fix: handle nullable, remove duplication #60
Conversation
| field :day_of_week, type: Objects::Day do | ||
| field :day_of_week, type: Objects::Day, null: true do |
There was a problem hiding this comment.
Adding this with no other logic changes made the validation spec fail. So this should ensure we don't accidentally introduce regressions in future.
paulsturgess
force-pushed
the
fix-nullable
branch
2 times, most recently
from
245e471 to
39b24da
Compare
paulsturgess
force-pushed
the
fix-nullable
branch
from
39b24da to
1432f56
Compare
| field :my_partial_polymorph, type: Objects::MonthPolymorph, include: "number", null: true do | ||
| description "Partial polymorph" | ||
| end | ||
| field :my_partial_polymorph, type: Objects::MonthPolymorph, include: "number", null: true |
There was a problem hiding this comment.
I removed the description here to test that we still set nullable correctly even when there is no description for the field.
paulsturgess
requested review from
bencromwell
and removed request for
bencromwell
paulsturgess
force-pushed
the
fix-nullable
branch
from
56a2956 to
e5f1671
Compare
paulsturgess
changed the title
paulsturgess
changed the title
| field :time, type: Objects::Time, include: "unix,day_of_week,year[as_string]", null: true do | ||
| field :time, | ||
| type: Objects::Time, | ||
| include: "unix,day_of_week,year[*,-as_integer,-as_array_of_strings,-as_array_of_enums]", null: true do |
There was a problem hiding this comment.
I changed this to make sure we handle negative declarations in the include string
There was a problem hiding this comment.
Whilst I can't attest to the Ruby the end result is where it needs to be! Also appreciate the inclusion of OpenAPI 3.1 labelled comments.
Perhaps in a magical completionist Future where George RR Martin has finished writing his book and the tooling for OpenAPI 3.1 is there, we can use those as markers to update the generation.
Follow on from #56
There were some places we didn't correctly set nullable for refs (sibling attrs are not allowed for refs).
We can set sibling attrs on refs, if we wrap them in
allOf. This PR moves that fix into the central helper that generates most refs and we can enable it by specifyingsibling_props: true.It also fixes an issue where an apia endpoint declares the same error response (error code and http status) more than once. this can happen if say an ArgumentLookup declares an error response and then the endpoint itself does too.
the schema will no longer include endpoints in groups marked with
no_schema.Katapult has some of these for legacy endpoints. And these were also problematic as they used the same names
for a lot of api objects which were causing clashes.
the generator can also now handle when
include: true(instead of a string). this is not the correct way to use the option, but Katapult uses it by accident presumably.partial objects are now only generated when absolutely necessary. previously all child objects of a parent partial object were considered partial and this generated a unique entry in the schema. now we only generate partial child objects when absolutely necessary.
closes: #55