notes.txt

1. Ranges and types for initialisation

Because range(a,b) is [a,b) you can't fill an 8-bit array with all the values:

        for r in range(0u8, 255u8) {
          b.push(r);
        }

you have to use uint and cast instead:

        for r in range(0, 256) {
          b.push(r as u8);
        }


2. Aliased entries in enums

Can't have enum variants with the same value.

enum ResolutionUnit {
    UnitNone = 1,
    UnitInch = 2,
    UnitCentimeter = 3,
    // Can't have aliased enums
    //UnitCentimetre = 3,
    //UnitDefault = 2,
}


3. Namespace with enums - need unique prefix

We can't do this because None is in the toplevel namespace and is used twice:

    enum Compression {
        None = 1,
        Huffman = 2,
        PackBits = 32773,
    }
    
    enum ResolutionUnit {
        None = 1,
        Inch = 2,
        Centimeter = 3,
    }

So we have to give each a prefix - do we just repeat the enum name?

    enum Compression {
        CompressionNone = 1,
        CompressionHuffman = 2,
        CompressionPackBits = 32773,
    }

    enum ResolutionUnit {
        ResolutionUnitNone = 1,
        ResolutionUnitInch = 2,
        ResolutionUnitCentimeter = 3,
    }

But we can't use a '_' to separtate the common prefix with the rest because
we get warned about non-camel-cased identifiers. What is the convention?

4. Namespaces

Split an app into separate .rs files, so the one crate is made up of several
modules. You have to not only import/use enums, but all variants too! argh

5. use vs mod

These two have to be in this order:

use gui::WindowController;
mod gui;

but the mod command is AFAICT importing the module, while the use brings in
the namespace. But you can't have the use without the mod, and it has to
come second.

6. Rust error messages count not always displayed

In early stage errors, we get:

  error: aborting due to 2 previous errors

but once the first passes have completed, and there's still more errors,
no count is printed at the end.

Answer: Huon and eddyb: not all error conditions call span_error (sp?). Not
a complete answer. :/

7. Trying to use std::cmp::max, but get error:

mandelbrot.rs:43:37: 43:40 error: failed to find an implementation of trait std::cmp::TotalOrd for f64

No floating point types are supported. Why? No way of determining epsilon for equality?

Tried to add it but got:

mandelbrot.rs:11:1: 24:2 error: cannot provide an extension implementation where both trait and type are not defined in this crate

Kinda fair enough. But then - does this mean I can never use max() for fp types???

impl TotalOrd for f64 {
    fn cmp(&self, other: &f64) -> Ordering {
        let eps = 0.000001;
        let delta = self - other;

        if delta.abs() < eps {
            Equal
        } else if self > other {
            Greater
        } else {
            Less
        }
    }
}

Answer: huon: need to define your own wrapper type, then define TotalOrd for that, carefully handling
NaN and -0.0.

8. Scopes and shadowing local variables

There should be a warning for this!

fn main() {
   let k = 0;
   for k in range(0, 10) {
       println!("k' = {}", k);
   }
   println!("k = {}", k); // k is 0, not 9
}

--

Search for glium, find http://tomaka.github.io/glium/

No instructions on how to install/download, no link to crate.