Specification validation as part of shortcuts

[p1c2u]

No description provided.

[codecov]

Codecov Report

Merging #686 (1752d35) into master (0abae09) will decrease coverage by 0.08%.
The diff coverage is 86.66%.

@@            Coverage Diff             @@
##           master     #686      +/-   ##
==========================================
- Coverage   91.22%   91.15%   -0.08%     
==========================================
  Files         126      126              
  Lines        3580     3640      +60     
  Branches      429      430       +1     
==========================================
+ Hits         3266     3318      +52     
- Misses        265      273       +8     
  Partials       49       49              

Files	Coverage Δ
openapi_core/shortcuts.py	100.00% <100.00%> (ø)
...penapi_core/unmarshalling/request/unmarshallers.py	79.67% <100.00%> (+0.10%)	⬆️
openapi_core/unmarshalling/unmarshallers.py	85.71% <100.00%> (+0.34%)	⬆️
openapi_core/validation/request/validators.py	87.50% <100.00%> (+0.70%)	⬆️
openapi_core/validation/response/validators.py	79.89% <100.00%> (+1.17%)	⬆️
openapi_core/validation/validators.py	95.27% <100.00%> (+0.27%)	⬆️
openapi_core/unmarshalling/request/protocols.py	73.91% <50.00%> (-5.04%)	⬇️
openapi_core/unmarshalling/response/protocols.py	76.92% <50.00%> (-4.90%)	⬇️
openapi_core/validation/request/protocols.py	70.37% <50.00%> (-3.55%)	⬇️
openapi_core/validation/response/protocols.py	71.42% <50.00%> (-3.58%)	⬇️

[andersk]

This seems wasteful since Spec.from_dict has already validated the spec. This now validates the spec three times, once in Spec.from_dict, once in validate_request, and once in validate_response:

from openapi_core import Spec, validate_request, validate_response
from openapi_core.testing import MockRequest, MockResponse

spec = Spec.from_dict(
    {
        "openapi": "3.1.0",
        "info": {"title": "Test", "version": "1.0.0"},
        "paths": {"/": {"get": {"responses": {"200": {"description": "OK"}}}}},
    },
)
request = MockRequest("http://localhost", "get", "/")
validate_request(request, spec=spec)
response = MockResponse("")
validate_response(request, response, spec=spec)

This adds a very expensive per-request and per-response overhead when the spec is large.

Moreover, the documented workaround doesn’t work; this still validates the spec (as can be seen by misspelling title, say):

from openapi_core import Spec, validate_request, validate_response
from openapi_core.testing import MockRequest, MockResponse

spec = Spec.from_dict(
    {
        "openapi": "3.1.0",
        "info": {"titel": "Test", "version": "1.0.0"},
        "paths": {"/": {"get": {"responses": {"200": {"description": "OK"}}}}},
    },
    validator=None,
)
request = MockRequest("http://localhost", "get", "/")
validate_request(request, spec=spec, spec_validator_cls=None)
response = MockResponse("")
validate_response(request, response, spec=spec, spec_validator_cls=None)

[p1c2u]

Hi @andersk

Guess I need to rethink the change. Thanks for your feedback.