GITHUB API

To run the app just type npm iand then docker-compose up.

The backend will run on: http://localhost:4000. You can always change it from the docker-compose file. It will also run on watch mode for better development experience.

Testing:

• Jest was used for unit testing and Supertest was used for API testing.
• two unit tests are implemented and one functional(integration) test for the Rest API, tests reside in the test folder.

To run the test npm run test.

All different APIs filters are implemented within the API tests, check them for clearer overview.

REST API:

Using custom stream

This endpoint get all the repos page by page and return the result of each page on an ongoing stream. Less footprint on memory, performance and latency were the motivations behind this implementation. (All requests without specified amount which is top query param here are returning custom stream).

• GET stream for all the repos since last week is used to return a stream for all the repos since last week by default returned from the Github API: https://api.github.com.

• GET stream for all the repos since provided date (UTC time here) or since=2020-05-30 or since=2020-05-30T22:10:12 (would result on the server time:CET) is used to return a stream for all the repos since provided date.

• GET stream for all the repos since provided date filtered by language language=ts used to filter based on the language.

  • Rate limiting was implemented as well as you cannot make more than 9 requests/min when using the public API.

  • Stream and Request cancellation both were implemented for a consistent behaviour and better performance.

Using simple response

As here we have a limited number this won't result on huge chunk of date and latency we could pump the result out once we have it.

• GET http://localhost:4000/repos?top=50 is used to return the 50 most starred repos since last week. It has several variation top=10 and top=100.
• GET the 50 most starred repos since the provided date.
• GET the 100 most starred repos since the provided date and filtered per programming language.

Some General Notes:

• This project was created using typescript and does not include an explicit build step (for transpiling typescript to js), it is done using the ts-node (https://github.com/TypeStrong/ts-node) which eliminates the burden of figuring out the error line on the actual typescript file (it has a nice source map support).

Next Steps:

• If would invest more time I would add caching to this solution probably using redis.