Eclipse Community Forums: M2T (model-to-text transformation)

Home » Modeling » M2T (model-to-text transformation) » a difficult query to solve(find all the routes between one class and another)

Show: Today's Messages :: Show Polls :: Message Navigator

a difficult query to solve [message #1804554]

Thu, 28 March 2019 17:06

Jonathan Marino

Messages: 8
Registered: May 2015

Junior Member

I am doing my engineering thesis on code generation for C language and I'm evaluating acceleo.
I have the next problem:
We are in uml using multiple inheritance. Let's say we have the following class diagram:

A
/ | \
B C D
| | \ |
| | E
\ | /
F

And we want to get all possible paths from A to F
((A, B, F), (A, C, F), (A, C, E, F), (A, D, E, F))
How can I write an OCL query that throws that result for that diagram? is there a way to solve something like this with JAVA instead of OCL?

Report message to a moderator

Re: a difficult query to solve [message #1804609 is a reply to message #1804554]

Fri, 29 March 2019 13:18

Tomasz abczy?ski

Messages: 11
Registered: May 2015

Junior Member

For the first question I should to think a bit but I suppose that the recursion will fit it.
Something like this (probably a bit more complicated):

[query private superClasses(c: Class) : OrderedSet(Class) = superClasses(c.superClass) /]

The answer for the last question is - yes, java wrapper. Look at tutorial:
https://wiki.eclipse.org/Acceleo/Getting_Started#Java_services_wrappers

Tomasz Babczyński

Report message to a moderator

Re: a difficult query to solve [message #1804620 is a reply to message #1804609]

Fri, 29 March 2019 15:28

Ed Willink

Messages: 7655
Registered: July 2009

Senior Member

Hi

(Apologies for not responding to your revised message on the OCL newsgroup. Your initial message came in at the same time as many SPAM philosophical queries which you inappropriate reference to diamond reinforced.)

The problem involves a nested iteration of indeterminate depth. This can sometimes be solved by a closure() iteration, but more usually by a recursion.

Given a helper function superPaths() : Set(Sequence(Class)) the problem degenerates to prepending each of the local superClasses on to the superPaths of each superClass.

Body something like: self.superClasses->collectNested(superClass | superPaths(superClass )->prepend(superClass ))->asSet()

In EMF-Java, you can define getSuperPaths() as a derived property with a cache.

In OCLinEcore, you can do much the same, the cache should happen automatically.

In Acceleo, the inherent caching of each operation invocation will automatically build the 'derived-property' cache as a Map from Class to Sequence(Class).

(If you use derived property caches, you should be aware that they will go stale if you change the metamodel.)

Regards

Ed Willink

Report message to a moderator

Re: a difficult query to solve [message #1805335 is a reply to message #1804620]

Fri, 12 April 2019 02:01

Jonathan Marino

Messages: 8
Registered: May 2015

Junior Member

Hello,
Thanks for the prompt reply.
I am not able to resolve the issue.
Based on your answer I coded the following query:

[query public superPaths(aClass: Class) : Sequence(Sequence(Class))=
	if (aClass.superClass->isEmpty()) then
		Sequence(Sequence(Class)){Sequence(Class){aClass}}
	else
		aClass.superClass->collectNested(superClass1 | superPaths(superClass1 ).prepend(aClass ))->asSequence()
	endif
/]

And I generated the following template to verify the result

[for (classes:Sequence(Class) | superPaths())]
	Route [i/]:
	[for (class: uml::Class | classes)separator(',')][class.name/][/for]
[/for]

But with a simple model:

Class1    Class6
|       /       |
Class3       Class2
       \     /
        Class4
          |
        Class5

For Class4 I get:
Route 1:
Class4, Class3, Class6, Class4, Class3, Class1
Route 2:
Class4, Class2, Class6

And for Class5:
Route 1:
Class5, Class4, Class3, Class6, Class4, Class3, Class1, Class5, Class4, Class2, Class6

It is close, apparently after a second level of inheritance the sequences become flattened. I do not know if I'm facing a bug or there's something wrong with the query ...

[Updated on: Fri, 12 April 2019 02:04]

Report message to a moderator

Re: a difficult query to solve [message #1805345 is a reply to message #1805335]

Fri, 12 April 2019 06:58

Ed Willink

Messages: 7655
Registered: July 2009

Senior Member

Hi

In my example I used Set(Sequence(T)). I have no idea why you use Sequence(Sequence(T)) ; it may malfunction.

If your example doesn't work, then you need to debug it. Try simpler cases. If it flattens find out where the flattening occurs.

