[hibernate-issues] [Hibernate-JIRA] Commented: (HV-421) Reconsider behavior of parameter validation for inheritance hierarchies

[Gunnar Morling (JIRA)]

    [ http://opensource.atlassian.com/projects/hibernate/browse/HV-421?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=39879#action_39879 ] 

Gunnar Morling commented on HV-421:
-----------------------------------

I think we even should raise a ConstraintDefinitionException if preconditions for the same method are defined at multiple places in the hierarchy. That way a user clearly sees what the problem is instead of wondering why parts of the preconditions are silently ignored. So this situation would cause an exception:

{code:java}
public class A {
 public void foo(@NotNull String s) { ... } 
}


public class B extends A {
  public void foo(@Size(min=5) String s) { ... } 
}
{code}

This also includes the case where the highest version of a method has no constraints at all, as this already represents a precondition (a very weak one, though) which must not be strengthened in sub-types:

{code:java}
public class A {
  public void foo(String s) { ... } 
}


public class B extends A {
  public void foo(@NotNull String s) { ... } 
}
{code}

Finally no preconditions at all may be used when multiple interface inheritance/implementation comes into play:

{code:java}
public interface A {
  public void foo(String s);
}


public interface B {
  public void foo(@NotNull String s);
}

public class C implements A, B {
  public void foo(String s) { ... } 
}
{code}

If this was allowed, a caller of foo() on a variable with static type A and runtime type C would have to fulfill the preconditions from B, meaning a strenghtening of the precondition from A.

I think these restrictions actually seem worse than they are in reality. Implementing multiple interfaces with the same method should happen very rarely and defining the precondition at the up-most version of a method should not be a big problem either. WDYT?

> Reconsider behavior of parameter validation for inheritance hierarchies
> -----------------------------------------------------------------------
>
>                 Key: HV-421
>                 URL: http://opensource.atlassian.com/projects/hibernate/browse/HV-421
>             Project: Hibernate Validator
>          Issue Type: Bug
>          Components: engine
>            Reporter: Gunnar Morling
>            Assignee: Gunnar Morling
>             Fix For: 4.2.0.Beta2
>
>
> Let A extend B and A#foo() override B#foo(). When validating an invocation of A#foo() the current implementation will evaluate all parameter constraints defined at A#foo() *and* B#foo(). That way foo()'s preconditions defined in B are strengthened by A. 
> According to the ["Programming by contract"|http://en.wikipedia.org/wiki/Programming_by_contract] article on WP this is not allowed, subtypes may only weaken preconditions defined by supertypes. The common implementation pattern for this is to combine the preconditions within a hierarchy by a logical OR, meaning the weakest precondition in the hierarchy applies.
> Note that postconditions (return value constraints) may be strengthened (but not weakened) by subtypes. Therefore the current implementation (AND combination) should be correct here.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://opensource.atlassian.com/projects/hibernate/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira