pyala

Python 3.10 to Scala 2.13 Transpiler

Warning Project under development. Not available right now. Come back later.

Tutorial

How to use pyala

Define your python function and add type annotations to its input parameters.
Declare new variables with empty annotated assignment to specify their local scope.
Use one of pyala.to_str, pyala.to_object or pyala.to_file to transpile your code.

from pyala import pyala

# define your functions
def foo():
    ...

def goo():
    ...

# get the source code of a single function
print(pyala.to_str(foo))

# get the source code of multiple functions bundled in an object
print(pyala.to_object(foo, goo, object_name='Example'))

# save the source code in a file
print(pyala.to_file(foo, goo, filepath='/tmp/Example.scala'))

Examples

def foo():
    y: int
    if x < 0:
        y = x**2
    else:
        y = x * 2
    return y + x

Discrepancies

Here are a list of considerations to take into account due to discrepancies between scala and python.

Imports

Imports are not supported except for:

import math

The module scala.math is automatically imported and will mimic python built-in functions related to math, and the library math itself. This means that this code is valid, even though it uses an import:

import math
def foo(x: float):
    return math.sqrt(x)

Reference types

A python reference cannot change type because scala does not allow it.

For instance, this python code is invalid in scala:

x = 4
x = True

because x is defined as a scala.Integer and cannot change to scala.Boolean.

Local Scope

In python, you can return a variable defined in a local scope. This is not allowed in scala.

For instance, this is valid python code:

def foo():
    if True:
        x = 5
    return x

but this is not valid is scala:

def foo() = {
    if (true) {
        val x = 5
    }
    x
}

However, you can access variable defined in a parent scope, like this:

def foo() = {
    var x: Integer = 0
    if (true) {
        x = 5
    }
    x
}

For this reason, pyala requires that you add an annotated assignment for each variable that you want to create:

def foo():
    x: int = 0
    if True:
        x = 5
    return x

Variable function arguments

You cannot assign a new value to a function argument. This code is invalid:

def foo(x: int):
    x = x + 1
    return x

You must create a new variable:

def foo(x: int):
    z = x + 1
    return z

F-string

String interpolation (f-string) is typed in scala. This means that this code is invalid:

def foo(x: int):
    return f"{x:0.2f}"

You can only using float with "f" format:

def foo(x: int):
    z: float = float(x)
    return f"{z:0.2f}"

*args and **kwargs

First, **kwargs are not supported. Second, *args must always comes last in the argument list. This means that this code:

def foo(*x: int, z:int = 0):
    return sum(x) - z

is transpiled to:

def foo(z: Int = 0, x: Int*) = {
  (x.sum - z)
}

Note the order of the parameters is z then x instead of x then z.

You can call this function correctly in python with goo(3,4,5,6,z=7) but it will give a compilation error in scala because the correct call is goo(7,3,4,5,6).

Be extra careful when calling function defined in this way. The order will be correct in python but not in scala.

operator

FloorDiv

The return of floor div will always be a scala.Long, even if you use float.

Eg.: In python: 3.0//2 -> 1.0 (float) In scala: 3.0//2 -> 1 (Long)

hex

In python, hex(-42) returns "-0x2a", in scala it returns "0xffffffd6".

References

Python AST: https://docs.python.org/3/library/ast.html
Built-in Functions: https://docs.python.org/3/library/functions.html
Math library: https://docs.python.org/3/library/math.html

Not Supported

expr NamedExpr

Lambda

Lambda expressions are not supported:

lambda x: x +1

operator MatMul

for loop with continue

async

The keyword async is not supported

ltrottier / pyala