skip recursion checks for non-recursive definitions #989
Conversation
| state.recursion_guard.decr_depth(); | ||
| output | ||
| let definition = state.definitions.get(self.validator_id).unwrap(); | ||
| if definition.recursive { |
There was a problem hiding this comment.
If I comment out this block I get a segfault in the test suite which gives me confidence the recursion detection is working correctly.
| let value_id = extra.rec_guard.add(value, self.serializer_id).map_err(py_err_se_err)?; | ||
| let comb_serializer = extra.definitions.get(self.serializer_id).unwrap(); | ||
| let r = comb_serializer.serde_serialize(value, serializer, include, exclude, extra); | ||
| let definition = extra.definitions.get(self.serializer_id).unwrap(); | ||
| let r = definition | ||
| .value | ||
| .serde_serialize(value, serializer, include, exclude, extra); | ||
| extra.rec_guard.pop(value_id, self.serializer_id); |
There was a problem hiding this comment.
I wonder if we can / should skip recursion on serializing too?
| let value_id = extra.rec_guard.add(value, self.serializer_id)?; | ||
| let comb_serializer = extra.definitions.get(self.serializer_id).unwrap(); | ||
| let r = comb_serializer.to_python(value, include, exclude, extra); | ||
| let definition = extra.definitions.get(self.serializer_id).unwrap(); | ||
| let r = definition.value.to_python(value, include, exclude, extra); | ||
| extra.rec_guard.pop(value_id, self.serializer_id); |
There was a problem hiding this comment.
Similar question about serializer recursion.
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #989 +/- ##
==========================================
- Coverage 92.99% 92.98% -0.01%
==========================================
Files 106 106
Lines 15844 15881 +37
Branches 35 35
==========================================
+ Hits 14734 14767 +33
- Misses 1103 1107 +4
Partials 7 7
Continue to review full report in Codecov by Sentry.
|
CodSpeed Performance ReportMerging #989 will improve performances by 52.98%Comparing Summary
Benchmarks breakdown
|
|
This unfortunately doesn't work: from typing import Optional
from pydantic import BaseModel
from pydantic_core import CoreSchema, SchemaValidator
class ModelA(BaseModel):
b: 'Optional[ModelB]'
class ModelB(BaseModel):
a: Optional[ModelA]
cyclic_data = {}
cyclic_data['a'] = {'b': cyclic_data}
schema: CoreSchema = {
'type': 'definitions',
'schema': {
'type': 'definition-ref',
'schema_ref': '__main__.ModelB:5168279104',
},
'definitions': [
{
'type': 'model',
'cls': ModelA,
'schema': {
'type': 'model-fields',
'fields': {
'b': {
'type': 'model-field',
'schema': {
'type': 'nullable',
'schema': {
'type': 'definition-ref',
'schema_ref': '__main__.ModelB:5168279104'
}
}
}
},
'model_name': 'ModelA'
},
'ref': '__main__.ModelA:5168251248'
},
{
'type': 'model',
'cls': ModelB,
'schema': {
'type': 'model-fields',
'fields': {
'a': {
'type': 'model-field',
'schema': {
'type': 'nullable',
'schema': {
'type': 'definition-ref',
'schema_ref': '__main__.ModelA:5168251248'
}
}
}
},
'model_name': 'ModelB'
},
'ref': '__main__.ModelB:5168279104'
}
]
}
validator = SchemaValidator(schema)
validator.validate_python(cyclic_data) # segfault |
|
Right yes that's a more complex case; I think we could analyze all the dependencies of definitions in a similar fashion to how this detects recursion, to detect cycles. Will probably follow up next week. |
|
That's already what we're doing in Python. If we have to do that here there's not much of a win. |
Change Summary
Detects recursion at the
pydantic-corelevel, so thatpydanticdoesn't have to, enabling low-level performance wins without needing to inline core schemas.Related issue number
Related to pydantic/pydantic#7611
Checklist
pydantic-core(except for expected changes)