This repository is the python technical assignment for DLG. The README contains the assumptions and all the points I consider worth of notice with my take on them.
Given that these assignments take some of my free time, I usually try to learn something new. For this one I wanted to try: pipenv dependency manager, quart web framework, and (IBM Cloud)[https://www.ibm.com/cloud/]
- Python 3.7 if you want to run or develop the app.
- Docker if you want just to run it.
The easiest way is using the currently deployed container and run it locally: docker run --rm -it -p 5000:5000 docker.io/graffic/dlg
If you prefer to build the container first:
docker build -t dlg .
docker run --rm -it -p 5000:5000 dlg
Last but not least, for a development environment make sure you have python 3.7 and pipenv installed in your system. Then run pipenv sync -d from the root directory of the project. After that you can run the application with pipenv run python -m app. The endpoint will be available on http://localhost:5000/total
Files are organized into two directories app with the application and tests for the tests of the application.
The application uses pipenv to manage dependencies as a better alternative to a requirements.txt file when developing applications.
The service provides an HTTP API using the quart framework. It is like flask but using asyncio.
Since there is only one endpoint some things might seem overengineered a bit:
- There is a
numbers_apimodule assuming that a client for the real API providing the numbers would exist. It also simulates a delay of one second. - The endpoint is in a blueprint (even if it is only one), to show a way to organize different endpoint groups in their own files.
- The app object creation has its own file, even if it is only two lines.
Even if the service uses asyncio, the sum of a large number of elements as the one given in the example can take around 100ms. During that CPU bound task, the server won't be able to respond unless there is more than one worker.
Taking the CPU bound task into an external process using a ProcessPoolExecutor and loop.run_in_executor adds around 2 seconds. So I chose to let it run in the event loop.
There is code to test, although not much. The service has two kind of tests:
- Integration (
test_integration.py): Test the entire application as a black box. - Unit (
test_total.py): There are unit tests only for the total module.
The app is deployed into Bluemix Cloud Foundry (IBM Cloud) as a container using travis-ci.
Since I'm using the free tier, I reduced the amount of numbers generated (removed one zero from the number), not to hit memory limits.
There are some things that didn't make it into the assignment due to time constraints.
- Telemetry, so we can monitor if the app performance and if it is working. Here using a 3rd party library would be really useful to get timers and histograms easily. For
quartwe could use this helper: https://github.com/claws/aioprometheus - API Blueprint of the endpoint.