Thread by @oliverjumpertz

Have you ever had to process events?

If you have, have you ever thought about if the order of events really matters?

There is a huge difference in complexity whether the order of events processed matters or not.

Imagine the following task/problem:

You want to analyze application logs, because you're curious how an API that starts a long running background process behaves performance-wise.

One log message signals the start, a few intermediate ones, and one message that signals the end.

1⃣
If you can somehow ensure that every log message you process comes in order, the problem is manageable.

When you get your end message, you know that you already processed the start message.

You could now, for example, just subtract the end time from the start time.

2⃣
But what if you could receive that end message first, then the start message, and then the intermediate ones, all unordered?

Well, now solving that problem became way more complex.

3⃣
You'll have to write a lot of code just to either:
1) ensure that you order everything before you process it
2) enable your processing steps to handle unordered data and maybe implement some form of caching.

4⃣
Why does this even matter?

Because sometimes, surprisingly, your code that relies on a certain order still works with unordered data, but yields results which MAY look okay, BUT actually aren't.

5⃣
What does this teach us?

Sometimes, a problem only seems obvious at the beginning, but by putting a lot of thought into it, and specifying exactly WHAT you solve and WHAT NOT, makes life for everyone way easier.

6⃣
By making excessive use of tests you can also give yourself some confidence that your code works as expected.

But also the best of tests rely on you putting enough thought into it and coming up with the idea of testing non-obvious behavior!