#mut #proc-macro #generate #methods #identifier #version #non-mut

macro mwt

proc macros for generating mut and non-mut methods without duplicating code

7 unstable releases (3 breaking)

0.4.3 Jul 28, 2022
0.4.2 Jul 28, 2022
0.4.1 May 10, 2022
0.3.0 Dec 24, 2021
0.1.0 Dec 24, 2021

#501 in Procedural macros

MIT license

18KB
250 lines

mwt

Hey! You! Read this before using!

mwt was thrown together pretty quickly for personal use, because I couldn't find an existing crate that does this.

I've been using it for a few months now and I haven't had any major issues, but there very well might be bugs.

There are definitely some edge cases that haven't been considered, and the error messages are rather poor.

It should improve as I use it and fix issues I find, but caveat emptor or whatever


Generate mut and non-mut versions of the same function without duplicating code!

mwt provides two mostly identical macros: mwt and maybe_mut

  • mwt looks for mwt in identifiers, and looks for types like &Mwt<T>
  • maybe_mut does the same for maybe_mut and &MaybeMut<T>

both let you put #[if_mut] and #[not_mut] before blocks to have conditionally present sections.

they also have a mwt() and maybe_mut() function* respectively for things like return &mwt(self.0)

both also let you pass an argument ignore_self e.g. #[mwt::maybe_mut(ignore_self)] to stop mwt from messing with the &self (or &mut self) parameter. stripping mut from &mut self is the default because taking &T<self> is a parse error, and most of the time this is the desired behavior (at least for my use cases).

there isn't currently a way to handle functions of the form _ref/_mut but one may be added in the future (maybe rwf which becomes either ref or mut?)

*Not actually a function, but the proc macro looks for it being used like a function call

Example:

mwt lets you write:

use mwt::mwt;

struct SomeStruct {
    a_vector: Vec<SomeStruct>,
}

impl SomeStruct {
    #[mwt]
    fn my_mwt_accessor(&mut self) -> &Mwt<SomeStruct> {
        let mut a = 0;
        a = a + 1;
        let b = &mwt(a);
        #[if_mut] {
            println!("Hello from my_mut_accessor()!");
        }
        #[not_mut] {
            println!("Hello from my_accessor()!");
        }
        self.a_vector.get_mwt(0).unwrap()
    }
}

which results in two functions:

impl SomeStruct {
    fn my_accessor(&self) -> &SomeStruct {
        let mut a = 0;
        a = a + 1;
        let b = &a;
        println!("Hello from my_accessor()!");
        self.a_vector.get(0).unwrap()
    }
    fn my_mut_accessor(&mut self) -> &mut SomeStruct {
        let mut a = 0;
        a = a + 1;
        let b = &mut a;
        println!("Hello from my_mut_accessor()!");
        self.a_vector.get_mut(0).unwrap()
    }
}

How to use

e.g.

#[mwt::mwt]
fn my_mwt_method(&'a mut self, other_param: i32) -> &Mwt<bool> {
    #[if_mut] {
        //code for only the mut version of the function
        let i = 0;
    }
    #[not_mut] {
        // code for only the non-mut version of the function
        let i= 1;
    }
    // do something with i
    self.get_mwt_flag_by_index(i)
}

Basically write the mutable version of your function, but for identifiers, replace mut with mwt and for types replace &mut T with &Mwt<T>

You can also use Mwt in types e.g. MyMwtType becomes MyMutType and MyType

Alternatively you can use mwt::maybe_mut if you feel that's more readable. example:

#[mwt::maybe_mut]
pub fn get_maybe_mut<T: 'static + Component>(&mut self) -> Option<&MaybeMut<T>> {
    // use #[if_mut]{} and #[not_mut]{} for conditional behavior
    //      or for cases where mwt isn't powerful enough yet
    //      like .as_ref() vs .as_mut()
    #[if_mut] { println!("if_mut"); }
    #[not_mut] { println!("not_mut"); }
    //   use &MaybeMut<T> for &mut types
    let map: &MaybeMut<HashMap<TypeId,Box<dyn Component>>>
    //    and &maybe_mut(...) for taking mut references
            = &maybe_mut(self.components);
    // use _maybe_mut_ in function calls, etc.
    map.get_maybe_mut(&TypeId::of::<T>())
        .and_then(|c| c.cast_maybe_mut::<T>())
}

results in two functions:

pub fn get_<T: 'static + Component>(& self) -> Option<&T> {
    println!("not_mut"); 
    let map: &HashMap<TypeId,Box<dyn Component>> = &self.components;
    map.get(&TypeId::of::<T>()).and_then(|c| c.cast::<T>())
}
pub fn get_mut<T: 'static + Component>(&mut self) -> Option<&mut T> {
    println!("if_mut");
    let map: &mut HashMap<TypeId,Box<dyn Component>> = &mut self.components;
    map.get_mut(&TypeId::of::<T>()).and_then(|c| c.cast_mut::<T>())
}

What's it actually doing?

mwt::mwt basically just replaces the function with two copies (i.e. a non-mut and mut version) and does a few things on those:

  • replace any occurrences of type references like &Mwt<T> with &T and &mut T respectively
  • replace any occurences of mwt(expr) with expr and mut expr respectively
  • for the non-mut version of the function:
    • it takes all identifiers it finds and trims any starting "mwt_" and ending "_mwt" and replaces "_mwt_" with "_"
    • it takes all types it finds and removes any instances of "Mwt"
  • for the mut version of the function:
    • it takes all identifiers it finds and replaces any instances of "mwt" with "mut"
    • it takes all types it finds and replaces any instances of "Mwt" with "Mut"
  • to allow for other ways behavior can differ, the mut version strips any occurences of #[not_mut]{...} and the non-mut version strips any occurrences of #[if_mut]{...} (the ones that aren't stripped have their braces removed, so be aware of that)
  • to allow for differing types, MwtAlt<First, Second> is replaced with either First or Second in the mut and non-mut versions respectively

mwt::maybe_mut is identical just with different strings. (Mwt -> MaybeMut, mwt -> maybe_mut, MwtAlt -> MutOrElse)


Found a bug? Need a feature?

Please file an issue or submit a pull request!

Dependencies

~1.5MB
~35K SLoC