Eclipse Community Forums: EMF » Question about constraints vs. invariants

Home » Modeling » EMF » Question about constraints vs. invariants

Question about constraints vs. invariants [message #687031]

Mon, 23 May 2011 15:55

Eclipse User

Originally posted by:

Hi,

The EMF delegates mechanism provides two ways of validating the
EObjects, constraints and invariants. AFAIU, invariants are meant to be
small and quick checks that can happen often while constraints are meant
to implement more global checks. What I don't understand is how this has
lead to the particular implementation of these. In particular, invariant
are implemented using modelled EOperation with a specific signature. Why
has this particular implementation been chosen and how can the arguments
of the EOperation be used in the body?

Regards,

Hallvard

Re: Question about constraints vs. invariants [message #687032 is a reply to message #687031]

Mon, 23 May 2011 16:11

Eclipse User

Hi Hallvard

As validated constraints, there is little difference between Ecore
invariants and Ecore constraints., Both should be invoked during validation.

The significant difference is that an Ecore invariant is callable
whereas an Ecore constraint is not. This may be useful from a more
sophisticated environment that allows selective invocation and
re-invocation of constraints.

The arguments can only be used from Java bodies.

Regards

Ed Willink

On 23/05/2011 20:55, Hallvard Trætteberg wrote:
> Hi,
>
> The EMF delegates mechanism provides two ways of validating the
> EObjects, constraints and invariants. AFAIU, invariants are meant to
> be small and quick checks that can happen often while constraints are
> meant to implement more global checks. What I don't understand is how
> this has lead to the particular implementation of these. In
> particular, invariant are implemented using modelled EOperation with a
> specific signature. Why has this particular implementation been chosen
> and how can the arguments of the EOperation be used in the body?
>
> Regards,
>
> Hallvard

Re: Question about constraints vs. invariants [message #687033 is a reply to message #687032]

Mon, 23 May 2011 17:46

Eclipse User

Originally posted by:

On 23.05.11 22.11, Ed Willink wrote:
>
> The significant difference is that an Ecore invariant is callable
> whereas an Ecore constraint is not. This may be useful from a more
> sophisticated environment that allows selective invocation and
> re-invocation of constraints.

Yes, that seems like a useful property of invariants.

> The arguments can only be used from Java bodies.

What can the arguments be used for in that case? Provide better
messages? I'm implementing support for Javascript-based delegates
(similar to the OCL support), so I wonder if there are special use cases
I should support.

Hallvard

Re: Question about constraints vs. invariants [message #687034 is a reply to message #687033]

Tue, 24 May 2011 01:12

Eclipse User

Hi

>> The arguments can only be used from Java bodies.
>
> What can the arguments be used for in that case? Provide better
> messages? I'm implementing support for Javascript-based delegates
> (similar to the OCL support), so I wonder if there are special use
> cases I should support.
>
As I wrote that I considered adding a magic 'oclContextTuple' to access
the context from OCL, but very low priority and certainly not at RC2.

The 'DiagnosticChain diagnostics' argument is needed if you take control
since it is the diagnostic output 'stream'. If you have a conventional
control flow then it is used automatically.

The 'Map<Object, Object> context' argument allows arbitrary objects and
values to be passed in, and in some cases updated. A common use is
provision of a SubstitutionLabelProvider to pretty print the
descriptions of interesting objects in diagnostics.

Regards

Ed Willink

Previous Topic:	deleting eobject
Next Topic:	(no subject)

Goto Forum:

-=] Back to Top [=-

Current Time: Sat Jul 05 08:32:22 EDT 2025

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter