mys 0.35.0

Strongly typed Python to C++ transpiler.

🐁 Mys

The Mys (/maɪs/) programming language - an attempt to create a statically typed Python-like language that produces fast binaries.

Mys is heavily inspired by Python’s syntax and Rust’s packaging.

Project homepage: https://github.com/eerimoq/mys

🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧

IMPORTANT INFORMATION

Mys is currently this README (the language specification), the examples folder that contains Mys packages, and the tests folder that contains Mys source code (.mys) and it’s manually written “generated” C++ code (.mys.cpp).

The language and build system implementation is still in a very early stage. Some arithmetic, print and conditional statements works, but not much more.

🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧 🚧

Installation

Install Python 3.6 or later, and then install Mys using pip.

$ pip install mys

You must also have a recent versions of g++ and make installed.

Tutorial

First of all, create a package called foo with the command mys new foo, and then enter it. This package is used in throughout the tutorial.

Two files are created; Package.toml and src/main.mys. Package.toml contains the package configuration and src/main.mys the application source code.

The source code is the hello world application. It looks like this:

def main():
    print('Hello, world!')

Build and run the application with the command mys run. It prints Hello, world!, just as expected.

Replace the code in src/main.mys with the code below. It examplifies how to use functions, classes, exceptions, types and command line arguments. The syntax is almost identical to Python, so most readers should easily understand it.

NOTE: This code does not yet work. This is just an example of what an application could look like in the future. The Fibonacci example works, so try that instead!

def func_1(a: int) -> (int, Final[str]):
    return 2 * a, 'Bar'


def func_2(a: int, b: int=1) -> int:
    for i in range(b):
        a += i * b

    return a


def func_3(a: Optional[int]) -> int:
    if a is None:
        return 0
    else:
        return 2 * a


def func_4(a: int) -> {int: [float]}:
    return {
        1: [],
        10 * a: [7.5, -1.0]
    }


def func_5():
    try:
        raise Exception()
    except:
        print('func_5():      An exception occurred.')


class Calc:

    def __init__(self, value: int):
        self.value = value

    def triple(self):
        self.value *= 3


def main(args: [str]):
    value = int(args[1])
    print('func_1(value):', func_1(value))
    print('func_2(value):', func_2(value))
    print('func_3(None): ', func_3(None))
    print('func_3(value):', func_3(value))
    print('func_4(value):', func_4(value))
    func_5()
    calc = Calc(value)
    calc.triple()
    print('calc:         ', calc)

Build and run it.

$ mys run 5
func_1(value): (5, 'Bar')
func_2(value): 7
func_3(None):  0
func_3(value): 10
func_4(value): {1: [], 50: [7.5, -1,0]}
func_5():      An exception occurred.
calc:          Calc(value=15)

Built-in functions and classes

Built-in functions and classes
abs()	all()	any()	bool()	bytes()
chr()	dict()	divmod()	enumerate()	float()
format()	int()	len()	list()	min()
max()	open()	ord()	print()	range()
reversed()	round()	str()	sum()	tuple()
zip()

All built-ins aims to behave like their Python counterparts, with the following differences.

• abs() only supports integer and floating point numbers.

• all() and any() only supports lists of bool().

• min() and max() only supports lists of integer and floating point numbers, and a fixed number of integer and floating points parameters.

• sum() only supports lists of integer and floating point numbers.

Types

Variables may all be set to None if declared as Optional.

Variables declared as Final can’t be modified.

Type	Example	Comment
int	1, -1000	An integer. Usually 32 or 64 bits.
float	5.5, -100.0	A floating point number. Usually 32 bits.
str	'Hi!'	A unicode string.
bytes	b'\x00\x43'	A sequence of bytes.
tuple(T1, T2, ...)	(5.0, 5, 'foo')	A tuple with items of types T1, T2, etc.
list(T)	[5, 10, 1]	A list with items of type T.
dict(TK, TV)	{5: 'a', -1: 'b'}	A dictionary with keys of type TK and values of type TV.

Packages

A package contains modules that other packages can use. All packages contains a file called lib.mys, which is imported with import <package>.

