| Strange result in query [message #667299] |
Wed, 27 April 2011 21:54  |
Eclipse User |
|
|
|
Hi,
I have tried the following recursive query (it traverses up a chain of
Package objects, to compute the full name):
[query public packageFullName(pack : Package, suffix : String) : String =
if pack.oclIsUndefined() then suffix else
packageFullName(pack._package, pack.name + '.' + suffix) endif
/]
If called on a chain of packages with names org.eclipse.tests it returns
<undefined-object>.org.eclipse where <undefined-object> is the result of
calling oclUndefined.toString(). It is as if the pack.name expression in
pack.name + '.' + suffix is interpreted as pack._package.name.
However, the following works:
[query public packageFullName(pack : Package, suffix : String) : String =
if pack.oclIsUndefined() then suffix else let suffix : String =
pack.name + '.' + suffix in packageFullName(pack._package, suffix) endif
/]
I.e. the argument is computed before the recursive call to
packageFullName. I cannot understand why the result should differ and be
wrong in the first case.
Hallvard
|
|
|
| Re: Strange result in query [message #667790 is a reply to message #667299] |
Mon, 02 May 2011 09:02  |
| Eclipse User |
|
|
|
Hi Hallvard,
Seems to me like a variable leak in Acceleo. Could you raise a bug against M2T/Acceleo for this?
As for what you are trying to achieve ... Wouldn't it be easier with something like :
[ancestors(Package)->reverse().name->sep('.')/]
i.e "retrieve all ancestors of type "Package", reverse the list so that the root is first in the list, get all their names, then separate these names with dots". This is what we usually use when computing qualified names.
Laurent Goubet
Obeo
|
|
|
| Re: Strange result in query [message #667805 is a reply to message #667790] |
Mon, 02 May 2011 09:44 |
| Eclipse User |
|
|
|
On 02.05.11 11.02, Laurent Goubet wrote:
>
> Seems to me like a variable leak in Acceleo. Could you raise a
> https://bugs.eclipse.org/bugs/enter_bug.cgi?product=M2T& component=Acceleo for
> this?
Done: https://bugs.eclipse.org/bugs/show_bug.cgi?id=344424
> As for what you are trying to achieve ... Wouldn't it be easier with
> something like :
>
> [ancestors(Package)->reverse().name->sep('.')/]
I didn't know about the ancestors function, thanks! However, the root
package is not the root of the model, so I think this would give me an
extra element that I need to remove (before or after doing reverse, I
guess).
Ideally, there is I would like to use a function that traverses along a
named relation (not necessarily the ancestor chain), 'package' in my
case, and collects them until it is null/empty. Does such a function
exist? Can I perhaps use OCL iterate construct?
Hallvard
|
|
|
| Re: Strange result in query [message #667806 is a reply to message #667805] |
Mon, 02 May 2011 09:48 |
| Eclipse User |
|
|
|
On 02.05.11 11.44, Hallvard Trætteberg wrote:
> On 02.05.11 11.02, Laurent Goubet wrote:
>>
>> Seems to me like a variable leak in Acceleo. Could you raise a
>> https://bugs.eclipse.org/bugs/enter_bug.cgi?product=M2T& component=Acceleo
>> for
>> this?
>
> Done: https://bugs.eclipse.org/bugs/show_bug.cgi?id=344424
>
>> As for what you are trying to achieve ... Wouldn't it be easier with
>> something like :
>>
>> [ancestors(Package)->reverse().name->sep('.')/]
>
> I didn't know about the ancestors function, thanks! However, the root
> package is not the root of the model, so I think this would give me an
> extra element that I need to remove (before or after doing reverse, I
> guess).
Sorry, I didn't notice the Package (oclType) argument, which filters out
non-Package object. Thanks for the tip!
Hallvard
|
|
|
| Re: Strange result in query [message #667809 is a reply to message #667806] |
Mon, 02 May 2011 10:07 |
| Eclipse User |
|
|
|
On 02.05.11 11.48, Hallvard Trætteberg wrote:
> On 02.05.11 11.44, Hallvard Trætteberg wrote:
>> On 02.05.11 11.02, Laurent Goubet wrote:
>>>
>>> Seems to me like a variable leak in Acceleo. Could you raise a
>>> https://bugs.eclipse.org/bugs/enter_bug.cgi?product=M2T& component=Acceleo
>>>
>>> for
>>> this?
