You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Why do these specs include reference to internal null fields? No idea, but as far as I'm aware it's valid to do so, and it's inconvenient that these specs are unparseable due to this issue.
This could be fixed by changing the check to not fail on null when dereferencing the last token in a path (before the last token, it should still be an error). I think undefined should still throw as it does now, because it's not a valid value in JSON, and so should never appear explicitly in a JSON schema (again, not an expert) unlike null.
I'm not totally sure if that change would have any other consequences though. This check was introduced by #153, and my best guess is that it was intended to handle the en-route (non-final) failure case, detecting it and creating an explicit error type for that as part of the error improvements there, but I can't be 100% sure there isn't another effect.
Happy to open a quick PR to make the change, if we can assume that's true and change this accordingly.
The text was updated successfully, but these errors were encountered:
It looks like that's currently pinned to 9.0.6, "to avoid circle bug" according to APIDevTools/swagger-parser@ef57642 (from @philsturgeon), I'm not sure exactly what that means but maybe it's resolved in later releases since then? It'd be great to see this updated to propagate the latest fixes here through that package too.
When resolving a reference, if any value en route including the final value is
null
then resolution fails. This happens here:json-schema-ref-parser/lib/pointer.ts
Line 96 in a5b3946
I think (not a JSON schema expert) this is incorrect. At the very least, it breaks parsing of some specs from https://github.com./APIs-guru/openapi-directory, including:
Why do these specs include reference to internal null fields? No idea, but as far as I'm aware it's valid to do so, and it's inconvenient that these specs are unparseable due to this issue.
This could be fixed by changing the check to not fail on
null
when dereferencing the last token in a path (before the last token, it should still be an error). I thinkundefined
should still throw as it does now, because it's not a valid value in JSON, and so should never appear explicitly in a JSON schema (again, not an expert) unlike null.I'm not totally sure if that change would have any other consequences though. This check was introduced by #153, and my best guess is that it was intended to handle the en-route (non-final) failure case, detecting it and creating an explicit error type for that as part of the error improvements there, but I can't be 100% sure there isn't another effect.
Happy to open a quick PR to make the change, if we can assume that's true and change this accordingly.
The text was updated successfully, but these errors were encountered: