paths: fix Operation request urljoin

[olichtne]

When adding operation path to the base_url, the use of the + operator can lead to doubling of / characters if the base_url ends with a /.

Using urljoin should fix this. This is also a cleaner way to work with url paths.

This causes some minor issues for us when interacting with our rest api server where the doubled / character is not matched correctly and results in 404 errors.

Even though the doubled / characters are defined in RFCs as something that should be simply skipped and don't effectively change the path, some http servers can be configured either way.

An example of a reproducer is to use a schema that looks as follows:

openapi: 3.0.2
info:
  title: LNST Dashboard
  version: ''
  description: API to the interact with the dashboard db directly - upload or get
    and process data as you wish
servers:
  - url: https://api.example.com/
paths:
  /artifacts/:
    get:
      operationId: listArtifacts
...

vs.

openapi: 3.0.2
info:
  title: LNST Dashboard
  version: ''
  description: API to the interact with the dashboard db directly - upload or get
    and process data as you wish
servers:
  - url: https://api.example.com
paths:
  /artifacts/:
    get:
      operationId: listArtifacts
...

Note that the difference is the final / in the server url.

For the first schema with the current code of openapi3 the request api.call_listArtifacts() would call:

https://api.example.com//artifacts

Whereas with the fix the same request would call:

https://api.example.com/artifacts

[Dorthu]

This behaves unexpectedly when the servers listed include a path fragment. For example:

openapi: 3.0.2
info:
  title: LNST Dashboard
  version: ''
  description: API to the interact with the dashboard db directly - upload or get
    and process data as you wish
servers:
  - url: https://api.example.com/v2
paths:
  /artifacts/:
    get:
      operationId: listArtifacts
...

A request to the listArtifacts operation here would result in a request to https://api.example.com/artifacts/ instead of https://api.example.com/v2/artifacts/; this appears to be the behavior of urljoin:

>>> from urllib.parse import urljoin
>>> urljoin("https://api.example.com/v2", "/artifacts/")
'https://api.example.com/artifacts/'
>>> urljoin("https://api.example.com/v2/", "/artifacts/")
'https://api.example.com/artifacts/'

Maybe stripping the extra / if it's present in both the server name and the path would be a simpler solution?

[olichtne]

hmmm, good point

stripping the extra /

i feel like there should be a better way to do this, doing string editing operations on path fragments for a simple join operation feels weird to me.

i'll think about this next week when i have some time for this.

thanks for looking at the PR.