Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
pydantic
json_schema.py
Project Total

This report was generated by python-coverage-comment-action

CodSpeed Performance Report

Merging #11420 will not alter performance

_{Comparing Jack70248:decimal_constraint (88da428) with main (10af6a8)}

Summary

✅ 46 untouched benchmarks

please review - thanks!

Apologies, I've spotted a couple of outstanding issues. I'll pull this PR back to draft until I can push an update:

• The lookahead is redundant when both max_digits and decimal_places are set
• The lookahead isn't currently working with the requirement for arbitrary leading zeroes in validation This was actually fine

I'll also add in support for arbitrary trailing zeroes in validation, on the basis that Pydantic successfully validates 1111.000 with a constraint of 'max_digits':4

This should be ready for review now, thanks @sydney-runkle

Thanks and sorry for the delayed response. We are almost here. One issue I noticed is that during serialization, we don't make use of the normalized version:

import re
from decimal import Decimal
from typing import Annotated

from pydantic import TypeAdapter, Field


ta = TypeAdapter(Annotated[Decimal, Field(max_digits=4)])

reg = ta.json_schema(mode='serialization')['pattern']

val = ta.validate_python("0000001.001000")
print(val)
#> 1.001000
dumped = ta.dump_python(val, mode='json')
print(dumped)
#> 1.001000

re.compile(reg).match(dumped)  # None

Thanks and sorry for the delayed response. We are almost here. One issue I noticed is that during serialization, we don't make use of the normalized version:

import re
from decimal import Decimal
from typing import Annotated

from pydantic import TypeAdapter, Field


ta = TypeAdapter(Annotated[Decimal, Field(max_digits=4)])

reg = ta.json_schema(mode='serialization')['pattern']

val = ta.validate_python("0000001.001000")
print(val)
#> 1.001000
dumped = ta.dump_python(val, mode='json')
print(dumped)
#> 1.001000

re.compile(reg).match(dumped)  # None

Thanks @Viicos , and no problem on the delay. I've got a few things on for the next week but will be able to push an update later next week.

Hi @Viicos

I might need a hand to wrap my head around the solution to this problem. It sounds like we want the output of the TypeAdaptors dump_python() method (and BaseModel's dump_model() method?) to return the normalised Decimal form: str(Decimal("1.001000").normalize()) -> "1.001", so that it correctly matches against the serialization regex pattern.

It looks like the default serializers for this come from pydantic_core, but there's mechanisms to overwrite these through defining custom serializers such as a PlainSerializer or using something like core_schema.plain_serializer_function_ser_schema(), but it's not clear to me the best way to overwrite the TypeAdaptors/BaseModels serializer with a custom serializer, or if this should instead be changed in pydantic-core?

Let me know if I've misunderstood anything, thanks!

I might need a hand to wrap my head around the solution to this problem. It sounds like we want the output of the TypeAdaptors dump_python() method (and BaseModel's dump_model() method?) to return the normalised Decimal form: str(Decimal("1.001000").normalize()) -> "1.001", so that it correctly matches against the serialization regex pattern.

We could, but this would be a breaking change so it can only be considered for v3.

Handled in #11420.