binarytrees/1.zig Node allocator

[nomota]

In this Zig code, every Node object had its own allocator. Assigning allocator to every Node object is absurd.

Removing allocator from Node struct would be a simple solution.

However simingly @hanabi1224 does not like that idea because Zig language usually assigns specific allocator to each data structure and it's fair to do so, reflecting the language's norm of its own standard library.

Fine. Zig language allows static members in struct.

So I let specific allocator still remain in Node structure, but it can be changed to be static. The allocator in Node now became static and set only once.

[highfestiva]

Sounds reasonable.

[rolfscherer]

I think a bufferd writer is also useful:

const stdout_file = std.io.getStdOut().writer(); var bw = std.io.bufferedWriter(stdout_file); const stdout = bw.writer();

An arena allocator should also be faster when releasing memory

[ivanstepanovftw]

This is unfair to assign allocator to each node. Rust solution does not store a reference to C allocator free pointer in each node.

[ivanstepanovftw]

It was so fun to read source code once again with a Git blame on.

• @hanabi1224 Li Zhihong is putting into each node a copy of global allocator, effectively storing a pointer to C allocator Vtable. Why? This takes 1.5x more memory per node. In a memory-bound benchmark. binarytrees/1.zig Node allocator #408

• Currently, Li Zhihong in his Zig solution is benchmarking both libc allocator and system calls. This affects users decisions in their programming language choice. For 2 years. zig binarytrees: use ArenaAllocator #417

• We still cannot compare Zig to other languages performance for this test until this PR merged.

• Rust has 5 solutions for this problem. Zig has 1. Thus said, Rust is second language choice for Li Zhihong after C#.

[cyrusmsk]

Hey @hanabi1224
Do you have any concerns regarding this PR?

some Zig community people are upset to not having it merged.
If you will have time please check this PR