Try to evaluate forward refs after model created #2588

uriyyo · 2021-03-28T20:31:05Z

Change Summary

Try to evaluate forward refs after model created.

So it will not be required to use update_forward_refs in the case of self-referencing model:

from __future__ import annotations

from pydantic import BaseModel
from typing import Optional


class Foo(BaseModel):
    foo: Optional[Foo]


assert Foo.__fields__['foo'].type_ is Foo

Checklist

Unit tests for the changes exist
Tests pass on CI and coverage remains at 100%
Documentation reflects the changes where applicable
changes/<pull request or issue id>-<github username>.md file added describing change
(see changes/README.md for details)

PrettyWood

Can you please rebase/merge with master to trigger the new CI?

pydantic/main.py

samuelcolvin · 2021-05-09T13:03:42Z

you have conflicts and the wrong change.

Please update.

samuelcolvin

looking good, but this also needs a note in the docs.

pydantic/main.py

uriyyo · 2021-05-09T15:23:44Z

@samuelcolvin I have updated PR. Could you please review it again?

PrettyWood

Cool feature! Nice to be able to do everything automatically

changes/2588-uriyyo.md

docs/usage/postponed_annotations.md

PrettyWood · 2021-09-04T18:05:08Z

pydantic/main.py

@@ -374,6 +373,8 @@ def is_untouched(v: Any) -> bool:
        cls = super().__new__(mcs, name, bases, new_namespace, **kwargs)
        # set __signature__ attr only for model class, but not for its instances
        cls.__signature__ = ClassAttribute('__signature__', generate_model_signature(cls.__init__, fields, config))
+        cls.__try_update_forward_refs__()


What is the impact on perf for models without forward refs? Do we have a way to know whether or not we should try to update?

I wrote a simple benchmark:

import timeit from pydantic import BaseModel def callback(): class Order(BaseModel): a: int b: float c: str t = timeit.timeit( callback, number=10_000, globals={"BaseModel": BaseModel}, ) print(f'{t:.5f}')

Results:

With __try_update_forward_refs__ - 3.17182

Without __try_update_forward_refs__ - 3.15772

PrettyWood · 2021-09-04T18:05:47Z

Hi @uriyyo
Can you please rebase/merge and update? 🙏

Co-authored-by: Eric Jolibois <em.jolibois@gmail.com>

…d-refs # Conflicts: # pydantic/main.py

PrettyWood

Yes I run the benchmark seems ok performance-wise

This branch

▶ BENCHMARK_REPEATS=20 make benchmark-pydantic
python benchmarks/run.py pydantic-only
testing pydantic, 20 times each
    pydantic ( 1/20) time=0.622s, success=48.65%
    pydantic ( 2/20) time=0.617s, success=48.65%
    pydantic ( 3/20) time=0.598s, success=48.65%
    pydantic ( 4/20) time=0.596s, success=48.65%
    pydantic ( 5/20) time=0.597s, success=48.65%
    pydantic ( 6/20) time=0.606s, success=48.65%
    pydantic ( 7/20) time=0.595s, success=48.65%
    pydantic ( 8/20) time=0.598s, success=48.65%
    pydantic ( 9/20) time=0.591s, success=48.65%
    pydantic (10/20) time=0.597s, success=48.65%
    pydantic (11/20) time=0.595s, success=48.65%
    pydantic (12/20) time=0.605s, success=48.65%
    pydantic (13/20) time=0.603s, success=48.65%
    pydantic (14/20) time=0.601s, success=48.65%
    pydantic (15/20) time=0.611s, success=48.65%
    pydantic (16/20) time=0.603s, success=48.65%
    pydantic (17/20) time=0.600s, success=48.65%
    pydantic (18/20) time=0.596s, success=48.65%
    pydantic (19/20) time=0.592s, success=48.65%
    pydantic (20/20) time=0.605s, success=48.65%
    pydantic best=0.591s, avg=0.601s, stdev=0.008s

    pydantic best=98.503μs/iter avg=100.240μs/iter stdev=1.335μs/iter version=1.8.2

1.8

▶ BENCHMARK_REPEATS=20 make benchmark-pydantic
python benchmarks/run.py pydantic-only
testing pydantic, 20 times each
    pydantic ( 1/20) time=0.608s, success=48.65%
    pydantic ( 2/20) time=0.588s, success=48.65%
    pydantic ( 3/20) time=0.616s, success=48.65%
    pydantic ( 4/20) time=0.632s, success=48.65%
    pydantic ( 5/20) time=0.599s, success=48.65%
    pydantic ( 6/20) time=0.588s, success=48.65%
    pydantic ( 7/20) time=0.591s, success=48.65%
    pydantic ( 8/20) time=0.595s, success=48.65%
    pydantic ( 9/20) time=0.589s, success=48.65%
    pydantic (10/20) time=0.612s, success=48.65%
    pydantic (11/20) time=0.599s, success=48.65%
    pydantic (12/20) time=0.594s, success=48.65%
    pydantic (13/20) time=0.623s, success=48.65%
    pydantic (14/20) time=0.591s, success=48.65%
    pydantic (15/20) time=0.592s, success=48.65%
    pydantic (16/20) time=0.595s, success=48.65%
    pydantic (17/20) time=0.595s, success=48.65%
    pydantic (18/20) time=0.600s, success=48.65%
    pydantic (19/20) time=0.625s, success=48.65%
    pydantic (20/20) time=0.587s, success=48.65%
    pydantic best=0.587s, avg=0.601s, stdev=0.014s

    pydantic best=97.825μs/iter avg=100.158μs/iter stdev=2.254μs/iter version=1.8

PrettyWood

LGTM!

samuelcolvin · 2021-12-05T14:21:04Z

this is great, thank you so much.

* Try to evaluate forward refs after model created * Upadate docs and remove code duplication * Update changes/2588-uriyyo.md Co-authored-by: Eric Jolibois <em.jolibois@gmail.com> * Update docs/usage/postponed_annotations.md Co-authored-by: Eric Jolibois <em.jolibois@gmail.com> * Remove unused import Co-authored-by: Eric Jolibois <em.jolibois@gmail.com>

PrettyWood reviewed Apr 5, 2021

View reviewed changes

pydantic/main.py Outdated Show resolved Hide resolved

PrettyWood added the awaiting author revision awaiting changes from the PR author label Apr 5, 2021

uriyyo force-pushed the feature/auto-forward-refs branch from fc89c5a to 286b730 Compare April 5, 2021 18:06

PrettyWood added ready for review and removed awaiting author revision awaiting changes from the PR author labels Apr 5, 2021

github-actions bot added awaiting author revision awaiting changes from the PR author and removed ready for review labels May 9, 2021

github-actions bot assigned uriyyo May 9, 2021

samuelcolvin reviewed May 9, 2021

View reviewed changes

pydantic/main.py Outdated Show resolved Hide resolved

pydantic/main.py Outdated Show resolved Hide resolved

uriyyo added 2 commits May 9, 2021 17:45

Try to evaluate forward refs after model created

f00aa8e

Upadate docs and remove code duplication

cb5ee00

uriyyo force-pushed the feature/auto-forward-refs branch from 286b730 to cb5ee00 Compare May 9, 2021 15:20

uriyyo requested a review from samuelcolvin May 9, 2021 15:23

github-actions bot added ready for review and removed awaiting author revision awaiting changes from the PR author labels May 9, 2021

github-actions bot assigned PrettyWood and samuelcolvin and unassigned uriyyo May 9, 2021

PrettyWood reviewed Sep 4, 2021

View reviewed changes

PrettyWood assigned uriyyo and unassigned samuelcolvin and PrettyWood Sep 4, 2021

PrettyWood added awaiting author revision awaiting changes from the PR author and removed ready for review labels Sep 4, 2021

uriyyo and others added 3 commits September 4, 2021 21:20

Update changes/2588-uriyyo.md

6b8f0d1

Co-authored-by: Eric Jolibois <em.jolibois@gmail.com>

Update docs/usage/postponed_annotations.md

31ba3e1

Co-authored-by: Eric Jolibois <em.jolibois@gmail.com>

Merge remote-tracking branch 'origin/master' into feature/auto-forwar…

06583fb

…d-refs # Conflicts: # pydantic/main.py

Remove unused import

db847c2

PrettyWood reviewed Sep 5, 2021

View reviewed changes

PrettyWood approved these changes Sep 5, 2021

View reviewed changes

PrettyWood added ready for review and removed awaiting author revision awaiting changes from the PR author labels Sep 5, 2021

samuelcolvin merged commit 415eb54 into pydantic:master Dec 5, 2021

PrettyWood mentioned this pull request Jan 20, 2022

Regression in v1.9.0: dataclass as a field type on itself #3675

Closed

3 tasks

followben mentioned this pull request Feb 11, 2022

Migrate to Pydantic ketgo/py-abac#22

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

Try to evaluate forward refs after model created #2588

Try to evaluate forward refs after model created #2588

Uh oh!

uriyyo commented Mar 28, 2021

Uh oh!

PrettyWood left a comment

Uh oh!

Uh oh!

samuelcolvin commented May 9, 2021

Uh oh!

samuelcolvin left a comment

Uh oh!

Uh oh!

Uh oh!

uriyyo commented May 9, 2021

Uh oh!

PrettyWood left a comment

Uh oh!

Uh oh!

Uh oh!

PrettyWood Sep 4, 2021

Uh oh!

uriyyo Sep 4, 2021

Uh oh!

PrettyWood commented Sep 4, 2021

Uh oh!

PrettyWood left a comment

Uh oh!

PrettyWood left a comment

Uh oh!

samuelcolvin commented Dec 5, 2021

Uh oh!

Uh oh!

Uh oh!

Try to evaluate forward refs after model created #2588

Try to evaluate forward refs after model created #2588

Uh oh!

Conversation

uriyyo commented Mar 28, 2021

Change Summary

Checklist

Uh oh!

PrettyWood left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

samuelcolvin commented May 9, 2021

Uh oh!

samuelcolvin left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

uriyyo commented May 9, 2021

Uh oh!

PrettyWood left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

PrettyWood Sep 4, 2021

Choose a reason for hiding this comment

Uh oh!

uriyyo Sep 4, 2021

Choose a reason for hiding this comment

Uh oh!

PrettyWood commented Sep 4, 2021

Uh oh!

PrettyWood left a comment

Choose a reason for hiding this comment

This branch

1.8

Uh oh!

PrettyWood left a comment

Choose a reason for hiding this comment

Uh oh!

samuelcolvin commented Dec 5, 2021

Uh oh!

Uh oh!