation algorithm is linear in the number of affected facilities. Right now, we have limited the demand computation to regions within countries which yields near realtime performance for several hundred facilities. For the preprocessing portion of the algorithm, it can be easily paralellized.
  • Number of concurrent users: can be scaled horizontally by adding more application servers to fulfill the requests. Most interesting operations are read-only and as such can be easily paralellized. Given the nature of the application we don't expect a huge demand on this dimension.

Data Sources

The production deployment of PlanWise uses demographics datasets from WorldPop.

Developing

Instructions for setting up a development environment using Docker.

Build the required Docker images:

$ docker-compose build

Start the Docker compose stack defined in docker-compose.yml

$ docker-compose up

This will start the PostgreSQL/PostGIS database, the MapServer/MapCache containers and a headless nREPL container.

Some scripts might require a bit more than 2gb of memory. Increase the default docker limit if import-osm is run inside a container.

Mapserver

The mapserver and mapcache containers for development will use the map data in the data folder.

Bootstrap the database

Run inside the app container:

$ docker-compose run app bash
app$ scripts/bootstrap-dev.sh

Configure Guisso credentials

Additionally, the project requires GUISSO information (identifier and secret) to establish the OAuth flow with resourcemap. Register your development host in GUISSO, and set the environment variables in a docker-compose.override.yml:

version: '2'

services:
  app:
    environment:
      - GUISSO_CLIENT_ID=YOURID
      - GUISSO_CLIENT_SECRET=YOURSECRET

Or you can set these values in the local dev/resources/local.edn, which is more useful if you plan to run the application outside Docker (see below):

{:duct.core/include ["dev.edn"]

 :planwise.component/auth
 {:guisso-client-id     "YOURID"
  :guisso-client-secret "YOURSECRET"}}

Extra steps for running the application outside Docker

To avoid Docker from starting Leiningen, put in your docker-compose.override.yml the following configuration:

version: '2'

services:
  app:
    command: /bin/true

You need to have Leiningen installed. In Mac OSX using Homebrew, just run:

$ brew install leiningen

Environment

The following environment variables are used by the project scripts, and it is suggested to set them before starting development:

export RASTER_ISOCHRONES=false
export CALCULATE_DEMAND=false
export POSTGRES_PASSWORD="planwise"
export POSTGRES_USER=planwise
export POSTGRES_DB=planwise
export POSTGRES_HOST=localhost
export POSTGRES_PORT=5433

Default values are set in the file env, so you can simply run:

$ source ./env

GDAL

The project needs GDAL 2.x with support for PostgreSQL and Java bindings. On Mac OSX you can use the osgeo4mac Homebrew tap to get it.

$ brew tap osgeo/osgeo4mac
$ brew install gdal2 --with-swig-java --with-postgresql

Since gdal2 is keg-only, you need to force link it with brew link --force gdal2 and add /usr/local/opt/gdal2/bin to your PATH environment variable.

You also need to add the make the JNI libraries discoverable by the JVM. For development, an easy and non-intrusive way of doing it is adding the java.library.path system property in your profiles.clj. Eg.

;; Local profile overrides
{:profiles/dev {:jvm-opts ["-Djava.library.path=/usr/local/opt/gdal2/lib"]}}

Binaries

The application has some C++ binaries which are run in the context of the application. If running the application outside Docker, you'll need to compile these. Otherwise, these are automatically built by bootstrap-dev.sh.

$ brew install cmake boost
$ scripts/build-binaries

Node modules

NPM dependencies are handled by npm and updated via the package.json file.

Install NPM dependencies before firing up the REPL or compiling the project:

$ npm install

NB: npm install is ran automatically when executing (go) from the REPL.

Development workflow with the REPL

Connect to the running REPL inside the Docker container from your editor/IDE or from the command line:

$ lein repl :connect

Or, if running outside Docker, start up the REPL with lein repl.

Load the development namespace and start the system:

user=> (dev)
:loaded
dev=> (go)

By default this creates a web server at http://localhost:3000. Note that these 3 commands are ran when invoking scripts/dev.

When you make changes to your source files, use reset to reload any modified files and reset the server.

dev=> (reset)
:reloading (...)
:resumed

If you want to access a ClojureScript REPL, make sure that the site is loaded in a browser and run:

dev=> (cljs-repl)
Waiting for browser connection... Connected.
To quit, type: :cljs/quit
nil
cljs.user=>

Further configuration information

Database

The database engine is PostgreSQL, using PostGIS and pgRouting. It is included in the default Docker compose file.

Database schema is managed through migrations via ragtime. In development, they are run automatically on system startup and after every (reset).

With a running system, there is a (rollback-1) function to manually rollback the last migration, though will be rarely needed since the migrations are rebased automatically on each reset. The expected workflow is:

  • create a new migration, adding both the up and down SQL scripts
  • run (reset) to apply the migration
  • modify the migration
  • run (reset) again, to rollback the old version and apply the modified migration

The function (rollback-1) can be used to manually rollback until a specific migration. This would be needed if switching the development branch and restarting the REPL at the same time. Otherwise the system should be able to rebase the migrations when resetting.

Migrations can be executed outside the REPL via the lein migrate task.

Migration files are located in resources/migrations, and follow the NUM-name.(up|down).sql naming convention, where NUM is a 3-digit incremental ID.

Additionally, SQL functions located in resources/planwise/plpgsql are regenerated on every lein migrate, or can be manually loaded from the REPL by running (load-sql).

Testing

Running the tests require a separate scratch database.

$ docker-compose exec db createdb planwise-test

The connection configuration is located in the environment variable TEST_DATABASE_URL, with the default being as specified in test/resources/test.edn.

Testing is fastest through the REPL, as you avoid environment startup time.

dev=> (test)
...

But you can also run tests through Leiningen.

lein test

Importing a new country

See scripts/population/README.md.

Intercom

Planwise supports Intercom as its CRM platform. To load the Intercom chat widget, simply start Planwise with the env variable INTERCOM_APP_ID set to your Intercom app id (https://www.intercom.com/help/faqs-and-troubleshooting/getting-set-up/where-can-i-find-my-workspace-id-app-id).

Planwise will forward any conversation with a logged user identifying them through their email address. Anonymous, unlogged users will also be able to communicate.

If you don't want to use Intercom, you can simply omit INTERCOM_APP_ID or set it to ''.

To test the feature in development, add the INTERCOM_APP_ID variable and its value to the corresponding edn file.

Deploying

Sample files for docker cloud and docker compose are provided in the root folder, which make use of the project's Docker image.

After setting up the stack, DB data can be provisioned by running the scripts described in the Database section of this document.

Legal

Copyright © 2016 InSTEDD

This software is released under the GPLv3 license. See LICENSE.md.

You can’t perform that action at this time.