I'm not sure why you call your answers 'close'. The only closeness in that some input elements appear. How can Class1 possibly be the middle of a path?

I do not have time to look at snippets of your functionality. You need to post an executable repro to get help.

Generally with Acceleo, introduction of let-variables can aid understanding and reduce opportunities for tooling malfunctions. Nested collections offer ample opportunities for both.

Regards

Ed Willink

Report message to a moderator

Re: a difficult query to solve [message #1806026 is a reply to message #1805345]

Tue, 30 April 2019 18:04

Jonathan Marino

Messages: 8
Registered: May 2015

Junior Member

Hello,
After many attempts I was not able to create the query with Set instead of Sequence. At least I receive the error: "Invalid type for iteration".

For now in my thesis I say that I was not able to solve it with OCL, but I would like to change it.

The repository executable with the query is at: https://github.com/jonyMarino/oopc_acceleo_issue.git

Thank you very much and I hope you can help me,
Regards,
Jonathan

PD: I said it was close because if you notice even where Class1 appears in the middle is that there are many paths one after another, the fact that Class5 does not appear three times tells me that part of the collection was flattened before the "prepend" and after that it seems that all the collection was flattened...

Report message to a moderator

Re: a difficult query to solve [message #1806040 is a reply to message #1806026]

Wed, 01 May 2019 09:23

Ed Willink

Messages: 7655
Registered: July 2009

Senior Member

Hi

The problem is that in
Quote:

Body something like: self.superClasses->collectNested(superClass | superPaths(superClass )->prepend(superClass ))->asSet()

collectNested is not a suitable operation for merging multiple Set(Sequence()) to give a larger Set; it gives a wrapped Set. You need to fall back on iterate.

You should find that the following works:

Quote:

[comment encoding = UTF-8 /]
[module generate('http://www.eclipse.org/uml2/5.0.0/UML')]

[query public superPaths(aClass: uml::Class) : OrderedSet(Sequence(uml::Class))=
let superSuperPaths : Set(Sequence(uml::Class)) = if aClass.superClass->isEmpty()
then OrderedSet(Sequence(Class)){Sequence(Class){}}
else aClass.superClass->iterate(superClass1;
result : Set(Sequence(uml::Class)) = Set(Sequence(uml::Class)){}
| result->union(superPaths(superClass1)))
endif in
superSuperPaths->collectNested(superPath : Sequence(uml::Class) | superPath->prepend(aClass))->asOrderedSet()
/]

[template public main(aPackage : uml::Package)]
[comment @main /]
[file (aPackage.name + '.txt', false, 'UTF-8')]
[for (aClass : uml::Class | aPackage.ownedType->filter(uml::Class))]
[genClass(aClass)/]
[/for]
[/file]
[/template]

[template public genClass (aClass : uml::Class) ]
[aClass.name/]
[for (path : Sequence(uml::Class) | superPaths(aClass))]
Route [i/]:[for (class : uml::Class | path) separator(',')] [class.name/][/for]
[/for]
[/template]

At least the following looks like a plausible result:

Quote:

Class4
Route 1: Class4, Class2, Class6
Route 2: Class4, Class3, Class6
Route 3: Class4, Class3, Class1

Class1
Route 1: Class1

Class5
Route 1: Class5, Class4, Class3, Class6
Route 2: Class5, Class4, Class3, Class1
Route 3: Class5, Class4, Class2, Class6

Class2
Route 1: Class2, Class6

Class6
Route 1: Class6

Class3
Route 1: Class3, Class1
Route 2: Class3, Class6

Regards

Ed Willink

Report message to a moderator

Re: a difficult query to solve [message #1806110 is a reply to message #1806040]

Thu, 02 May 2019 18:05

Jonathan Marino

Messages: 8
Registered: May 2015

Junior Member

Thank you very much Ed!

To complete the answer I added this query to get all the paths from any class to another:

[query public getPaths(fromClass: uml::Class, toClass: uml::Class) : OrderedSet(Sequence(uml::Class))=
fromClass.superPaths()->select(path: Sequence(uml::Class) | path->includes(toClass))->collectNested(path: Sequence(uml::Class) | path->subSequence(2,path->indexOf(toClass)))->asOrderedSet()
/]

I think that the JAVA way is very necessary since it allows debugging.
If I had known before that it served to solve this type of queries I would have fought much less with them and do it the JAVA form. Any way, the queries in OCL enrich the thesis very much.

Regards!

Report message to a moderator

Previous Topic:	Invoke method cannot used for OrderedSet ?
Next Topic:	[Acceleo] Disambiguate class reference

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

]

Current Time: Fri Apr 19 01:32:59 GMT 2024

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter