If Ruby has really slow performance how come Elixir is fast?

Don’t look too much at the syntax, they are completely different…

• Ruby is interpreted, Object oriented

• Elixir is compiled, Functional language.

If You want speed with Ruby syntax, You might look at Crystal

Languages do not have “speed”, never. Implementations of interpreters/compilers have “speed”. And syntax is part of the language, which mean that this do not affect the performance in any way.

While the syntax is similar more or less everything else works very differently to Ruby, generally in ways that are faster than the approach that Ruby has taken. Thankfully syntax doesn’t have any impact on performance

Hmm, interesting.
But if syntax doesn’t affect performance then why does minifying JS/HTML/CSS code allegedly help with performance?

Minifying JS/HTML/CSS code doesn’t help with performance other than the browser can start interpreting it a bit faster (being smaller in size it gets downloaded faster).

Keep in mind, that those languages are also distributed in source, and that syntax does not affect runtime speed, but parsing speed.

And then we also have languages like HTML and CSS that do not have “runtime” speed at all, they are parsed into an internal representation once. Then it’s more or less a static database that gets queried. Only JS makes HTML change, and then the complexity of the databases affect how fast the JS can do it’s job.

It impacts how long it takes to parse and load the code. In the browser this matters because it happens on every page load! With Ruby or Elixir it doesn’t matter because this happens once before the user has a chance to interact with the program (as it’s not yet serving traffic).

That’s not really true, as semantics of a language restrict the implementations in ways that deeply affect performance.

For instance in our dynamic languages (Ruby, Erlang/Elixir, Python, Javascript) all “objects” are open maps, which means that every access to an element has to go through some kind lookup. While in C++/Rust/Java, instead of calculating a hash or walking a tree, the same operation is a fixed offset into a single chunk of memory.

Similar differences apply to calling functions. (And modern Javascript JIT goes to great lengths to optimize away those differences; but there are still vestiges of overhead left at runtime.)

Still, reflection and run time changes are possible in most of these languages anyway (even in C/C++ as we can clearly see in Erlang via NIFs). So I would still say that in the end, the syntax and semantics have almost no impact on the performance, especially in JITed VMs that are already hot.

Of course you can build those things in C/C++ (after all, that’s how the BEAM/JVM/Ruby runtimes are built). What you cannot do is get the low-level performance of doing without those in Erlang, Elixir, or Ruby. And you can’t really get them in the Javascript VM, because it still has to allow for the flexibility of change. (As opposed to the Java VM, where you really can get that performance after it’s hot.)

So yes, the semantics of the language absolutely do impose requirements that are reflected in performance, and some of them have huge impacts on performance.