erezsh / runtype Goto Github PK

View Code? Open in Web Editor NEW

159.0 159.0 3.0 535 KB

Utilities for run-time type validation and multiple dispatch

License: MIT License

Python 100.00%

dataclass multiple-dispatch type-validation typing

runtype's Introduction

👋 Welcome to my Github profile! Have a look at my open-source projects:

Lark, an innovative parsing toolkit, that's grown to be the most popular parser in the Python ecosystem!
Data-diff, a tool for cross-database diffing, that was generously funded by Datafold.
Preql, an ambitious new programming language that compiles to SQL. (inactive, but I plan to resume working on it some day!)
Runtype, a collection of run-time type utilities including validation and multiple dispatch.
Many more!

👩‍💻 My stack is Python for everything, and Svelte+Typescript for webdev. I also write in C++, Rust, Julia, and many others.

🌎 I'm living the life of a digital nomad; I spend a lot of my time in Amsterdam.

👅 I speak Hebrew, English and pocito de Spanish.

💞️ I’m looking to collaborate! Let me know if you have a cool project in mind, or want to help with one of mine. I'm especially interested in programming languages, the semantic-web, machine learning, static analysis, and recommendation systems.

👷‍♂️ I am a freelancer, and may be available for hire for cool and interesting projects, either open-source and proprietry.

🔗 Would you like to know more? Visit my website.

runtype's People

Contributors

Stargazers

Watchers

Forkers

stjordanis 4144 mbrukman

runtype's Issues

Add support for Annotated in `is_subtype` (transparently ignoring the annotations)

In my project I am using pydantic and use its Field(...) configuration e.g. to define alias names for keys within Annotated, e.g.

x: Annotated[int, Field(alias="y")]

but I need to check is_subtype on the wrapped types (I don't use the annotation for any type-domain related info).

It would be great if Annotated could be transparently ignored.

I understand there are many "semantics" one could give to checking Annotated types, but as the annotation is quite arbitrary and tool specific, I think that simply ignoring the annotation and recursing down the actual type would be the "right" default behavior.

A more conservative solution I would also be happy with would be to have this as some keyword argument to is_subtype, called e.g. ignore_annotations to have that behavior and do with them whatever you believe would make more sense as the default.

But in my case, the annotation object (pydantic.FieldInfo) does not have even equality, so this would not work for me to check that the annotations must all be equal or something like that. Thats why I think it does not make any sense to try doing anything with the annotations but ignoring them, as it probably would not work in the general case.

Importing runtype breaks (at least) get_type_hints

Is runtype somehow messing with typing, typing_extensions or pydantic?

A thing that looks relevant in my logs is:

/home/admin/.pyenv/versions/3.8.10/lib/python3.8/typing.py:1264: in get_type_hints
    value = _eval_type(value, globalns, localns)
/home/admin/.pyenv/versions/3.8.10/lib/python3.8/typing.py:270: in _eval_type
    return t._evaluate(globalns, localns)
E   TypeError: _evaluate() missing 1 required positional argument: '_'

Even just importing runtype.validation triggers that problem, so I'd expect it does some monkey patching on import.

Support Union for is_subtype

Hi!

Thanks for creating this project! is_subtype should probably be part of typing or typing_extensions.

Currently, is_subtype does not work for Unions:

>>> import runtype
>>> runtype.is_subtype(str, str | None)
True
>>> runtype.is_subtype(str | None, str | None)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File ".../python3.10/site-packages/runtype/validation.py", line 30, in is_subtype
    return ct1 <= ct2
  File ".../python3.10/site-packages/runtype/pytypes.py", line 130, in __le__
    return issubclass(self.kernel, other.kernel)
TypeError: issubclass() arg 1 must be a class

Does it make sense to support this? I can try to add support for this feature.

In my project, I wanted to check that some Pydantic classes implement some Protocols. I cannot inherit from those Protocols, as both Pydantic and Protocol use (different) metaclasses. So I was comparing the __annotations__ in the Protocol with the ones in the pydantic implementation using is_subtype.

Thanks!

FYI: Now available from AUR

Just FYI, I have packaged runtype for ArchLinux:

https://aur.archlinux.org/packages/python-runtype/

ArchLinux users can install the package by running:

  yay -S python-runtype

runtype.isa([40, 2], tuple[int, int]) == True

Hello,

Test with runtype 0.2.3:

>>> import runtype
>>> runtype.isa([40, 2], tuple[int, int])
True

Expected behavior:

>>> import runtype
>>> runtype.isa([40, 2], tuple[int, int])
False

Remark: runtype does agree that a tuple is not a list:

>>> runtype.isa((40, 2), list[int])
False

is_subtype does not work for Literals

>>> is_subtype(Literal[1], Literal[1,2])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/admin/.cache/pypoetry/virtualenvs/metador-core-XpPben8F-py3.8/lib/python3.8/site-packages/runtype/validation.py", line 25, in is_subtype
    return t1 <= t2
TypeError: '<=' not supported between instances of 'OneOf' and 'OneOf'

Expected: True

`json` method doesn't work with list

As I wanted to add you to apischema benchmark, I've noticed an issue with serialization, which simply doesn't work with list.

from runtype import dataclass

@dataclass
class Bar:
    baz: int

@dataclass
class Foo:
    bars: list[Bar]

assert Foo([Bar(0)]).json() == {"bars": [Bar(baz=0)]}

As you can see, Bar instances are not serialized.

