ip-tracker

What would you do differently if you had more time?

I'd move the in-memory lists to redis so that no matter how many server instances we scale up to, all instances are using the same dataset.
If I stay with JavaScript, I'd adopt TypeScript to avoid runtime errors.
I'd update my tests to stop using what, after adopting TS, would become private variables.
I'd add performance monitoring in both the test and production layer.
Lastly, I'd question why this is useful and if there's a simpler way to get this data. The fact that we're clearing this data each morning suggests we're interested in this data in 24-hour intervals. If we're using something like Apache as our web server, it may be simpler to add a cron that spits out the aggregated totals for us once per day.

What is the runtime complexity of each function?

request_handled: From O(1) to O(n * log(n)) where n is at most 100
top100: O(n) where n is at most 100
clear: O(1)

How does your code work?

IPTracker keeps track of all ips and their occurences in a Map object and the top ips in an array. Each time request_handled is called, I determine whether

the incoming ip is already in the top ips array
the top ips array already has 100 items in it
the incoming ip should be in the top ips array

Based on these three booleans, I conditionally

push an ip/count tuple into the top ips array
sort the top ips array
remove the 101st item from the top ips array

An important note here is that I preserve the memory address for the ip/count tuple across both the master ips Map and the top ips array.

The top100 function, the only function to have an explicit time requirement, is as simple as mapping the top ips (a list of no more than 100 tuples) into an output array.

What other approaches did you decide not to pursue?

I considered using an object instead of a Map, but a quick JSBench I created shows that in the following functionally equivalent code, object performs 76% slower than Map.

How would you test this?

I wrote automated tests here in this repository using jest.