>>
>> Done: https://bugs.eclipse.org/bugs/show_bug.cgi?id=344424
>>
>>> As for what you are trying to achieve ... Wouldn't it be easier with
>>> something like :
>>>
>>> [ancestors(Package)->reverse().name->sep('.')/]
>>
>> I didn't know about the ancestors function, thanks! However, the root
>> package is not the root of the model, so I think this would give me an
>> extra element that I need to remove (before or after doing reverse, I
>> guess).
>
> Sorry, I didn't notice the Package (oclType) argument, which filters out
> non-Package object. Thanks for the tip!
Update: ancestors(Package)->reverse().name gives an error marker. I
think it's because reverse isn't recognized, since
ancestors(Package).name works. BTW, I'm using Acceleo 3.0.2 with Helios.
Hallvard
|
|
|
| Re: Strange result in query [message #667818 is a reply to message #667299] |
Mon, 02 May 2011 10:32 |
| Eclipse User |
|
|
|
Hallvard,
I know we've fixed a number of bugs with both the "reverse" and "ancestors" services ... but I don't remember which  . The 3.1.0 M7 build for Acceleo will be available tomorrow, and it should fix the issue you have with this query.
In the meantime, and if memory serves, you should be able to use this query instead :
[ancestors(Package)->asSequence()->reverse().name->sep('.')/]
As I seem to remember something about "ancestors" not being statically typed to a Sequence or OrderedSet (i.e. an ordered collection) and thus "reverse" not being applicable.
If that doesn't fix it, could you provide me with the message associated with this compilation error?
Laurent Goubet
Obeo
|
|
|
| Re: Strange result in query [message #667836 is a reply to message #667818] |
Mon, 02 May 2011 12:09 |
| Eclipse User |
|
|
|
On 02.05.11 12.32, Laurent Goubet wrote:
>
> [ancestors(Package)->asSequence()->reverse().name->sep('.')/]
This does not work, either.
> If that doesn't fix it, could you provide me with the message associated
> with this compilation error?
The error marker's message is "Unrecognized variable: (name)". As I
said, ancestors(Package).name->sep('.') works fine, so I think the
problem is that reverse() is missing.
Hallvard
|
|
|
| Re: Strange result in query [message #667884 is a reply to message #667299] |
Mon, 02 May 2011 15:12 |
| Eclipse User |
|
|
|
ha, that one  . It has been fixed too. The problem was that "reverse" lost the typing of the collection, ancestors(Package) returns a Collection(Package) but "Collection(Package)->reverse()" returns a Collection(T) ... and "name" does not exist on "T".
You can work around this by manually re-casting :
[ancestors(Package)->reverse().oclAsType(Package).name->sep('.')/]
These problems are fixed in Acceleo 3.1.
Laurent Goubet
Obeo
|
|
|
| Re: Strange result in query [message #667935 is a reply to message #667884] |
Mon, 02 May 2011 20:07 |
| Eclipse User |
|
|
|
On 02.05.11 17.12, Laurent Goubet wrote:
> ha, that one :p. It has been fixed too. The problem was that "reverse"
> lost the typing of the collection, ancestors(Package) returns a
> Collection(Package) but "Collection(Package)->reverse()" returns a
> Collection(T) ... and "name" does not exist on "T".
I understand. I can imagine it is a bit tricky to get everything correct.
> You can work around this by manually re-casting :
> [ancestors(Package)->reverse().oclAsType(Package).name->sep('.')/]
When you use -> it is a method on the collection (object), right. While
when you use . it applies to every element in the collection? So
..oclAsType(Package) means the same as ->oclAsType(Sequence(Package))
(not sure if that's correct ocl, but you understand what I mean)?
> These problems are fixed in Acceleo 3.1.
Which is coming to a place near you, soon :-)
Hallvard
|
|
|
| Re: Strange result in query [message #667971 is a reply to message #667935] |
Tue, 03 May 2011 07:39 |
| Eclipse User |
|
|
|
| Quote: |
When you use -> it is a method on the collection (object), right. While
when you use . it applies to every element in the collection? So
..oclAsType(Package) means the same as ->oclAsType(Sequence(Package))
(not sure if that's correct ocl, but you understand what I mean)?
|
Yup, though "oclAsType" cannot be used on a Collection (with a "->") AFAIK.
| Quote: |
| Quote: |
These problems are fixed in Acceleo 3.1.
|
Which is coming to a place near you, soon 
|
The M7 will be available today on the Milestones update site.
Laurent Goubet
Obeo
|
|
|
Powered by
FUDForum. Page generated in 0.06812 seconds
] 
Search
Help
Register
Login
Home
