Linting required flag provides me some errors when should not

[MathRobin]

I tried to lint an openapi spec file and got some strange errors. Maybe I'm wrong but when i check into samples, i think i'm not.

Steps to reproduce the behavior:

1. This OpenAPI file extract

...
/log:
    post:
      summary: Add a log to applicative logs
      description: Add a log to applicative logs
      operationId: setLog
      responses:
        '204':
          description: No content
        '401':
          description: You're not authorized. Please provide your token
      requestBody:
        description: ''
        content:
          application/json:
            schema:
              type: object
              properties:
                companyId:
                  required: true
                  type: number
...

2. Run this command with these arguments... npx openapi lint openapi.yml
3. See error

[1] openapi.yml:166:29 at #/paths/~1log/post/requestBody/content/application~1json/schema/properties/companyId/required

Expected type `array` but got `boolean`.

164 | properties:
165 |   companyId:
166 |     required: true
167 |     type: number
168 |   userId:

Error was generated by the spec rule.

openapi spec: 3.0.0 openapi-cli: 1.0.0-beta.69 node: 14.18.1

Am i wrong ? Many thanks for your work, it's great to have a tool like this one !

[adamaltman]

This is incorrect:

              type: object
              properties:
                companyId:
                  required: true
                  type: number

Do this instead:

              type: object
              required:
                 - companyId
              properties:
                companyId:
                  type: number

Let's leave this issue open because the problem message is incorrect (or unhelpful).

A message like:

• required incorrectly defined
• required needs to be a sibling to properties

I'm not coming up with a good message name off-the-top.

Update: I understand why it says it is an array and not boolean, but it is still defined at the wrong level and that seems like a more important problem.

[MathRobin]

oh ok, thanks for have solved my problem !

The surprising next problem for me is that why it's different between how to define for query parameters and for the body ? Is it normal looking to spec ?

Again, many thanks, that solved my problem !

I wasn't able to find an example like your working solution. Each time i found swagger-like (sorry i know it's a competitor) with required into property, not as a sibling.

Even Insomnia didn't understand required as an array, sibling of properties. But it understands when it's a member. Strange no?

[adamaltman]

Strange? I'll leave that to the community to decide. Different between parameters and schemas? Yes!

You got it exactly:

parameters:
  filter:
     required: true
     # ...

# schemas set required in a "required" array
schemas:
  adam:
    type: object
    required:
      - filter
    properties:
       filter:
         type: string

[MathRobin]

I don't know for schemas (I don't use them, I auto generate my openapi file, it was more easy to don't use them). But it's different between query string parameters and body. For the first every param needs a nested required boolean when for the last you need a required array listing fields, as a sibling to list of parameters. Why not use the same way. I'm sure that there a lot of smart people who have work on openapi spec. If they have chosen this way, it's for good reasons

[lornajane]

Thanks for this, both of you! I don't see that there's actions to take on Redocly CLI so I'm closing the issue.