Managing relationship creation and deletion in a REST API - rest

We are building a REST API with the following resources: Users, UserGroups. Currently we are supporting the following URI's:
/BASEAPI/VERSION/Users/
/BASEAPI/VERSION/Users/{id}/UserGroups
/BASEAPI/VERSION/UserGroups/
/BASEAPI/VERSION/UserGroups/{id}/Users
I like this better than including references in the objects which then have to be pulled on subsequent requests. It also means that we can avoid query params to filter the results. i.e. we don't have to support:
/BASEAPI/VERSION/UserGroups/{id}?user_id={user_id}
The problem is that it doesn't make creation and deletion semantics very clear. i.e. should a DELETE request to:
/BASEAPI/VERSION/Users/{id}/UserGroups/{group_id}
remove the UserGroup, or remove the user from the user group?
We've considered adding:
/BASEAPI/VERSION/UserGroupUsers
But something doesn't quite feel right about that, but maybe it's the best way to go. What do others think are best practices?

You need to figure out how you intend to represent the membership relationship between user and user group. It can be an attribute of the user, an attribute of the group, or a separate resource. Those are the only choices. How users are added to and removed from groups falls out naturally from your choice. Membership management becomes a PUT/DELETE to the user, the group, or the membership resource.
Personally, I find the separate resource to be the cleanest way to handle the issue, but you then need query parameters to poll for a specific user or group. Also, you'd need to change your second-level resource names, because it makes no sense for /userGroups/{id}/users to return a collection of userGroupUsers resources.

A URL addresses a resource. A GET on this URL returns the resource and a DELETE deletes it. If the DELETE would delete something different than the GET is returning something really is broken.
So if /BASEAPI/VERSION/Users/4711/UserGroups would return the UserGroups with the ID 0815 and 0816 the DELETE should delete both userGroups.
Question is: Does this make sense? What is happening to the other users in both userGroups?
If you want to remove a user from a group I would provide a PATCH Method.

Related

How to design REST API with one action but two different meanings?

Given the example of a shop API with an orders resource. You would want to delete one order by id
DELETE /orders/:orderId
Under the hood you run an update query and set canceled to true. But what if
A customer calls this endpoint:
You need a canceledByCustomer database flag
No additional permissions are required
An administrator calls this endpoint?
You need a rejectedByAdministrator database flag
Additional permissions are required
Would you keep the endpoint posted above and check internally, if the calling user tries to cancel the order of another user and if true, this is a reject action?
Would you add two query parameters cancel and reject and one of them MUST be true and one of them MUST be null/false?
Would you violate the design rules, create two different endpoints and add verbs to them like so?
DELETE /orders/:orderId/cancel => customer can call it
DELETE /orders/:orderId/reject => only administrators can call it
Does someone know about the best practises for such "domain driven" problems?
API endpoints don't have to correlate on what happens closer to the core, for example in your Aggregate Root or CommandHandler. In my opinion, make the API routes as verbose as possible, which means creating their own separate routes for each use case. Push the logic on what database flag to use (canceledByCustomer vs rejectedByAdministrator) closer down to the entity.

How to properly access children by filtering parents in a single REST API call

I'm rewriting an API to be more RESTful, but I'm struggling with a design issue. I'll explain the situation first and then my question.
SITUATION:
I have two sets resources users and items. Each user has a list of item, so the resource path would like something like this:
api/v1/users/{userId}/items
Also each user has an isPrimary property, but only one user can be primary at a time. This means that if I want to get the primary user you'd do something like this:
api/v1/users?isPrimary=true
This should return a single "primary" user.
I have client of my API that wants to get the items of the primary user, but can't make two API calls (one to get the primary user and the second to get the items of the user, using the userId). Instead the client would like to make a single API call.
QUESTION:
How should I got about designing an API that fetches the items of a single user in only one API call when all the client has is the isPrimary query parameter for the user?
MY THOUGHTS:
I think I have a some options:
Option 1) api/v1/users?isPrimary=true will return the list of items along with the user data.
I don't like this one, because I have other API clients that call api/v1/users or api/v1/users?isPrimary=true to only get and parse through user data NOT item data. A user can have thousands of items, so returning those items every time would be taxing on both the client and the service.
Option 2) api/v1/users/items?isPrimary=true
I also don't like this because it's ugly and not really RESTful since there is not {userId} in the path and isPrimary isn't a property of items.
Option 3) api/v1/users?isPrimary=true&isShowingItems=true
This is like the first one, but I use another query parameter to flag whether or not to show the items belonging to the user in the response. The problem is that the query parameter is misleading because there is no isShowingItems property associated with a user.
Any help that you all could provide will be greatly appreciated. Thanks in advance.
There's no real standard solution for this, and all of your solutions are in my mind valid. So my answer will be a bit subjective.
Have you looked at HAL for your API format? HAL has a standard way to embed data from one resources into another (using _embedded) and it sounds like a pretty valid use-case for this.
The server can decide whether to embed the items based on a number of criteria, but one cheap solution might be to just add a query parameter like ?embed=items
Even if you don't use HAL, conceptually you could still copy this behavior similarly. Or maybe you only use _embedded. At least it's re-using an existing idea over building something new.
Aside from that practical solution, there is nothing in un-RESTful about exposing data at multiple endpoints. So if you created a resource like:
/v1/primary-user-with-items
Then this might be ugly and inconsistent with the rest of your API, but not inherently
'not RESTful' (sorry for the double negative).
You could include a List<User.Fieldset> parameter called fieldsets, and then include things if they are specified in fieldsets. This has the benefit that you can reuse the pattern by adding fieldsets onto any object in your API that has fields you might wish to include.
api/v1/users?isPrimary=true&fieldsets=items

