Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Feature/11 combine profile update and fi association #38

Merged
merged 15 commits into from
Oct 5, 2023

Conversation

guffee23
Copy link
Contributor

Closes #11

Added UserProfile dto. Abstracted lei association method. Added lei association to admin update user api.
Updated json for all related calls.
Changed update test to incorporate association testing.
Added test where leis not present in call.
Updated associate api test.
@github-actions
Copy link

github-actions bot commented Sep 26, 2023

Coverage report

The coverage rate went from 84.32% to 83.91% ⬇️
The branch rate is 84%.

90% of new lines are covered.

Diff Coverage details (click to unfold)

src/entities/models/init.py

100% of new lines are covered (100% of the complete file).

src/entities/models/dto.py

100% of new lines are covered (100% of the complete file).

src/oauth2/oauth2_admin.py

50% of new lines are covered (33.33% of the complete file).
Missing lines: 92, 93

src/routers/admin.py

100% of new lines are covered (100% of the complete file).

Comment on lines 44 to 45
firstName: str
lastName: str
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The body was to look like:
{ "firstName": "First", "lastName": "Last", "leis": ["LEI1", "LEI2"] }
So, we need aliasing for the request as well as when generating the dict?
To match the request reqs and the syntax that keycloak expects, we would need it in the form above, but to change to snake case would require aliasing to snake when request is sent and when creating dict being sent to keycloak.
Just double-checking here. Should the request body be changed to snake case?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

request body to the api should be snake case, we would then alias it to camel case for keycloak.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just so I understand, our API's User resource will look like:

{
  "first_name":"Jane",
  "last_name":"Doe",
  "leis": ["123456789012","234567890123"],
}

Meanwhile, the Keycloak Admin API's UserRepresentation expects:

{
  "firstName": "Jane",
  "lastName": "Doe"
}

...and to associated LEIs to said User, it's reeeaaally associating a Keycloak User to a Keycloak Group which is named after a given LEI? So, to complete the transaction, and second call that's made to...

  • PUT /admin/realms/{realm}/users/{id}/groups/{groupId}

If yes, 👍. If no, I may need some help here. 😅


I think what @lchen-2101 is getting at is that we shouldn't couple our API format to Keycloak's, which I agree with.

As to Pydantic's to_camel, I'm not sure we want to do that. There's only 2 attributes we care about. Aaaand, what happens to leis on our model when you do the to_camel? Wouldn't that still be there in the data we'd be submitting to Keycloak? Seems like that might break.

To me, it doesn't seem that gross to just construct the dict at time of request in update_me(). 🤷 Looks like we're already doing it that way in upsert_group now anyway.

@@ -88,5 +88,9 @@ def associate_to_lei(self, user_id: str, lei: str) -> None:
detail="No institution found for given LEI",
)

def associate_to_lei_set(self, user_id, leis: Set[str]):
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

normally if we can, we'd like to keep the implementation detail of the collection out of the var name / method name; this one should be good with just associate_to_leis

Comment on lines 48 to 49
def get_user(self) -> Dict[str, Any]:
return {"firstName": self.firstName, "lastName": self.lastName}
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

pydantic should have something built-in that will generate the dict for you, I don't think you need to create a separate method for this.

@@ -19,12 +20,13 @@ async def get_me(request: Request):

@router.put("/me/", status_code=HTTPStatus.ACCEPTED, dependencies=[Depends(check_domain)])
@requires("manage-account")
async def update_me(request: Request, user: Dict[str, Any]):
oauth2_admin.update_user(request.user.id, user)
async def update_me(request: Request, user: UserProfile):
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should API functions that use the Keycloak client be async? It looks like python-keycloak uses requests under the hood, which is not async. I think they'd need to use something like httpx to support that. As it stands now, I think this could lead to unintentional blocking during these calls.

Dropping async is not a bad thing. FastAPI will just run these requests in a new thread, which I think is what we'd want for synchronous processes like this.

Comment on lines 44 to 45
firstName: str
lastName: str
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just so I understand, our API's User resource will look like:

{
  "first_name":"Jane",
  "last_name":"Doe",
  "leis": ["123456789012","234567890123"],
}

Meanwhile, the Keycloak Admin API's UserRepresentation expects:

{
  "firstName": "Jane",
  "lastName": "Doe"
}

...and to associated LEIs to said User, it's reeeaaally associating a Keycloak User to a Keycloak Group which is named after a given LEI? So, to complete the transaction, and second call that's made to...

  • PUT /admin/realms/{realm}/users/{id}/groups/{groupId}

If yes, 👍. If no, I may need some help here. 😅


I think what @lchen-2101 is getting at is that we shouldn't couple our API format to Keycloak's, which I agree with.

As to Pydantic's to_camel, I'm not sure we want to do that. There's only 2 attributes we care about. Aaaand, what happens to leis on our model when you do the to_camel? Wouldn't that still be there in the data we'd be submitting to Keycloak? Seems like that might break.

To me, it doesn't seem that gross to just construct the dict at time of request in update_me(). 🤷 Looks like we're already doing it that way in upsert_group now anyway.

Updated dto to snake case.
Changed associate lei method name.
Utilized pydantic dictionaries.
Updated tests to account for method name change.
Comment on lines 44 to 50
class UserProfile(BaseModel):
first_name: str = Field(alias="firstName")
last_name: str = Field(alias="lastName")
leis: Optional[Set[str]] = Field(alias="leis")

class Config:
alias_generator = to_camel
Copy link
Collaborator

@lchen-2101 lchen-2101 Oct 3, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

found a couple of things with the configs, the to_camel generator is actually doing pascal case; there is a to_lower_camel which would achieve the goal, without you having to specify Field; there is also an allow_population_by_field_name option which would allow the request payload to be snake case instead of just camel case; while it doesn't change the swagger doc to show camel case, it does allow the API to accept both camel case and snake case. So what you end up with would be:

class UserProfile(BaseModel):
    first_name: str
    last_name: str
    leis: Optional[Set[str]]

    class Config(BaseConfig):
        alias_generator = to_lower_camel
        allow_population_by_field_name = True

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

just a side note; when we update fastapi later, the included pydantic package also changes to 2.x from 1.x, and some of these configurations will be changed. Nothing to worry about here, but in the pydantic settings issue, the dependency will be updated, so more changes will happen here as well.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we need/want to have our API accept camelCase? Is that just an artifact of us using the same UserProfile model for both our API and for submitting to Keycloak's API? To me it seems more explicit and future-proof to decouple the two, and just do a simple translation to the expected JSON format at the time we call the Keycloak API. If we go this route, we have one oddball model object with extra Config section that likely needs an explainer comment for what it's doing there, and we'll need to keep the exclude={"leis"} bit up-to-date every time you add another field UserProfile.

This is all what I was getting at when I asked #38 (comment), so it's also possible I'm just missing something.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@hkeeler It has to do with keycloak and the dictionary vals it's expecting, which are in camelCase.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I get that. My question is, why are we doing these Pydantic gymnastics when we could just...

def update_me(request: Request, user: UserProfile):
    keycloak_user = {'firstName': user.first_name, 'lastName': user.last_name}
    oauth2_admin.update_user(request.user.id, keycloak_user)
    if user.leis:
        oauth2_admin.associate_to_leis(request.user.id, user.leis)

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

my thought would be still to have a well defined model to communicate with Keycloak instead of constructing a dict on the spot. I was hoping to leverage what's built into pydantic, and having the config take care of SerDe; but if we want to do away with the configs, having a method on the model would be more preferable.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

But that model is for our API, not Keycloak's, right? It feels like we're using the fact that the two APIs happen to be similar enough in message structure (currently) that we're using the same model for both, but doing so comes with the baggage of having their fields now coupled, having to understand/explain why one model is different from the rest, and having to adjust a disconnected method to exclude future fields we add.

Now, we could have a dedicated model for calls to Keycloak. I mean really, that should be the job of python-keycloak, but that seems to just be a very a light wrapper around their API. I think that's probably more than we need for this PR, though.

For now, I agree having a to_keycloak_user() on the user model works. It's simple and explicit.

@@ -88,5 +88,9 @@ def associate_to_lei(self, user_id: str, lei: str) -> None:
detail="No institution found for given LEI",
)

def associate_to_leis(self, user_id, leis: Set[str]):
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
def associate_to_leis(self, user_id, leis: Set[str]):
def associate_to_leis(self, user_id: str, leis: Set[str]):

Comment on lines 44 to 50
class UserProfile(BaseModel):
first_name: str = Field(alias="firstName")
last_name: str = Field(alias="lastName")
leis: Optional[Set[str]] = Field(alias="leis")

class Config:
alias_generator = to_camel
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we need/want to have our API accept camelCase? Is that just an artifact of us using the same UserProfile model for both our API and for submitting to Keycloak's API? To me it seems more explicit and future-proof to decouple the two, and just do a simple translation to the expected JSON format at the time we call the Keycloak API. If we go this route, we have one oddball model object with extra Config section that likely needs an explainer comment for what it's doing there, and we'll need to keep the exclude={"leis"} bit up-to-date every time you add another field UserProfile.

This is all what I was getting at when I asked #38 (comment), so it's also possible I'm just missing something.

Comment on lines 51 to 54
class Config:
allow_population_by_field_name = True


Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can this be removed now too?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes

Copy link
Member

@hkeeler hkeeler left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍

Copy link
Collaborator

@lchen-2101 lchen-2101 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@lchen-2101 lchen-2101 merged commit 0c9ef87 into main Oct 5, 2023
@lchen-2101 lchen-2101 deleted the Feature/11_combine_profile_update_and_fi_association branch October 5, 2023 17:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Add endpoint that combines user profile update and user fi association
3 participants