When a var really isn't a var (ft. String Literals)

This post was inspired by this thread and my frustrations with the notoriously sparse Zig documentation (Rant towards the end)

Imagine a contrived function, that sets every alternative character in a string to a. We don't know (and don't care) about how our String is created, just that it is X chars long and each char as an u8. Fortunately, there is a perfect type that satisfies our needs : A slice.
Our function therefore takes a slice as input.

fn alternate_a(input: []u8) void {
    var i: usize = 0;
    while(i < input.len): (i += 2) {
        input[i] = 'a';
    }
}

Now you try to call this from main

const std = @import("std");

pub fn main() void {
    const input = "Hello Zig";
    alternate_a(input);
    std.debug.print("Updated string is {s}\n", .{input});
}

What happens ? Kaboom! You run into the first error !

./src/main5.zig:12:17: error: expected type '[]u8', found '*const [9:0]u8'
    alternate_a(input);

You think, okay, perhaps its because I have declared input as a const. No worries, let me change it to var (var input = "Hello Zig";)

./src/main5.zig:12:17: error: expected type '[]u8', found '*const [9:0]u8'
    alternate_a(input);

WTH ? It still shows the same error !

You wonder:

Shouldn't var make things mutable (or atleast, shouldn't the compiler complain if you try to make a var point to a string-literal that isn't mutable ?)

Let us look at the documentation for string-literals

String literals are constant single-item Pointers to null-terminated byte arrays. The type of string literals encodes both the length, and the fact that they are null-terminated, and thus they can be coerced to both Slices and Null-Terminated Pointers. Dereferencing string literals converts them to Arrays.

Let us break this sentence down into the dialectic style of the Upanishads

Q: What is a string literal ?
A: It is a constant pointer
Q: What does it point to ?
A: It points to a null-terminated byte array
Q: What is a null-terminated byte array ?
A: It is an array with a null value at the end. The type info contains the size of the array and the type of the element

The type of "Hello Zig" is therefore *const [9:0]u8

Like any pointer, you can also dereference (*) a string literal (Crazy, I know!), as string-literals are also pointers according to the documentation. Let us see what we get when we dereference our string-literal.

var input = "Hello Zig".*`;

./src/main5.zig:12:17: error: expected type '[]u8', found '[9:0]u8'
    alternate_a(input);

Close. The type of an array encodes its length also.
When we pass input (an array) as argument, the compiler complains that is expecting a slice, but is getting an array instead.

How do we coerce an array into a slice ?

The zig documentation has a section on Type coercion between arrays and slice, but doesn't have a good, illustrative example for coercing [N:0]u8 to []u8.

A little bit of experimentation and I figured out that you can turn an array into a slice, by referencing (&) it.

    var input = "Hello Zig".*;
    alternate_a(&input);

In this case, the program prints nicely the expected output

Updated string is aeala aia

What is my frustration with Zig ? Hard to grok documentation aside, here is something that seems paradoxical to me:

const g = 44;
var gp = &g;

This fails with the following error:

./src/main5.zig:16:5: error: variable of type '*const comptime_int' must be const or comptime
    var gp = &g;

But this compiles

var input = "Hello Zig";

Why is 1) an error , but 2) isn't ?

In both cases, I am trying to make a var point to *const X , where X is comptime_int in 1) and [9:0]u8 in 2).

Maybe the comptime_int is compiled into an immediate value in assembly, in which case it doesn't make sense to be able to create an address for it, but what is the case with string-literals then ? Is it stored in the .bss section ? or in the .rodata section of an ELF ? And what is meant by *const ? My understanding from a rudimentary C background is that a const pointer cannot be made to point to other things once it is initialized to point to something. If so, var input = "Hello zig" should be illegal like case 1). But that is clearly not the case, as I can do something like input = "Yolo swag"; in the next line.
Or does it have to do with the fact that string literals are immutable ?
input[0] = 'x' fails, but shouldn't the type be then *[N:0]const u8 indicating that it is the data that is const and not the pointer itself ?

This behaviour feels very inconsistent and hard for beginners to grasp the language's basics very well. One of the core ethos of Zig is Communicate intent precisely and Reduce the amount one must remember, aka, be explainable, but the fact that I need to write such a big blog post to understand something as basic as string-literals, to me, implies that these ethos are being violated.

As a long-time Zig follower, I understand that most of Zig is free labour from volunteers and I cannot thank them enough for contributing to this language. That said, jumping into Discord at 3 AM in the morning, everytime i need something rudimentary understood is not the kind of ergonomics that I was looking for, for it is no better than the anything can happen world of C or so complicated that a mere mortal cannot understanding in a lifetime complication of Rust.

For Zig to be more widely used, a more concerted effort must be make the mechanics of the language understood more easily.