REST Design: What Http verb should be used to retrieve a dynamic resource?

I have a scenario in which I have REST API which manages a Resource which we will call Group. A Group contains members and the group resource is dynamic - whenever you retrieve it, you get the latest data (so a query must run server side to update the number of members in a group - in other words, the result of the request is to modify the data, since the results of running the query are stored).
Given a *group_id* it should return a minimal amount of information like
{
group_id: "5t7yu8i9io0op",
group_name: "That's my name",
size: 34
}
So a GET to this resource causes the resource to change, since a subsequent GET could return a new value for 'size'. This tells me it is not idempotent and so you should use POST to retrieve this resource. Am I correct in this conclusion?
If I am correct, do you think it is advisable to also provide a GET method that only returns the currently stored data for the group (eg. so the size could be out of date, even the name too). I suppose in this case I should return a last-modified date as one of the fields so that the user knows how up-to-date the resource is and can then elect to use the POST method...but then I am left wondering why would anyone do that, so why not ONLY provide the POST method and forget about GET?
Confused I am!
Thanks in advance.
[EDIT]
#Satish posted a link in his/her answer to the HTTP specs. In section 9.1.1. it ends with this sentence:
Naturally, it is not possible to ensure that the server does not generate side-effects as a result of performing a GET request; in fact, some dynamic resources consider that a feature. The important distinction here is that the user did not request the side-effects, so therefore cannot be held accountable for them.
So in my scenario, the requester does not really care about the side-effect that the value for 'size' is recomputed as a direct result of making the request. They want the group information and it just so happens that to provide accurate, up-to-date group data, the size query must be run in order to update that value. Whilst making the request causes data to change implies this should be a POST, the user did not request that side-effect and so therefore a GET request would be acceptable and more intuitive, would it not? And therefore still be restful according to this sentence.
[2nd EDIT]
#Satish asks a very important question in the comments. So for others who read this I'll explain further about this problem:
Normally you would not run the group query to update its size from a REST request. As members are added or removed from a group, you would update the computed size of that group, store it and then a simple GET request would always return the correct size. However, our situation is more complicated in that a group is only stored as a query definition in ElasticSearch (kind of like a view in an RDBMS). Members do not get added/removed to and from groups. They get added to a much larger set of data (a collection in MongoDB). There are hundreds, potentially thousands, of different 'group definitions' so it is not practical to recompute size for every group when the collection changes. We cannot know when an item is added/removed to/from the collection which groups might change size - you only know by running the group definition who is in that group and what the size is. I hope that clears things up. :)
You should use GET. Even if dynamic resource is changing, you did not request for that change through your request and you are not accountable for that change. Ref: http://www.w3.org/Protocols/rfc2616/rfc2616-sec9.html
In your case when you do a GET you retrieve some information about the Group. You don't modify the group structure. Ok, the group can be changed by an external entity so your next GET may bring you another data. Am I right? Who modifies the structure of the group and when?
So you should use GETbecause the resource it will be modified from somewhere else and not by your call that tries to do a read operation.
EDIT
After your edited the question I just want to add that I agree about the side effects.
It matters if you sent data or a change command explicitly to the server or you just read something and you don't have to pay attention for what the server side is doing to gave you the response. More intuitively:
GET - Requests data from a specified resource
POST - Submits data to be processed to a specified resource
It is the combination of GET and POST. So you should use POST.
Refer : http://adarshdchaurasia.wordpress.com/2013/09/26/http-get-vs-post/
You should not use GET because if you use GET method then search engines may cache the responses. It may cause unintentional data update at your server side, which you do not want. GET method is meant to return content without updating anything on server. POST is meant to updated the things at server and return result against that operation.

