Raise warnings when deprecated fields are filled at model instantiation

[changhc]

Change Summary

Raise warnings when deprecated fields receive values at model instantiation. My previous attempt at pydantic/pydantic#10865 was rejected because it led to performance issues, so I'm hoping that by moving the logic to pydantic-core will work better. At least doing it here means that the code won't need to loop over the input values one more time.

One extra config is added to the model-field schema: deprecation_msg (Option<String>). A field is marked as deprecated when this config value is present. The actual deprecation message will be passed in from pydantic; this keeps handling the type of the deprecation message simple, and most importantly, there is already logic over there dealing with possible cases of deprecation messages and thus we should not duplicate the logic here.

Also tested with pydantic and it worked.

Related issue number

Part of pydantic/pydantic#8922

Checklist

• Unit tests for the changes exist
• Documentation reflects the changes where applicable
• Pydantic tests pass with this pydantic-core (except for expected changes)
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

Selected Reviewer: @sydney-runkle

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.61%. Comparing base (ab503cb) to head (0c07f9a).
Report is 235 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1551      +/-   ##
==========================================
- Coverage   90.21%   89.61%   -0.61%     
==========================================
  Files         106      112       +6     
  Lines       16339    17895    +1556     
  Branches       36       40       +4     
==========================================
+ Hits        14740    16036    +1296     
- Misses       1592     1839     +247     
- Partials        7       20      +13     

Files with missing lines	Coverage Δ
python/pydantic_core/core_schema.py	94.84% <100.00%> (+0.07%)	⬆️
src/validators/model_fields.rs	95.17% <100.00%> (-1.31%)	⬇️

... and 54 files with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6472887...0c07f9a. Read the comment docs.

---- 🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests
• JS Bundle Analysis - Avoid shipping oversized bundles

[codspeed-hq]

CodSpeed Performance Report

Merging #1551 will not alter performance

_{Comparing changhc:8922-init-deprecated-field (8e85a60) with main (4477692)}

Summary

✅ 155 untouched benchmarks

[changhc]

Please review.

Meanwhile, I'd like to ask the reviewer's opinion on how we can make this thing "less slow". Since this implementation calls import, it will surely affect performance. Codspeed pointed out performance regression with my first commit because there I called import once and for all fields even when there were no deprecated fields. Now that import is moved so that it is called only when a warning needs to be raised. Performance looks fine according to the report, but this is because we don't have any benchmark for deprecated fields.

Since we know that performance will definitely be impacted, we can try to mitigate this. For instance, instead of calling import each time a deprecated field is filled, we can collect warnings and raise them all together in one warning at the end. There are some issues with this approach though. First, the final warning message will need to be more specific because all deprecation messages will be mixed up. We will need to indicate which fields are deprecated and which message belongs to which field for clarity, but this will be inconsistent with what is implemented in pydantic now, i.e. print the deprecation warning only without any additional information. Maybe we can just implement that here and change pydantic later on.

Another issue is when that final warning should be raised. Currently ModelFieldsValidator operates in a "fail early" fashion, i.e. when any field value fails validation, the whole model validation is terminated. Even if we raise the deprecation warning right before any error return, we won't be able to mention all deprecated fields unless we iterate through all fields separately first, but this will probably lead to performance issues. If we don't collect all deprecation warning but simply raise whatever we have observed, the user experience might not be good because users will be like "finding out the remaining "undetected" deprecation fields after fixing validation issues" instead of being able to see all issues upfront and fixing them in one go. This issue already exists with the current implementation though.