lythp 0.0.1

Python as a LISP

Lythp

It's Python turned into a LISP!

(def greet (name)
    """A function which greets someone.

        >>> (greet "Jim")
        Hello Jim!

    """
    (print (+ "Hello " name "!"))
)

The goal is not to preserve LISP traditions, keywords, and features; rather, the goal is to have fun fitting Python into a LISP syntax in the most natural way I can find, using Python's built-in tokenizer.

Here is a slightly more interesting example:

# A global dict for caching function values:
(= _fib_cache (dict))


(def fib (n)
    """Returns the nth Fibonacci number.
    Maintains a global cache of values, to avoid needless recalculation."""
    (if
        ((< n 0) (raise (ValueError "Less than 0")))
        ((< n 2) n) # Base cases: 0, 1
        ((in _fib_cache n) ([n] _fib_cache))
        (else
            (= value (+ (fib (- n 1)) (fib (- n 2)) ))
            (= [n] _fib_cache value)
            value
        )
    )
)


# Print out the first 10 Fibonacci numbers:
(for x (range 10) (print (fib x)))

For more examples, see: examples

The interpreter

Currently, Lythp has its own runtime, instead of transpiling to Python. For instance, variables are stored in a stack of contexts, i.e. a list of dicts. It would be nice to transpile to Python instead, either by using exec or by generating bytecode directly.

In any case, make sure you're in a python3 virtual environment:

python3 -m venv venv
. venv/bin/activate

Then run files like so:

./lythp.py examples/fac.lsp

You can also run the interpreter in REPL mode like so:

./lythp.py

It may be helpful to install rlwrap to improve the REPL experience:

rlwrap ./lythp.py

You can run the unit tests like so:

pip install pytest pytest-cov
pytest

And run all example programs like so:

./run.sh

Syntax

The basic syntax is quite simple: the source code is chopped into tokens using Python's built-in tokenize module, and a tree structure representing the program is then built, with the following types of node:

• Names: x, None, True, ,, *, ==
• Literals: 100, "Hello!", """Docstrings too!"""
• Parentheses: (...)
• Brackets: [...]
• Braces: {...}

NOTE: f-strings (e.g. f"Hello, {name}!") are not currently supported.

As usual in a LISP, parentheses represent statements and function calls. Here is the syntax of specific statements:

Built-ins and literals

The following values have the same syntax as in Python:

None
True
False
123
"hello"
r"(\n+)"
b"beep boop"
...  # the "ellipsis" object

TODO: describe lists, dicts, sets, etc

Attribute and item lookup: ., [...]

Basic usage:

# Python
obj.x
arr[i]
obj.x.y.z[n + 1]

# Lythp
(.x obj)
([i] arr)
(.x.y.z[+ n 1] obj)

Assignment: =

Basic usage:

# Python
x = 1
x += 1

# Lythp
(= x 1)
(+= x 1)

Assigning with setattr/setitem:

# Python
obj.x.y.z[n + 1] = value

# Lythp
(= .x.y.z[+ n 1] obj value)

NOTE: there is no equivalent of Python's destructuring (e.g. x, y = a, b).

Function calls

Basic usage:

# Python
x + y
f(x, y, z=3)
obj.method(x)

# Lythp
(+ x y)
(f x y [z 3])
((.method obj) x)

Args and kwargs:

# Python
f(x, y, *args, z=3, **kwargs)

# Lythp
(f x y [*args] [z 3] [**kwargs])

Lambdas and function definitions: lambda, def

Basic usage:

# Python
def f(x, y, z=3): ...etc...
f = lambda x: x + 1

# Lythp
(def f (x y [z 3]) ...etc...)
(= f (lambda (x) (+ x 1)))

NOTE: while Python only allows a single expression per lambda, Lythp allows statements, like a regular function definition.

Args and kwargs:

# Python
def f(x, /. y, *args, z=3, **kwargs): ...etc...

# Lythp
(def f (x [/] y [*args] [z 3] [**kwargs]) ...etc...)

Classes

# Python
class A(B, C):
    x = 3
    def __init__(self, value):
        ...etc...

a = A(value)

# Lythp
(class A (B C)
    (= x 3)
    (def __init__ (self value)
        ...etc...
    )
)

(= a (A value))