Types

Types of arguments of arguments and return types can be specified using type hints as usual. You can use anything from typing. Under the hood, Plum uses Beartype, which means that all types and type hints supported by Beartype are also supported by Plum. Here are a few examples:

from typing import Union, Optional, List, Dict

from plum import dispatch


@dispatch
def f(x) -> str:
    return "fallback"


@dispatch
def f(x: int, *xs: int) -> str:
    return "one or more ints"


@dispatch
def f(x: Union[int, str]) -> str:
    return "int or str"


@dispatch
def f(x: list) -> str:
    return "list"


@dispatch
def f(x: List[int]) -> str:
    return "list of int"


@dispatch
def f(x: Optional[dict]) -> Optional[str]:
    return "dict or None"


@dispatch
def f(x: Dict[int, str]) -> str:
   return "dict of int to str"

Note: Although parametric types such as List[int] and Dict[int, str] are fully supported, they do incur a performance penalty. For optimal performance, is recommended to use parametric types only where necessary. Union and Optional do not incur a performance penalty.

The type system is covariant, as opposed to Julia’s type system, which is invariant. For example, this means that List[T1] is a subtype of List[T2] whenever T1 is a subtype of T2.

Performance and Faithful Types

Plum achieves performance by caching the dispatch process. Unfortunately, efficient caching is not always possible. Efficient caching is possible for so-called faithful types.

Definition: faithful type

A type t is faithful if, for all x, the following is true:

isinstance(x, t) == issubclass(type(x), t)

For example, int is faithful, since type(1) == int; but Literal[1] is not faithful, since issubclass(int, Literal[1]) is false.

Methods which have signatures that depend only on faithful types will be performant. On the other hand, methods which have one or more signatures with one or more unfaithful types cannot use caching and will therefore be less performant.

Example:

from typing import Literal

from plum import dispatch


@dispatch
def add_5_faithful(x: int):
    return x + 5


@dispatch
def add_5_unfaithful(x: Literal[1]):
    return x + 5

>>> %timeit add_5_faithful(1)  
585 ns ± 6.2 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

>>> %timeit add_5_unfaithful(1)  
6.24 µs ± 68.9 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

Plum implements is_faithful, which is a function that attempts to establish whether a type is faithful or not:

>>> from plum import is_faithful

>>> is_faithful(int)
True

>>> is_faithful(Literal[1])
False

If you implement, e.g., a type with a custom __instancecheck__, then is_faithful will detect this and conservatively say that your type is not faithful. You can tell Plum whether your type is faithful or not by setting __faithful__:

...

class MyClass(metaclass=MyMeta):
    __faithful__ = True   # Yes, `MyClass` is faithful!

    ...

ModuleType

A niche use case is that you might want to depend on types from packages you have not yet imported. This can be useful if these packages either bring a lot of dependencies or are slow to load. This is possible with ModuleType.

Important

After the dependency is imported, you must clear all cache using clear_all_cache! If you do not, due to existing caches, dispatch may behave erroneously.

Example:

from plum import dispatch, clear_all_cache, ModuleType

EagerTensor = ModuleType("tensorflow.python.framework.ops", "EagerTensor")


@dispatch
def f(x: EagerTensor):
    return "An eager TF tensor!"

>>> try: f(1)
... except Exception as e: print(f"{type(e).__name__}: {e}")
NotFoundLookupError: `f(1)` could not be resolved...

>>> g.methods
List of 1 method(s):
    [0] f(x:
        plum.type.ModuleType[tensorflow.python.framework.ops.EagerTensor])
            <function f at ...> @ ...

>>> import tensorflow as tf  # Very slow...

>>> clear_all_cache()  # Clear dispatch cache.

>>> f(tf.ones(5))
'An eager TF tensor!'

The object EagerTensor is a type. You can resolve it to what it points to with resolve_type_hint:

>>> EagerTensor
plum.type.ModuleType[tensorflow.python.framework.ops.EagerTensor]

>>> from plum import resolve_type_hint

>>> resolve_type_hint(EagerTensor)
tensorflow.python.framework.ops.EagerTensor

PromisedType

Another problem that can occur is that you want to depend on a type from your package, but you just cannot yet access it because of circular imports. In this case, you use PromisedType to create a proxy type and then deliver the dependency when it is available.

Important

You must deliver the dependency before the proxy type is used! That is, you cannot use the function that uses the proxy type as a type hint before the dependency is delivered.

from plum import dispatch, clear_all_cache, PromisedType

ProxyInt = PromisedType("SpecialInt")  # Proxy for `int`


@dispatch
def f(x: ProxyInt):
    return "An integer!"

# Deliver the type that `ProxyInt` should point to. Do this before `f` is first used!
ProxyInt.deliver(int)

>>> f(1)
'An integer!'

Like for PromisedType, the object ProxyInt is a type. You can resolve it to what it points to with resolve_type_hint:

>>> ProxyInt
<class 'plum.type.PromisedType[SpecialInt]'>

>>> from plum import resolve_type_hint

>>> resolve_type_hint(ProxyInt)
<class 'int'>