RESTful URI to select "My" items

I want to expose my Item resource in these ways:
GET /Item (to show all Items)
GET /Item/401 (to show only the Item with key value 401)
So far so good. Now to request only the Items that belong to Me, I can think of these possible URIs constructions:
GET /Item/Mine
GET /MyItem
GET /Item/owner=42 (where my ID is 42)
I don't like the last one because it simplifies my client programming to rely on existing authentication to figure out "My" ID. I happen to be coding on ASP.NET WebAPI, but I'm really asking about the principle and design, not the implementation.
My question: Which of the three ways to represent "My" Items is most consistent with the RESTful design concept? Is it even legitimate to ask for "My" Items as such?
If users should not be able to see items that aren't "theirs", then this is an authorization issue. Just use GET /Item, and have your authorization module strip out the items that aren't theirs.
If users can see all items, and you just want to grab the items that belong to a particular user, then you should be parameterizing the call to GET /Item. Something like GET /Item/?owner=42. In this case, the user isn't limited to asking for their items .. they can ask for the items of any owner. That buys them flexibility without costing you anything.
If you have a /User resource, you could also add an endpoint /User/{id}/Item, which
returns the same collection as /Item/?owner=
And this isn't germane, but I find that plural collections (/Items, /Users) read a lot better than singulars. YMMV.
I would use GET /Items to get all itmes.
And GET /Items?owner=42.
GET /Item/Mine is not consistent since it contradicts your way of identifying the resource ie, GET /Item/401.
GET /MyItem seems seems unnecessary since the resources you are looking for can be retrieved using the other REST url paths.

RESTful API creates a globally unique resource

In our system we have accounts which contain items. An item is always associated with a single account but also has a globally unique id in the system. Sometimes it is desirable to work with an item when only its id is known.
Is it incorrect to allow access to a subordinate resource (the item) from outside it's owner (the account)? In other words, is it wrong to have 2 URI's to the same resource? This is a little tricky to explain so here is an example:
POST /inventory/accountId
#Request Body contains new item
#Response body contains new item's id
GET|PUT|DELETE /inventory/accountId/guid #obviously works and makes sense
GET|PUT|DELETE /inventory/guid #does this make sense?
Perhaps I should rethink my resource layout and not use accounts to create items but instead take the account as a query string parameter or field on the item?
POST /inventory
# Request body contains item w/ account name set on it
GET|POST|DELETE /inventory/uuid #makes sense
GET|POST|DELETE /inventory/accountId/uuid #not allowed
I think having two URIs point to the same item is asking for trouble. In my experience, these sorts of things lead to craziness as you scale out (caching, multiple nodes in a cluster going out of sync and so on). As long as the item's ID is indeed globally unique, there's no reason no to simply refer to it as /inventory/uid
POST /inventory/accountId
GET|PUT|DELETE /inventory/accountId/guid #obviously works and makes sense
GET|PUT|DELETE /inventory/guid #does this make sense?
It makes the most sense when /inventory/guid redirects to /inventory/accountId/guid (or, I'd argue, vice-versa). Having a single canonical entity, with multiple URI's redirecting to it, allows your caching scheme to remain the most straightforward. If the two URI's instead return the same data, then a user is inevitably going to PUT a new representation to one and then be confused when it GETs an old copy from the other because the cache was only invalidated for the former. A similar problem can occur for subsequent GETs on the two. Redirects keep that a lot cleaner (not perfectly synchronous, but cleaner).
Whether to make items subordinate to accounts depends on whether items can exist without an account. If the data of an item is a subset of the data of an account, then go ahead and make it subordinate. If you find that an account is just one kind of container, or that some items exist without any container, then promote them to the top level.
In other words, is it wrong to have 2 URI's to the same resource?
No. It is not wrong to have multiple URI's identifying the same resource. I don't see anything wrong with your first approach as well. Remember URI's are unique identifiers and should be opaque to clients. If they are uniquely identifying a resource then you don't have to worry too much about making your URLs look pretty. I am not saying resource modeling is not important but IMO we shouldn't spend too much time on it. If your business needs that you have guid directly under inventory and also under individual accounts, so be it.
Are you concerned about this because of a potential security hole in letting data be available to unauthorized users? Or is your concern purely design driven?
If security is not your concern, I agree that it is perfectly fine to have 2 URIS pointing to the same resource.