There are two kinds of packages; library packages and application packages. The only difference is that application packages contains a file called src/main.mys, which contains the application entry point def main(...). Application packages produces an executable when built (mys build), libraries does not.

A package:

my-package/
├── Package.toml
├── README.rst
├── src/
│   ├── lib.mys
│   └── main.mys         # Only part of application packages.
└── tst/
    └── test_lib.mys

The mys command line interface:

mys new      Create a new package.
mys build    Build the appliaction.
mys run      Build and run the application.
mys clean    Remove build output.
mys publish  Publish a release.

Importing packages

• Import a package with import <package>.

• Import a module with import <package>[.<sub-package>]*.<module>.

• Import selected functions and classes with from <package>[.<sub-package>]*.<module> import <function/class>.

Use import ... as <name> to use a custom name.

Here are a few examples:

import mypkg1
import mypkg2.mod1
import mypkg2.subpkg1.mod1
from mypkg3.subpkg1.mod1 import func1
from mypkg3.subpkg1.mod1 import func2 as func3

def foo():
    mypkg1.func()
    mypkg2.mod1.func()
    mypkg2.subpkg1.mod1.func()
    func1()
    func3()

Package internal imports

Here are a few examples:

import .mod1
import ..subpkg1.mod1
from ...subpkg1.mod1 import func1

def foo():
    mod1.func()
    subpkg1.mod1.func()
    func1()

Extending Mys with C++

Extending Mys with C++ is extremly easy and flexible. Strings that starts with mys-embedded-c++ are inserted at the same location in the generated code.

def main():
    a: int = 0

    '''mys-embedded-c++

    int b = 2;
    a++;
    '''

    print('a + b:', a + b)

Memory management

Integers and floating point numbers are allocated on the stack, passed by value to functions and returned by value from functions, just as any C++ program.

Strings, bytes, tuples, lists, dicts and classes are normally allocated on the heap and managed by C++ shared pointers. Objects that are known not to outlive a function are allocated on the stack.

Reference cycles are not detected and will result in memory leaks.

There is no garbage collector.

Major differences to Python

• All variables must have a known type at compile time. The same applies to function parameters and return value.

• Threads can run in parallel. No GIL exists.

  WARNING: Data races will occur when multiple threads uses a variable at the same time, which will likely make the program crash.

• Integers and floats have a platform dependent maximum size, usually 32 or 64 bits.

• Decorators does not exist.

• Variable function arguments *args and **kwargs are not supported, except to some built-in functions.

• Async is not supported.

• Generators are not supported.

• The majority of the standard library is not implemented.

• Dictionary keys must be integers, floats, strings or bytes.

• Strings, bytes and tuple items are mutable by default. Mark them as Final to make them immutable.

• Classes and functions are private by default. Decorate them with @public to make them public. Variables are always private.

Text editor settings

Visual Code

Use the Python language for *.mys files by modifying your files.associations setting.

See the official Visual Code guide for more detils.

"files.associations": {
    "*.mys": "python"
}

Emacs

Use the Python mode for *.mys files by adding the following to your .emacs configuration file.

(add-to-list 'auto-mode-alist '("\\.mys\\'" . python-mode))

Performance

ToDo: Create a benchmark and present its outcome in this section.

Build time

Mys should be slower.

Startup time

Mys should be faster.

Runtime

Mys should be faster.

Memory usage

Mys should use less memory.

Build process

mys run and mys build does the following:

1. Uses Python’s parser to transform the source code to an Abstract Syntax Tree (AST).

2. Generates C++ code from the AST.

3. Compiles the C++ code with g++.

4. Statically links the program with g++.

Notebook for the developer

Importing ideas:

// import pkg
#include "pkg/lib.mys.hpp"

// import pkg.mod
#include "pkg/mod.mys.hpp"

// Function alias when using import ... as <name>.
constexpr auto bar = [] (auto &&...args) {
    return foo(std::forward<decltype(args)>(args)...);
};

// Class alias when using import ... as <name>.
typedef <package>::<module>::MyClass <name>;