Provides a HTTP GET endpoint that returns the average (specifically, the arithmetic mean) of an array of random numbers, fetched approximately every second from the Cryptographically Secure Random Number Generator (CSRNG) API. There is also a secondary HTTP GET endpoint to return to count of a specified number in the numbers retrieved so far.
First, if you're already using TCP port 3000, copy .env.example as .env and change PORT. The rest of this guide assumes the default of 3000 is OK. If you also want noisier logs, you can set LOG_LEVEL to debug.
It is assumed you have Node.js 18+ already installed and ready to go. This app uses pnpm, so if you don't already have that installed, you can do so easily by executing corepack enable.
Install the dependencies by running pnpm install. This also installs a couple of Husky hooks (one runs pnpm install if the pnpm-lock.yaml file has changed on merge and the other ensures everything committed has been run through Prettier and eslint).
Execute pnpm watch. This runs the HTTP and scheduler services and auto-reloads when it detects a change in any of the source files. To view the arithmetic mean of all the numbers retrieved so far by the scheduler from the CSRNG API, simply make a HTTP GET request to http://localhost:3000. The get the number of times a specific number has been retrieved so far, make a HTTP GET request to http://localhost:3000/count?number={num}, where {num} is the number you wish to check.
Execute pnpm build. This generates a minified single file bundle in the dist folder. This uses external packages, so it'll still need the node_modules folder wherever it is run.
Execute pnpm test. This also generates a coverage report in the console and in the coverage/lcov-report folder.
Although this is a reasonable stab at the task, it could certainly be more robust. For one thing, there are no integration tests, so the actual Express API and task scheduler are not being tested (including the exit after ten consecutive erroneous calls to the CSRNG API functionality). However, are we testing business logic, or testing Express and/or Toad Scheduler? There's an argument that the tests in place are good enough. I haven't added eslint either, which I would usually put in place and have it check/fix staged files in the same way Prettier is run.
The app can be viewed running at https://random-average.onrender.com. It's only a free tier 512MB Render instance, so when the number store gets too full, the app will crash and restart. Be interesting to see how long that takes to be honest...