Home » Modeling » TMF (Xtext) » Problem with opposites
| Problem with opposites [message #629274] |
Mon, 27 September 2010 16:10  |
 Hallvard Traetteberg
Messages: 673
Registered: July 2009
Location: Trondheim, Norway |
Senior Member |
|
|
Hi,
I have a model with a Role, Person and Allocation objects. Role and
Person both have lists of Allocations and Allocation refers to the
others, and pairs of these references are opposites: Role.allocations is
the opposite of Allocation.role and Person.allocations is the opposite
of Allocation.person.
In my grammar, I have rules for creating Roles, Persons and Allocation,
where only the latter includes grammar fragments for referring to the
others: as follows (slightly simplified):
Role: roleName=EString 'providing' workload=EInt 'units';
Person: 'Person' name=EString 'provides' workload=EInt 'units';
Allocation: person=[Person|EString] 'fills' role=[Role|EString] 'role';
In my test file, I first have all the Roles and Persons. Then the
Allocations follows:
"Hallvard Trætteberg" fills "tdt4250 course responsible" role
"Hallvard Trætteberg" fills "tdt4250 lecturer" role
"Babak Farshchian" fills "tdt4250 lecturer" role
However, I get the following error marker for the first line, and
similiar for each Person/Allocation and Role/Allocation pairt (six in
total):
The opposite features 'person' of AllocationImpl@498b4d
{platform:/resource/rptest/model/h10.rpxt#//@allocations.0}' and
'allocations' of PersonImpl@78bb58
{platform:/resource/rptest/model/h10.rpxt#Hallvard Trætteberg}' do not
refer to each other
I interpret this to mean that the objects that should refer to each
other have inconsistent lists/values. How can this happen?
Hallvard
|
|
|
| Re: Problem with opposites [message #629292 is a reply to message #629274] |
Mon, 27 September 2010 16:38  |
Ed Willink
Messages: 7655
Registered: July 2009 |
Senior Member |
|
|
Hi
This should not happen but very easily does during development.
Are you sure that you validated your meta-model?
Are you sure you re-genmodelled after changing your meta-model?
Are you sure that you regenerated the editor after re-genmodeling?
Regards
Ed Willink
On 27/09/2010 17:10, Hallvard Trætteberg wrote:
> Hi,
>
> I have a model with a Role, Person and Allocation objects. Role and
> Person both have lists of Allocations and Allocation refers to the
> others, and pairs of these references are opposites: Role.allocations is
> the opposite of Allocation.role and Person.allocations is the opposite
> of Allocation.person.
>
> In my grammar, I have rules for creating Roles, Persons and Allocation,
> where only the latter includes grammar fragments for referring to the
> others: as follows (slightly simplified):
>
> Role: roleName=EString 'providing' workload=EInt 'units';
> Person: 'Person' name=EString 'provides' workload=EInt 'units';
> Allocation: person=[Person|EString] 'fills' role=[Role|EString] 'role';
>
> In my test file, I first have all the Roles and Persons. Then the
> Allocations follows:
>
> "Hallvard Trætteberg" fills "tdt4250 course responsible" role
> "Hallvard Trætteberg" fills "tdt4250 lecturer" role
> "Babak Farshchian" fills "tdt4250 lecturer" role
>
> However, I get the following error marker for the first line, and
> similiar for each Person/Allocation and Role/Allocation pairt (six in
> total):
>
> The opposite features 'person' of AllocationImpl@498b4d
> {platform:/resource/rptest/model/h10.rpxt#//@allocations.0}' and
> 'allocations' of PersonImpl@78bb58
> {platform:/resource/rptest/model/h10.rpxt#Hallvard Trætteberg}' do not
> refer to each other
>
> I interpret this to mean that the objects that should refer to each
> other have inconsistent lists/values. How can this happen?
>
> Hallvard
|
|
| | |
| Re: Problem with opposites (and proxies) [message #629353 is a reply to message #629274] |
Mon, 27 September 2010 22:05 |
Hallvard Traetteberg
Messages: 673
Registered: July 2009
Location: Trondheim, Norway |
Senior Member |
|
|
I've tried tracing what happens, and here's what I see (in the debugger):
1) Object creation phase: Person names are set.
2) (Partly) as expected: The Allocation objects are linked to Person
proxies, i.e. different ones than in phase 1. E.g. these do not have
names. After the linking, these Person proxies correctly refer back to
the Allocation objects (the opposite invariant).
3) During the validation phase the Person proxies are resolved, hence,
the Allocation objects are "backpatched" to refer to the correct Person
objects. However, the references from the Person proxies back to
Allocation objects, that are due to the opposite invariant, are not set
as part of resolving the proxy. Hence, the proxy still refers to the
Allocation object, and the real ones do not.
How could this have happened? BTW, I've removed AllocationImpl and
PersonImpl, reloaded the genmodel and regenerated the code.
Hallvard
On 27.09.10 18.10, Hallvard Trætteberg wrote:
>
> I have a model with a Role, Person and Allocation objects. Role and
> Person both have lists of Allocations and Allocation refers to the
> others, and pairs of these references are opposites: Role.allocations is
> the opposite of Allocation.role and Person.allocations is the opposite
> of Allocation.person.
>
> In my grammar, I have rules for creating Roles, Persons and Allocation,
> where only the latter includes grammar fragments for referring to the
> others: as follows (slightly simplified):
>
> Role: roleName=EString 'providing' workload=EInt 'units';
> Person: 'Person' name=EString 'provides' workload=EInt 'units';
> Allocation: person=[Person|EString] 'fills' role=[Role|EString] 'role';
>
> In my test file, I first have all the Roles and Persons. Then the
> Allocations follows:
>
> "Hallvard Trætteberg" fills "tdt4250 course responsible" role
> "Hallvard Trætteberg" fills "tdt4250 lecturer" role
> "Babak Farshchian" fills "tdt4250 lecturer" role
>
> However, I get the following error marker for the first line, and
> similiar for each Person/Allocation and Role/Allocation pairt (six in
> total):
>
> The opposite features 'person' of AllocationImpl@498b4d
> {platform:/resource/rptest/model/h10.rpxt#//@allocations.0}' and
> 'allocations' of PersonImpl@78bb58
> {platform:/resource/rptest/model/h10.rpxt#Hallvard Trætteberg}' do not
> refer to each other
>
> I interpret this to mean that the objects that should refer to each
> other have inconsistent lists/values. How can this happen?
>
> Hallvard
|
|
|
| Re: Problem with opposites (and proxies) [message #629356 is a reply to message #629353] |
Mon, 27 September 2010 22:35 |
Hallvard Traetteberg
Messages: 673
Registered: July 2009
Location: Trondheim, Norway |
Senior Member |
|
|
I should have googled first:
https://bugs.eclipse.org/bugs/show_bug.cgi?format=multiple&a mp;id=282486
Sorry for the noise.
Setting resolve proxies to false "solved" the problem. Will this be
fixed to work with proxies in the future?
Hallvard
On 28.09.10 00.05, Hallvard Trætteberg wrote:
> I've tried tracing what happens, and here's what I see (in the debugger):
> 1) Object creation phase: Person names are set.
> 2) (Partly) as expected: The Allocation objects are linked to Person
> proxies, i.e. different ones than in phase 1. E.g. these do not have
> names. After the linking, these Person proxies correctly refer back to
> the Allocation objects (the opposite invariant).
> 3) During the validation phase the Person proxies are resolved, hence,
> the Allocation objects are "backpatched" to refer to the correct Person
> objects. However, the references from the Person proxies back to
> Allocation objects, that are due to the opposite invariant, are not set
> as part of resolving the proxy. Hence, the proxy still refers to the
> Allocation object, and the real ones do not.
>
> How could this have happened? BTW, I've removed AllocationImpl and
> PersonImpl, reloaded the genmodel and regenerated the code.
>
> Hallvard
>
> On 27.09.10 18.10, Hallvard Trætteberg wrote:
>>
>> I have a model with a Role, Person and Allocation objects. Role and
>> Person both have lists of Allocations and Allocation refers to the
>> others, and pairs of these references are opposites: Role.allocations is
>> the opposite of Allocation.role and Person.allocations is the opposite
>> of Allocation.person.
>>
>> In my grammar, I have rules for creating Roles, Persons and Allocation,
>> where only the latter includes grammar fragments for referring to the
>> others: as follows (slightly simplified):
>>
>> Role: roleName=EString 'providing' workload=EInt 'units';
>> Person: 'Person' name=EString 'provides' workload=EInt 'units';
>> Allocation: person=[Person|EString] 'fills' role=[Role|EString] 'role';
>>
>> In my test file, I first have all the Roles and Persons. Then the
>> Allocations follows:
>>
>> "Hallvard Trætteberg" fills "tdt4250 course responsible" role
>> "Hallvard Trætteberg" fills "tdt4250 lecturer" role
>> "Babak Farshchian" fills "tdt4250 lecturer" role
>>
>> However, I get the following error marker for the first line, and
>> similiar for each Person/Allocation and Role/Allocation pairt (six in
>> total):
>>
>> The opposite features 'person' of AllocationImpl@498b4d
>> {platform:/resource/rptest/model/h10.rpxt#//@allocations.0}' and
>> 'allocations' of PersonImpl@78bb58
>> {platform:/resource/rptest/model/h10.rpxt#Hallvard Trætteberg}' do not
>> refer to each other
>>
>> I interpret this to mean that the objects that should refer to each
>> other have inconsistent lists/values. How can this happen?
>>
>> Hallvard
>
|
|
|
Goto Forum:
Current Time: Tue Apr 16 18:34:48 GMT 2024
Powered by FUDForum. Page generated in 0.61367 seconds |
] 
Search
Help
Register
Login
Home