Dispatch and Runtype Dataclasses

In #24 you mentioned how multiple dispatch works fine for constructors. To demonstrate, you introduced a Color class that exhibits this behavior...

from runtype import Dispatch

dp = Dispatch()

class Color:
    @dp
    def __init__(self, hex: str):
        self.hex = hex

    @dp
    def __init__(self, r: int, g:int, b: int):
        self.hex = '#%02x%02x%02x' % (r,g,b)


a = Color(10, 20, 30)
b = Color('#abc')

print(a.hex, b.hex) # Prints: #0a141e #abc

Experimenting with your example, I tried making the Color class into a dataclasses.dataclass, and it still worked great! However, when I made it into a runtype.dataclass, I received the following error...

Traceback (most recent call last):
  File "C:\Work\temp-runtype\main.py", line 20, in <module>
    a = Color(10, 20, 30)
  File "C:\Users\owner\AppData\Local\pypoetry\Cache\virtualenvs\temp-runtype-7V6D2ZSe-py3.10\lib\site-packages\runtype\dispatch.py", line 37, in dispatched_f
    return f(*args, **kw)
  File "C:\Work\temp-runtype\main.py", line 17, in __init__
    self.hex = '#%02x%02x%02x' % (r,g,b)
  File "<string>", line 4, in __setattr__
dataclasses.FrozenInstanceError: cannot assign to field 'hex'

A curious thing to discover from runtype's own dataclass decorator. Is this a bug, or have I missed something?

Using dataclass and multiple dispatch with constructor

Thank you for putting this package together. It is very useful.

One problem I ran into is using dataclass with multiple dispatch on the constructor. Here is an example:

Example

from runtype import Dispatch 
from runtype import dataclass

dp = Dispatch()

@dataclass 
class SomeType:
    
    x: int = 1
    y: int = 1

@dataclass
class MyClass:
    
    x: int = 0
    y: int = 0
    
    @dp
    def __init__(self, some_type : SomeType):
        self.x = some_type.x
        self.y = some_type.y

Error

 MyClass()
Traceback (most recent call last):

  File "/home/dfish/anaconda3/lib/python3.8/site-packages/runtype/dispatch.py", line 84, in find_function_cached
    return self._cache[sig]

KeyError: (<class '__main__.MyClass'>,)


During handling of the above exception, another exception occurred:

Traceback (most recent call last):

  File "/tmp/ipykernel_54810/2409861197.py", line 1, in <module>
    MyClass()

  File "/home/dfish/anaconda3/lib/python3.8/site-packages/runtype/dispatch.py", line 36, in dispatched_f
    f = root.find_function_cached(args)

  File "/home/dfish/anaconda3/lib/python3.8/site-packages/runtype/dispatch.py", line 86, in find_function_cached
    f = self.find_function(args)

  File "/home/dfish/anaconda3/lib/python3.8/site-packages/runtype/dispatch.py", line 72, in find_function
    raise DispatchError(f"Function '{self.name}' not found for signature {self.get_arg_types(args)}")

DispatchError: Function '__init__' not found for signature (<class '__main__.MyClass'>,)

It would be nice to use dataclass and multiple dispatch together, but its not clear to me if that is possible.

No support for type Literal error when using typing_extensions >= 4.6.0

This worked up to typing_extensions version 4.5.0:

from runtype import is_subtype
from typing_extensions import Literal
is_subtype(Literal[1], Literal[1, 2])

Now with typing_extensions 4.6.0 and upward (and Python 3.8) I get this:

Traceback (most recent call last):
  File "/local/home/a.pirogov/.cache/pypoetry/virtualenvs/metador-core-AXME58BY-py3.8/lib/python3.8/site-packages/runtype/pytypes.py", line 541, in to_canon
    return self.cache[t]
KeyError: typing_extensions.Literal[1]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/local/home/a.pirogov/.cache/pypoetry/virtualenvs/metador-core-AXME58BY-py3.8/lib/python3.8/site-packages/runtype/pytypes.py", line 544, in to_canon
    res = _type_cast_mapping[t]
KeyError: typing_extensions.Literal[1]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/local/home/a.pirogov/.cache/pypoetry/virtualenvs/metador-core-AXME58BY-py3.8/lib/python3.8/site-packages/runtype/validation.py", line 28, in is_subtype
    ct1 = type_caster.to_canon(t1)
  File "/local/home/a.pirogov/.cache/pypoetry/virtualenvs/metador-core-AXME58BY-py3.8/lib/python3.8/site-packages/runtype/pytypes.py", line 546, in to_canon
    res = self._to_canon(t)
  File "/local/home/a.pirogov/.cache/pypoetry/virtualenvs/metador-core-AXME58BY-py3.8/lib/python3.8/site-packages/runtype/pytypes.py", line 532, in _to_canon
    raise NotImplementedError("No support for type:", t)
NotImplementedError: ('No support for type:', typing_extensions.Literal[1])

erezsh / runtype Goto Github PK

runtype's Introduction

runtype's People

Contributors

Stargazers

Watchers

Forkers

runtype's Issues

Add support for Annotated in `is_subtype` (transparently ignoring the annotations)

Importing runtype breaks (at least) get_type_hints

Support Union for is_subtype

FYI: Now available from AUR

runtype.isa([40, 2], tuple[int, int]) == True

is_subtype does not work for Literals

`json` method doesn't work with list

Dispatch and Runtype Dataclasses

Using dataclass and multiple dispatch with constructor

No support for type Literal error when using typing_extensions >= 4.6.0

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent