-
-
Notifications
You must be signed in to change notification settings - Fork 50
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Feature Request] Pseudo-callable (i.e., non-function object implementing the __call__()
protocol) support
#211
Comments
ohnoes. A bug, everybody! An actual bug! It's been months since I had one of these. I almost don't know what to do, anymore. Paaaaani– Wait. Panic never solved anything except causing me to bald yet further. It's time to roll up the unwashed sleeves and get greasy with the @beartype codebase again. 🐒 |
...heh. So, it comes to this. @beartype doesn't appear to explicitly support pseudo-callables (i.e., objects masquerading as functions by defining the Let's rectify that, stat! And... there goes the |
beartype
+ ft.wraps
+ class instances__call__()
protocol) support
This commit is the first in a commit chain adding support for **pseudo-callables** (i.e., objects masquerading as functions by defining the `__call__()` dunder method), en-route to resolving feature request #211 kindly submitted by Google X typing guru @patrick-kidger (Patrick Kidger). Specifically, this commit internally details a sane path forward for implementing this in about a microsecond. We're on the clock here, people! ...okay, we're not. But let's time ourselves anyway. Gotta break records like a speed runner here. (*Extravagant extravaganza!*)
This commit is the next in a commit chain adding support for **pseudo-callables** (i.e., objects masquerading as functions by defining the `__call__()` dunder method), en-route to resolving feature request #211 kindly submitted by Google X typing guru @patrick-kidger (Patrick Kidger). Specifically, this commit defines a new private `beartype._decor.decorcore._beartype_method_bound()` decorator decorating bound method descriptors with `@beartype`. (*Dalliance, man! Mandala!*)
This commit is the next in a commit chain adding support for **pseudo-callables** (i.e., objects masquerading as functions by defining the `__call__()` dunder method), en-route to resolving feature request #211 kindly submitted by Google X typing guru @patrick-kidger (Patrick Kidger). Specifically, this commit resolves `pyright` complaints raised by recent versions of `pyright` on our new `beartype._decor.decorcore._beartype_method_bound()` decorator decorating bound method descriptors with `@beartype`. (*Indiscriminate incrimination!*)
This commit is the next in a commit chain adding support for **pseudo-callables** (i.e., objects masquerading as functions by defining the `__call__()` dunder method), en-route to resolving feature request #211 kindly submitted by Google X typing guru @patrick-kidger (Patrick Kidger). Specifically, this commit finalizes our draft implementation of this functionality. Superficially, everything behaves as expected. But let's actually test this, shall we? (*Importune apportion!*)
This commit is the last in a commit chain adding support for **pseudo-callables** (i.e., objects masquerading as functions by defining the `__call__()` dunder method), resolving feature request #211 kindly submitted by Google X typing guru @patrick-kidger (Patrick Kidger). Specifically, this commit exhaustively tests our implementation of this functionality. This commit's for you, Dr. Kidger! (*Illustrious lustrous illusions!*)
This commit is the *real* last in a commit chain adding support for **pseudo-callables** (i.e., objects masquerading as functions by defining the `__call__()` dunder method), resolving feature request #211 kindly submitted by Google X typing guru @patrick-kidger (Patrick Kidger). Specifically, this commit cleans up this functionality a bit in preparation for generalizing @beartype to monkey-patch runtime type-checking into arbitrary objects. Mere anarchy is loosed upon the typing scene! (*Ostentatious ostriches ostensibly tense in riches!*)
Resolved by cab50aa. Bwa-ha! Against all odds, we did something. Interestingly, resolving this pushes @beartype considerably closer to a full-blown runtime type-checking monkey-patcher. After all, @beartype is now only one small step for mankind away from being able to:
Oh – and I think your example's a bit busted, maybe? The low-level from beartype import beartype
from functools import wraps
def f(self, x: int):
return x
class A:
@wraps(f)
def __call__(self, x):
return f(self, x)
a = A()
beartype(a)
# This prints:
# 0xFEEDFACE
print(a(0xFEEDFACE))
# This raises the expected type-checking exception:
# beartype.roar.BeartypeCallHintParamViolation: @beartyped __main__.f()
# parameter x='What horror hath @beartype wraught this time!?' violates
# type hint <class 'int'>, as str 'What horror hath @beartype wraught
# this time!?' not instance of int.
print(a('What horror hath @beartype wraught this time!?')) Thanks so much for flying Air Beartype. We hope you enjoy your flight. To your left, you can see a giant balloon shaped like a bear currently being filled with Sidewinder missiles by an F-35A Lightning II. Meals consisting exclusively of honey and maple syrup will begin shortly. 🎈 |
It's worth bearing (ahem) in mind that your example does something different to my original -- I was wrapping the class, not the What do you mean? |
Ahh... I'm afraid your example doesn't work, though. It is sad. I tried it, too; You can readily verify this yourself at a REPL: e.g., $ ipython3.11
>>> from beartype import beartype
>>> from functools import update_wrapper
>>> def f(x: int): pass
>>> class A:
... def __call__(self, a): pass
>>> a = A()
>>> a = update_wrapper(a, f)
>>> a.__call__.__annotations__
{}
|
It ends up being placed on the class instance as Perhaps
|
nods furiously
That... sounds pretty non-standard, magical, and implicit. I mean, sure. Technically, @beartype can do anything. Nobody in CPython's typing community acknowledges the existence of runtime type-checking. They've left everything up to our perverse discretion. Always a mistake, Guido! Now, we're making up the rules as we go along. But you should probably just propagate type hints onto the from beartype import beartype
from functools import update_wrapper
def f(self, x: int):
return x
class A:
def __call__(self, x):
return f(self, x)
a = A()
update_wrapper(A.__call__, f)
beartype(a)
# Works! Totally works! I am shocked face as well.
print(a(0xFEEDFACE))
print(a('What horror hath @beartype wraught this time!?')) Do not ask what @beartype can do for Google X. Ask what Google X can do for @beartype. 😄 Kidding! Just kidding! Still, the sentiment stands. Typing is an anarchic battle royale and you're personally invited. Or... does even the above not work for your specific use case? Does Google really need @beartype to start making up its own normative pseudo-standards? We could do that, if Google is asking. What Google wants, Google gets. Even cray-cray stuff is on the table. |
Being placed on Intended behaviour here is that each instance of I think doing so is canonical / not sketchy! (Sadly, I'm only using beartype in hobby projects right now. FWIW the complexity of beartype's codebase probably makes it a tough security review before it could be used anywhere internally.) |
Exactly. That's exactly it. Oh... I see. @patrick-kidger doesn't know yet. Gods. Should I be the one to break it to him? I hate being that guy. 😶🌫️ Sadly, you've just hit a deep Python constraint without knowing it. Unlike almost everything else, Thus, >>> assert a.__call__.__func__ is A.__call__
True When you call Again, this is a low-level Python constraint. Although most methods don't work like that,
Hah! The jokes on you, Google. Against all odds, @beartype has already insinuated itself into over 100 upstream open-source packages – some of which I guarantee Google is already using, because they are literally Google projects. You and I are equally shocked. AudioLM-Pytorch is my personal favourite. It's Google. It's @beartype. As an obsessive-compulsive RyMer (i.e., RateYourMusic user), I proudly look forward to the day when Google AI will synthesize a realtime feed of @beartype-checked afro-samba música popular brasileira synthwave that I can jack directly into my neocortex. 🎧 🎶 |
This isn't quite right. You are correct that The reason is that built-in magic methods are looked up on the type: Incidentally, this has an exception. If the class is a Ephemerality of bound methods is unrelated. This shouldn't affect beartype's ability to check class instances as callables. For example, here's a wrapper that fixes beartype's behaviour here (just by wrapping the instance into a function, which has the desired behaviour): import beartype
import functools as ft
# Library code
def fixed_beartype(fn):
@ft.wraps(fn)
def wrapped(*args, **kwargs):
return fn(*args, **kwargs)
return beartype.beartype(wrapped)
# User code
class Kls:
def __call__(self, x):
return x * 2
a = Kls()
b = Kls()
def f(x: int):
...
def g(x: str):
...
ft.update_wrapper(a, f)
ft.update_wrapper(b, g)
assert a.__annotations__ == {"x": int}
assert b.__annotations__ == {"x": str}
a = fixed_beartype(a)
b = fixed_beartype(b)
a(1) # okay
b("hi") # okay
try:
a("hi")
except beartype.roar.BeartypeCallHintParamViolation:
pass # okay
else:
raise Exception
try:
b(1)
except beartype.roar.BeartypeCallHintParamViolation:
pass # okay
else:
raise Exception This works because My expectation is that
I don't think that's a Google project. Lucidrains does reimplementations of the work of various research groups; Google amongst them. Beartype isn't currently used internally. |
This minor release delivers pulse-quickening support for **pandera (pandas) type hints,** **PEP 484,** **PEP 585**, **PEP 591**, **PEP 647**, **PEP 3119**, and **pseudo-callables.** This minor release resolves **12 issues** and merges **2 pull requests.** But first, a quiet word from our wondrous sponsors. They are monocled QA wizards who serve justice while crushing bugs for humanity. High fives, please! ## Beartype Sponsors * [**ZeroGuard:** The Modern Threat Hunting Platform](https://zeroguard.com). *All the signals, All the time.* Thunderous applause echoes through the cavernous confines of the Bear Den. 👏 🐻❄️ 👏 And now... the moment we've waited for. A heinous display of plaintext that assaults all five senses simultaneously. ## Compatibility Added * **Pandera (pandas) type hints** (i.e., ad-hoc PEP-noncompliant type hints validating pandas `DataFrame` objects, produced by subscripting factories published by the `pandera.typing` subpackage and validated *only* by user-defined callables decorated by the ad-hoc PEP-noncompliant `@pandera.check_types` runtime type-checking decorator), resolving feature request #227 kindly submitted by @ulfaslakprecis (Ulf Aslak) the Big Boss Typer. @beartype now: * Transparently supports pandera's PEP-noncompliant `@pandera.check_types` decorator for deeply runtime type-checking arbitrary pandas objects. * *Always* performs a rudimentary `O(1)` `isinstance()`-based type-check for each Pandera type hint. Doing so substantially improves usability in common use cases, including: * Callables annotated by one or more pandera type hints that are correctly decorated by @beartype but incorrectly *not* decorated by the pandera-specific `@pandera.check_types` decorator. * (Data)classes annotated by one or more pandera type hints. * Pandera type hints passed as the second argument to statement-level @beartype type-checkers – including: * `beartype.door.is_bearable()`. * `beartype.door.die_if_unbearable()`. * Implements a non-trivial trie data structure to efficiently detect all type hints produced by subscriptable factories in the `pandera.typing` submodule. Let us pretend this never happened, @ulfaslakprecis. * **PEP 484- and 585-compliant generator constraints.** This release relaxes prior constraints erroneously imposed by @beartype prohibiting both asynchronous and synchronous generator callables from being annotated as returning unsubscripted standard abstract base classes (ABCs) defined by the `collections.abc` module. Now, @beartype permits: * Asynchronous generator callables to be annotated as returning the unsubscripted `collections.abc.AsyncGenerator` type. * Synchronous generator callables to be annotated as returning the unsubscripted `collections.abc.Generator` type. * **PEP 591** (i.e., `typing.Final[...]` type hints), partially resolving issue #223 kindly submitted by the acronym known only as @JWCS (Jude). @beartype now trivially reduces *all* `typing.Final[{hint}]` type hints to merely `{hint}` (e.g., `typing.Final[int]` to `int`). In other words, @beartype no longer raises exceptions when confronted with final type hints and instead at least tries to do the right thing. This still isn't *quite* what everyone wants @beartype to do here; ideally, @beartype should also raise exceptions on detecting attempts to redefine instance and class variables annotated as `Final[...]`. Doing so is *definitely* feasible and exactly what @beartype should *eventually* do – but also non-trivial, because whatever @beartype *eventually* does needs to preserve compatibility with all implementations of the `@dataclass` decorator across all versions of Python now and forever. Cue that head-throbbing migraine. It's comin'! Oh, I can feel it! * **PEP 647** (i.e., `typing.TypeGuard[...] type hints`), resolving feature request #221 kindly submitted by Google X researcher extraordinaire @patrick-kidger. @beartype now trivially reduces *all* `typing.TypeGuard[...]` type hints to the builtin `bool` type. ## Compatibility Improved * **PEP 3119.** @beartype now detects both **non-isinstanceable classes** (i.e., classes whose metaclasses define PEP 3119-compliant `__instancecheck__()` dunder methods unconditionally raising `TypeError` exceptions) and **non-issubclassable classes** (i.e., classes whose metaclasses define PEP 3119-compliant `__subclasscheck__()` dunder methods unconditionally raising `TypeError` exceptions) more narrowly for safety, resolving issue #220 kindly submitted by *ex*traordinary Google X researcher @patrick-kidger (Patrick Kidger). Notably, @beartype now *only* accepts `TypeError` exceptions as connoting non-isinstanceability and non-issubclassability. Previously, @beartype broadly treated any class raising any exception whatsoever when passed as the second parameter to `isinstance()` and `issubclass()` as non-isinstanceable and non-issubclassable. Sadly, doing so erroneously raises false positives for isinstanceable and issubclassable metaclasses that have yet to be fully "initialized" at the early time the `@beartype` decorator performs this detection. ## Features Added * **Pseudo-callable monkey-patching support.