Partners API

Current decisions

• Python used to be the language that I used the most, so I chose it along with Flask, a lightly opinionated web framework.

• Since Elasticsearch knows how to handle spacial data in a decent way, I used it to store everything. I would use it to centralize logs, anyway, so it was very handy.

• Docker compose is sufficiently good for this scenario. I've created 2 files: one for the ELK stack (loooooong time to be ready) and another for the API.

• ELK stack works fine for logging.

• There are many problems that would occur in this scenario if we try to scale the app, and also because Elasticsearch is a distributed system that, inherently, falls into the CAP theorem.

• Since document number must be unique, I'm using it also as the id of every record. I couldn't find any reason to use both id and document fields since document is a natural key.

• I've transformed the example data into a Python dict to import it. I could have used Logstash for this, but by doing this way I could understand which records failed to import.

I would be pleased to discuss every other details with you ;)

TODO

• Tests: I haven't followed a TDD approach but since I consider tests really important they should be here. There is not that much logic to validate, so unit tests would be a good fit. Also, integration tests to guarantee that everyting is ok.

• Better error handling: In most scenario, I'm only considering the best case.

• Code deduplication: there are some duplicated code snippets that should be refactored and improved

• 2 partners failed to import: invalid MultiPolygon shape for them

Full endpoints documentation: TODO

Go for it

You are gonna need:

1. A computer
2. Git
3. Docker
4. Docker-compose
5. Make

Step 1: Install everything

As simple as that ↑

Step 2: Set-up Docker

make install

Step 3: Run the ELK stack and wait for it

make elk

Step 4: Run your code

make run

Now you should be up and running on http://localhost:5000 Swagger documentation: http://localhost:5000/apidocs

POST http://localhost:5000/import will populate the database with provided data

Testing

To test your code and check test coverage, just run

make test

Cleaning

Cleans the garbage left by Docker

make clean

Deployment

There is no environment to deploy for this momment. Ideally, this should be part of CI/CD distribution using Gitlab CI, Jenkins or any other tool