[news.eclipse.modeling.gmt] Re: [Epsilon] Problem with ECL

[Dimitrios Kolovos <dskolovos@xxxxxxxxx>]

Hi Steve,

I can't reproduce this. For a Left model that contains only one Class A 
and a Right model that contains only one Class B I get the following result:

A <ClassA> B

which is correct. There is only one pair to be compared (why did you 
expect 4?), and when the ClassA rule is executed it returns true and so 
A and B are recorded as matching.

I suspect that both of your two models have 2 classes (A and B) each. 
That would explain the result: the 4 pairs first pass from rule ClassA 
which characterizes the A<->A and A<->B pairs as matching, and the B<->A 
and B<->B pairs as non-matching (your clause is l.name = 'A') and as a 
result, since at that point the engine has decided about all pairs, rule 
ClassB is not needed and thus is not executed.

Cheers,
Dimitrios

Steve Barrett wrote:
> Hello,
>
> I'm trying to use ECL to help with some model merging. Unfortunately I'm 
> stuck.
>
> My left model consits of a single class A, and the right a single class 
> B. I'm using the follow code (class.ecl) to compare them and print out 
> the match trace:
>
>
> rule ClassA
>  match l : Left!Class with r : Right!Class {
>  compare : l.name = 'A' }
>
> rule ClassB
>  match l : Left!Class with r : Right!Class {
>  compare : l.name = 'B' }
>
> post {
>  -- Print name of matching elements and of the rule that matched them.
>  for (m in matchTrace.matches.select(m|m.isMatching)) {
>    (m.left.name + ' <' + m.getRule().name + '> ' + m.right.name).println();
>  }
> }
>
>
> I think that the four element pairings should cause each rule to fire 
> once, so the output is a surprise to me:
>
>
> A <ClassA> A
> A <ClassA> B
>
>
> Where are the B to A, B to B pairings? Why did rule ClassA match A with 
> B? And why does the second rule never fire?
>
> I can't help put think that I'm missing something really basic here, but 
> I have spent a bit of time trying to unravel this. Thanks for taking a 
> look.
>
> --Steve
>
> P.S. I couldn't get "m.rules.first().name" as given in the Complete OO 
> to DB comparison to work, which is why I'm using "m.getRule().name" 
> Could this be a source of trouble?