Skip to main content

Type-safe functional tools for Python

Project description


Koda is a collection of practical type-safe tools for Python.

At its core are a number of datatypes that are common in functional programming.


Maybe is similar to Python's Optional type. It has two variants: Nothing and Just, and they work in similar ways to what you may have seen in other languages.

from koda import Maybe, Just, nothing

a: Maybe[int] = Just(5)
b: Maybe[int] = nothing

To know if a Maybe is a Just or a Nothing, you'll need to inspect it.

from koda import Just, Maybe

maybe_str: Maybe[str] = function_returning_maybe_str()

# unwrap by checking instance type
if isinstance(maybe_str, Just):
    print("No value!")

# unwrap with structural pattern matching (python 3.10 +)
match maybe_str:
    case Just(val):
    case Nothing:
        print("No value!")

Maybe has methods for conveniently stringing logic together.

from koda import Just, nothing

def int_add_10(x: int) -> int:
    return x + 10

Just(5).map(int_add_10)  # Just(15)  # Nothing
Just(5).map(int_add_10).map(lambda x: f"abc{x}")  # Just("abc15")


from koda import Maybe, Just, nothing

def safe_divide(dividend: int, divisor: int) -> Maybe[float]:
    if divisor != 0:
        return Just(dividend / divisor)
        return nothing

Just(5).flat_map(lambda x: safe_divide(10, x))  # Just(2)
Just(0).flat_map(lambda x: safe_divide(10, x))  # Nothing
nothing.flat_map(lambda x: safe_divide(10, x))  # Nothing


Result provides a means of representing whether a computation succeeded or failed. To represent success, we can use OK; for failures we can use Err. Compared to Maybe, Result is perhaps most useful in that the "failure" case also returns data, whereas Nothing contains no data.

from koda import Ok, Err, Result 

def safe_divide_result(dividend: int, divisor: int) -> Result[float, str]:
    if divisor != 0:
        return Ok(dividend / divisor)
        return Err("cannot divide by zero!")

Ok(5).flat_map(lambda x: safe_divide_result(10, x))  # Ok(2)
Ok(0).flat_map(lambda x: safe_divide_result(10, x))  # Err("cannot divide by zero!") 
Err("some other error").map(lambda x: safe_divide_result(10, x))  # Err("some other error")

Result can be convenient with try/except logic.

from koda import Result, Ok, Err

def divide_by(dividend: int, divisor: int) -> Result[float, ZeroDivisionError]:
        return Ok(dividend / divisor)
    except ZeroDivisionError as exc:
        return Err(exc)

divided: Result[float, ZeroDivisionError] = divide_by(10, 0)  # Err(ZeroDivisionError("division by zero"))

Another way to perform the same computation would be to use safe_try:

from koda import Result, safe_try

# not safe on its own!
def divide(dividend: int, divisor: int) -> float:
    return dividend / divisor

# safe if used with `safe_try`
divided_ok: Result[float, Exception] = safe_try(divide, 10, 2)  # Ok(5)
divided_err: Result[float, Exception] = safe_try(divide, 10, 0)  # Err(ZeroDivisionError("division by zero"))


There are many other functions and datatypes included. Some examples:


Combine functions by sequencing.

from koda import compose
from typing import Callable

def int_to_str(val: int) -> str:
    return str(val)

def prepend_str_abc(val: str) -> str:
    return f"abc{val}"    

combined_func: Callable[[int], str] = compose(int_to_str, prepend_str_abc)
assert combined_func(10) == "abc10"


Try to get a value from a Mapping object, and return an unambiguous result.

from koda import mapping_get, Just, Maybe, nothing

example_dict: dict[str, Maybe[int]] = {"a": Just(1), "b": nothing}

assert mapping_get(example_dict, "a") == Just(Just(1))
assert mapping_get(example_dict, "b") == Just(nothing)
assert mapping_get(example_dict, "c") == nothing

As a comparison, note that dict.get can return ambiguous results:

from typing import Optional

example_dict: dict[str, Optional[int]] = {"a": 1, "b": None}

assert example_dict.get("b") is None
assert example_dict.get("c") is None

We can't tell from the resulting value whether the None was the value for a key, or whether the key was not present in the dict


Create a lazy function, which will only call the passed-in function the first time it is called. After it is called, the value is cached. The cached value is returned on each successive call.

from random import random
from koda import load_once

call_random_once = load_once(random)  # has not called random yet

retrieved_val: float = call_random_once()
assert retrieved_val == call_random_once()


Convert a Maybe to a Result type.

from koda import maybe_to_result, Just, nothing, Ok, Err

assert maybe_to_result("value if nothing", nothing) == Err("value if nothing")
assert maybe_to_result("value if nothing", Just(5)) == Ok(5)


Convert a Result to a Maybe type.

from koda import result_to_maybe, Just, nothing, Ok, Err

assert result_to_maybe(Ok(5)) == Just(5)
assert result_to_maybe(Err("any error")) == nothing 


Convert an Optional value to a Maybe.

from koda import to_maybe, Just, nothing

assert to_maybe(5) == Just(5)
assert to_maybe("abc") == Just("abc")
assert to_maybe(False) == Just(False)

assert to_maybe(None) == nothing


Koda is intended to focus on a small set of practical data types and utility functions for Python. It will not grow to encompass every possible functional or typesafe concept. Similarly, the intent of this library is to avoid requiring extra plugins (beyond a type-checker like mypy or pyright) or specific typchecker settings. As such, it is unlikely that things like Higher Kinded Types emulation or extended type inference will be implemented in this library.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

koda-0.11.0.tar.gz (9.6 kB view hashes)

Uploaded source

Built Distribution

koda-0.11.0-py3-none-any.whl (8.2 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page