Fix json serialization when computed field is excluded

[sydney-runkle]

Fix pydantic/pydantic#9015

Previously, we were getting the value of a computed field before we checked if it was excluded, which could lead to a recursion error in a case like that shown below.

This mirrors the behavior of the to_python logic already present, which currently works.

from __future__ import annotations

import hashlib

from pydantic import BaseModel, computed_field


class MediaFileInfo(BaseModel):
    url: str
    duration: float
    bitrate: int
    format: str
    codec: str

    @computed_field(repr=False)
    def id(self) -> str:
        str_obj = self.model_dump_json(exclude={"id"})
        hash_obj = hashlib.sha256()
        hash_obj.update(str_obj.encode("utf-8"))
        return hash_obj.hexdigest()[:12]


class VideoMediaFileInfo(MediaFileInfo):
    width: int
    height: int


class AudioMediaFileInfo(MediaFileInfo):
    language: str


if __name__ == "__main__":
    mf = AudioMediaFileInfo(
        url="https://bla.com",
        duration=123,
        bitrate=123,
        format="mp4",
        codec="mp4",
        language="fr"
    )
    print(mf.id)

This case is now fixed, and prints the id as expected! I think it makes sense to add a test case in pydantic for this 👍

Selected Reviewer: @samuelcolvin

[sydney-runkle]

Please review

[codspeed-hq]

CodSpeed Performance Report

Merging #1228 will not alter performance

_{Comparing fix_comp_field_ser_recursion (80c8f03) with main (8db3a6f)}

Summary

✅ 148 untouched benchmarks

[davidhewitt]

👍 a test would be nice!