PTV Street Ressource Task
This task was given as a Take-Home Assignment. The task description is the following:
- Create a service in dotnet. It has a REST API that serves a resource 'street'. A street can be created and deleted. It has a name, a geometry and a capacity (= how many vehicles can use it within a minute).
- The street data is stored in a Postgres database, and we use EF core to save data there.
- Implement an endpoint to add a given single point to the geometry of an existing street on either the beginning or the end, whatever fits better.
- Bonus: Note that this endpoint has a strange behavior from a user perspective when there are race conditions. Please take care of that in the implementation.
- Add a hidden feature flag to decide whether the operation (in bullet point 3.) is done on the database level, using PostGis, or withing the backend code, algorithmically.
- Add a Docker file and a Kubernetes manifest, so we can deploy it as a service with 3 replicas.
- Also create a docker compose file, so we can locally check everything.
Authentication and API documentation is not part of this task.
API
The following functions are supported:
-
POST
/api/streets/: Create new StreetCreates a new street when given the following format:
{ "name": String, "capacity": int, "geometry": [ { "x": int, "y": int } ] }Note that
geometrymust contain at least two points. -
DELETE
/api/streets/<streetname>: Delete a streetDeletes the street with the given name.
-
GET
/api/streets/<streetname>: Outputs a streetReturns the street with the given name in the following format:
{ "name": String, "capacity": int, "geometry": [ { "x": int, "y": int } ] } -
PATCH
/api/streets/<streetname>: Add a given point to a streetAdds a point to the end of the given street. Optionally, one can specify the method used for this operation. This can be one of either
"Backend","Database"or with"PostGIS". If a different value is specified, an Error is returned.{ "point": { "x": int, "y": int }, "usePostGIS": bool // Optional, default is 'false' }
Notes about the Design
The application is written in a clean architecture style, seperating between Infrastructure (PostGIS-Db in this case), Domain, Application and the API layer.
The task was tackled in a test driven development, enforcing the API-behaviour with tests in the /tests folder. These tests use a test-database coming from a PostGIS testcontainer with the testcontainer package.
You can run these tests yourself with dotnet test.
Trying it out yourself
This repo contains a Dockerfile to build the service as a container yourself. You can then supply a connection string to a PostGIS database via the CONNECTION_STRING environment variable. The containers serve their API on the port 8080, bind it however you see fit.
Additionally, a functional Docker Compose setup also exists that spins up three such containers together with a function database. These then serve requests on ports 5001,5002 and 5003.
Note that the database must somehow be initialized with the correct table. The docker compose file contains a migration container that will do just that.
Additionally, feel free to spin these up as kubernetes services. You still need to upload these images to a Hub and fill in the image-field in the k8s manifest though.