PATCH for non-resource associations doesn't work as expected [DATAREST-931]

[spring-projects-issues]

Destan Sarpkaya opened DATAREST-931 and commented

User has a repository and Phone has not. [id and some other fields are pruned in below examples for brevity, see full entities in example code]

@Entity
public class User {

        private String name;

	@Column(nullable = false)
	@Temporal(value = TemporalType.TIMESTAMP)
	private Calendar creationDate;

	@OneToMany(mappedBy = "user", cascade = CascadeType.ALL)
	private List<Phone> phones;
}

@Entity
public class Phone extends BaseEntity {

	@Column(nullable = false)
	@Temporal(value = TemporalType.TIMESTAMP)
	private Calendar creationDate;

	@ManyToOne(fetch = FetchType.EAGER)
	private User user;
}

In this case patch for /users/1 with following data works:

{
   "name":"John Doe X",
   "phones":[
      {
         "id":1,
         "creationDate":"2016-11-02T07:34:02.806+0000",
         "label":"work",
         "number":"00000000"
      }
   ]
}

but following data fails with not-null-constraint error for creationDate:

{
   "name":"John Doe X",
   "phones":[
      {
         "id":1,
         "label":"work",
         "number":"00000000"
      }
   ]
}

PATCH works selectively on resources however doesn't work as expected for inline data which is not a resource (in this case "phone")

---

Affects: 2.5.4 (Hopper SR4)

Attachments:

• person-repo.zip (6.55 kB)

Backported to: 2.5.5 (Hopper SR5), 2.4.7 (Gosling SR7)

[spring-projects-issues]

Oliver Drotbohm commented

Thanks for filing this. As you've already spent the time to create a sample project, any chance you add a test case that shows the erroneous behavior?

[spring-projects-issues]

Oliver Drotbohm commented

Nevermind, I got it. Looks like we need to apply some of our massaging to collections, too. Generally speaking, using a proper media type like JSON Patch will prevent problems like this as you can be much more specific about what your intentions are as a client

[spring-projects-issues]

Oliver Drotbohm commented

That should be fixed with the solution explained above. We now inspect arrays, too