** `@beartype` now supports **pseudo-callables** (i.e., otherwise uncallable objects masquerading as callable by defining the `__call__()` dunder method), resolving feature request #211 kindly submitted by Google X typing guru @patrick-kidger (Patrick Kidger). When passed a pseudo-callable whose `__call__()` method is annotated by one or more type hints, `@beartype` runtime type-checks that method in the standard way. ## Documentation Revised * **Literally everything,** also known as the release that migrated `README.rst` -> [Read the Docs (RtD)](https://beartype.readthedocs.io), resolving both issue #203 kindly submitted by @LittleBigGene (AKA the dynamo of the cell) and ancient issue #8 kindly submitted by @felix-hilden (AKA the Finnish computer vision art genius that really made all of this possible). Readable documentation slowly emerges from the primordial soup of @beartype's shameless past for which we cannot be blamed. @leycec was young and "spirited" back then. Specifically, this release: * Coerces our prior monolithic slab of unreadable `README.rst` documentation into a website graciously hosted by Read the Docs (RtD) subdividing that prior documentation into well-structured pages, resolving issue #203 kindly submitted by @LittleBigGene (AKA the dynamo of the cell). * Documents *most* previously undocumented public APIs in the @beartype codebase. Although a handful of public APIs remain undocumented (notably, the `beartype.peps` submodule), these undocumented APIs are assumed to either be sufficiently unpopular or non-useful to warrant investing additional scarce resources here. * Updates our installation instructions to note @beartype's recent availability as official packages in the official package repositories of various Linux distributions. Truly, this can only be the final mark of pride. These include: * Gentoo Linux's Portage tree. * Arch Linux's Arch User Repository (AUR). * Improves the Python code sample embedded in the ["Are We on the Worst Timeline?" subsection of our **Beartype Errors** chapter](https://beartype.readthedocs.io/en/latest/api_roar/#are-we-on-the-worst-timeline). Thanks to @JWCS for their related pull request (PR) #210, which strongly inspired this bald-faced improvement to the usability of our `beartype.typing` API. * Circumvents multiple long-standing upstream issues in the PyData Sphinx theme regarding empty left sidebars via the requisite `_templates/sidebar-nav-bs.html` template hack shamelessly copy-pasted into literally *every* project requiring this theme. This includes @beartype, because why not spew boilerplate that nobody understands everywhere? Sadly, doing so requires pinning to a maximum obsolete version of this theme that will surely die soon. And this is why I facepalm. These issues include: * pydata/pydata-sphinx-theme#90. * pydata/pydata-sphinx-theme#221. * pydata/pydata-sphinx-theme#1181. * Truncates our `README.rst` documentation to a placeholder stub that just directs everyone to RtD instead. * Improves `linecache` integration commentary. Specifically, a pull request by @faangbait (AKA the little-known third member of Daft Punk) improves internal commentary in our private `beartype._util.func.utilfuncmake.make_func()` factory function responsible for dynamically synthesizing new in-memory functions on-the-fly. Our suspicious usage of `None` as the second item of tuples added as values to the standard `linecache.cache` global dictionary has now been documented. Thanks so much for this stupendous contribution, @faangbait! ## Tests Improved * **Mypy integration.** This release improves our `test_pep561_mypy()` integration test to intentionally ignore unhelpful non-fatal warnings improperly emitted by mypy (which encourage usage of `typing_extensions`, oddly enough). * **Sphinx integration.** This release resolves multiple intersecting issues involving integration testing of Sphinx + @beartype, including: * `test_beartype_in_sphinx()` h0tfix is h0t. This release generalizes our test-specific `test_beartype_in_sphinx()` integration test to support arbitrary versions of Sphinx, resolving issue #209 kindly submitted by @danigm the sun-loving Málaga resident who frolics in the sea that Canadians everywhere are openly jealous of. Specifically, this release fundamentally refactors this integration test to fork a new Python interpreter as a subprocess of the current `pytest` process running the `sphinx-build` command. * A Python 3.7-specific failure in our continuous integration (CI) workflow caused by Sphinx attempting to call deprecated functionality of the third-party `pkg_resources` package. This release simply avoids installing Sphinx entirely under Python 3.7; although admittedly crude, it's unclear how else @beartype could possibly resolve this. Since Python 3.7 has almost hit its official End-Of-Life (EOL) and thus increasingly poses a security concern, this is hardly the worst resolution ever. Really! Believe what we're saying. Break nothing! It's the @beartype way. This is why @leycec cries like a mewling cat with no milk. (*Thrilling chills spill towards an untoward ontology!*)
With beartype version
0.12.0
::(
The text was updated successfully, but these errors were encountered: