Change Summary

Following work on the pydantic-core branch, this PR implements the config hooks to allow ser_json_temporal to be passed in.

Related issue number

Fixes #10454
Fixes #7940

Performance improvements

The main reason we wanted this was to improve performance when we want to serialize a datetime as a millisecond timestamp. Below is a small script using this branch:

from typing import Annotated

import pendulum
import datetime as dt
from pydantic import BaseModel, PlainSerializer
import tqdm

import time

RANGE = 10

dts = [dt for dt in pendulum.interval(
    pendulum.datetime(2023, 1, 1),
    pendulum.datetime(2024, 1, 1)
).range("seconds", 30)]


T_SerializedDatetime = Annotated[
    dt.datetime,
    PlainSerializer(
        lambda x: x.timestamp() * 1000,
        return_type=float,
        when_used="json",
    ),
]

class MyModelOld(BaseModel):
    """A model with a datetime field."""

    dts: list[T_SerializedDatetime]


class MyModelNew(BaseModel):
    """A model with a datetime field."""

    dts: list[dt.datetime]

    model_config = {
        "ser_json_temporal": "milliseconds"
    }

instance_old = MyModelOld(dts=dts)
instance_new = MyModelNew(dts=dts)

t = time.time()
for _ in tqdm.tqdm(range(RANGE)):
    instance_old.model_dump_json()
print(time.time() - t)

t = time.time()
for _ in tqdm.tqdm(range(RANGE)):
    instance_new.model_dump_json()
print(time.time() - t)

I've run this over a few different number of dts to try gauge performance. A note is that im not 100% sure how this is being built for me when im installing, i doubt its using PGO and not sure if its the dev or prod profile, so i'd guess performance will improve from these results.

For the second conversion:

and for the millisecond conversion:

So for us, this is a great improvement, esspecially for when we need to return large timeseries.

NOTE: currently based against a branch of pydantic-core, because theres no pydantic core release yet.

Checklist

• The pull request title is a good summary of the changes - it will be used in the changelog
• [] Unit tests for the changes exist
• Tests pass on CI
• Documentation reflects the changes where applicable
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers