MooseChef: to what package do methods belong?
Locked
MooseChef: to what package do methods belong?
 
|
Another MooseChef decision: aPackage queryOutgoingReferences returns all the FamixReferences made from aPackage. this is implemented by taking all methods (self methods) in the package and asking them what reference they make. And we have FAMIXPackage>>methods ^ self localMethods union: self extensionMethods So when asking a package what external references it makes, we get references stemming from methods that extend classes of other packages. We found this while computing a coupling metric for packages. So the solution would be easy to implement, but we are not sure what is the best solution ? - we could change FAMIXPackage>>methods (methods belong to the package that owns their class) - we could change queryOutgoingReferences (references/invocation/accesses stem from the methods in the classes in the package, not from the methods in the package) - or we could change our metric nicolas _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
Re: MooseChef: to what package do methods belong?
|
|
it would be good to methodExtens
aPackage methods = aPackage methodExtensions + aPackage methodDefined? aClass extendedMethods have at RPackage because I spent time on the API. On Jul 31, 2012, at 4:21 PM, Nicolas Anquetil wrote: > > Another MooseChef decision: > > aPackage queryOutgoingReferences > > returns all the FamixReferences made from aPackage. > > this is implemented by taking all methods (self methods) in the package and asking them what reference they make. > > And we have > FAMIXPackage>>methods > ^ self localMethods union: self extensionMethods > > So when asking a package what external references it makes, we get references stemming from methods that extend classes of other packages. > > We found this while computing a coupling metric for packages. > > So the solution would be easy to implement, but we are not sure what is the best solution ? > > - we could change FAMIXPackage>>methods (methods belong to the package that owns their class) > - we could change queryOutgoingReferences (references/invocation/accesses stem from the methods in the classes in the package, not from the methods in the package) > - or we could change our metric > > nicolas > _______________________________________________ > Moose-dev mailing list > [hidden email] > https://www.iam.unibe.ch/mailman/listinfo/moose-dev _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
Re: MooseChef: to what package do methods belong?
|
|
In reply to this post by Nicolas Anquetil
Hi Nicolas,
The current semantics for both methods and queryOutgoingReferences are quite fine. As I understand you want to count the amount of references that the code inside a package induces. So, this includes both the extension methods and the defined methods. Why do you see this as a problem? Cheers, Doru On 31 Jul 2012, at 16:21, Nicolas Anquetil wrote: > > Another MooseChef decision: > > aPackage queryOutgoingReferences > > returns all the FamixReferences made from aPackage. > > this is implemented by taking all methods (self methods) in the package and asking them what reference they make. > > And we have > FAMIXPackage>>methods > ^ self localMethods union: self extensionMethods > > So when asking a package what external references it makes, we get references stemming from methods that extend classes of other packages. > > We found this while computing a coupling metric for packages. > > So the solution would be easy to implement, but we are not sure what is the best solution ? > > - we could change FAMIXPackage>>methods (methods belong to the package that owns their class) > - we could change queryOutgoingReferences (references/invocation/accesses stem from the methods in the classes in the package, not from the methods in the package) > - or we could change our metric > > nicolas > _______________________________________________ > Moose-dev mailing list > [hidden email] > https://www.iam.unibe.ch/mailman/listinfo/moose-dev -- www.tudorgirba.com "Yesterday is a fact. Tomorrow is a possibility. Today is a challenge." _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
Re: MooseChef: to what package do methods belong?
|
|
we work on coupling metrics that measure how much a package depends on other packages (can you reuse it without the others). Even if it is trough a method added by extension, if package A use something in package B, then A needs B (or if you prefer A+C needs B) The current semantic does not show this because the extending method will be part of package C (the one that introduces the extension) and therefore C will depend on B (which seem correct to me), but A will not depend on B ... So with extensions, we can consider that both A and C depend on B. This is something that the coupling metrics do not consider. So we have to make a choice. For coherence with the same experiment on Java systems, we prefer to say that only A depends on B. But tis is not how famix and moose chef see things ... Additionaly, we have a metric whose results are useless in the presence of extensions. nicolas On 01/08/12 20:39, Tudor Girba wrote: > Hi Nicolas, > > The current semantics for both methods and queryOutgoingReferences are quite fine. > > As I understand you want to count the amount of references that the code inside a package induces. So, this includes both the extension methods and the defined methods. Why do you see this as a problem? > > Cheers, > Doru > > > On 31 Jul 2012, at 16:21, Nicolas Anquetil wrote: > >> Another MooseChef decision: >> >> aPackage queryOutgoingReferences >> >> returns all the FamixReferences made from aPackage. >> >> this is implemented by taking all methods (self methods) in the package and asking them what reference they make. >> >> And we have >> FAMIXPackage>>methods >> ^ self localMethods union: self extensionMethods >> >> So when asking a package what external references it makes, we get references stemming from methods that extend classes of other packages. >> >> We found this while computing a coupling metric for packages. >> >> So the solution would be easy to implement, but we are not sure what is the best solution ? >> >> - we could change FAMIXPackage>>methods (methods belong to the package that owns their class) >> - we could change queryOutgoingReferences (references/invocation/accesses stem from the methods in the classes in the package, not from the methods in the package) >> - or we could change our metric >> >> nicolas >> _______________________________________________ >> Moose-dev mailing list >> [hidden email] >> https://www.iam.unibe.ch/mailman/listinfo/moose-dev > -- > www.tudorgirba.com > > "Yesterday is a fact. > Tomorrow is a possibility. > Today is a challenge." > > > > > _______________________________________________ > Moose-dev mailing list > [hidden email] > https://www.iam.unibe.ch/mailman/listinfo/moose-dev _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
Re: MooseChef: to what package do methods belong?
|
|
Hi,
I still do not understand :). Let's be more specific. As I understand we could use the following example: - package A = FAMIX-Core - package B = Mondrian - package C = Moose-Finder The Moose-Finder extends FAMIX entities with visualization methods that depend on Mondrian. This decision was taken exactly so that FAMIX does not depend on Mondrian. So, why would you want to see FAMIX-Core depend on Mondrian? Cheers, Doru On Thu, Aug 2, 2012 at 9:24 AM, Nicolas Anquetil <[hidden email]> wrote: > > we work on coupling metrics that measure how much a package depends on other > packages (can you reuse it without the others). > > Even if it is trough a method added by extension, if package A use something > in package B, then A needs B (or if you prefer A+C needs B) > The current semantic does not show this because the extending method will be > part of package C (the one that introduces the extension) and therefore C > will depend on B (which seem correct to me), but A will not depend on B ... > > So with extensions, we can consider that both A and C depend on B. > This is something that the coupling metrics do not consider. > So we have to make a choice. > For coherence with the same experiment on Java systems, we prefer to say > that only A depends on B. > > But tis is not how famix and moose chef see things ... > > Additionaly, we have a metric whose results are useless in the presence of > extensions. > > nicolas > > > On 01/08/12 20:39, Tudor Girba wrote: >> >> Hi Nicolas, >> >> The current semantics for both methods and queryOutgoingReferences are >> quite fine. >> >> As I understand you want to count the amount of references that the code >> inside a package induces. So, this includes both the extension methods and >> the defined methods. Why do you see this as a problem? >> >> Cheers, >> Doru >> >> >> On 31 Jul 2012, at 16:21, Nicolas Anquetil wrote: >> >>> Another MooseChef decision: >>> >>> aPackage queryOutgoingReferences >>> >>> returns all the FamixReferences made from aPackage. >>> >>> this is implemented by taking all methods (self methods) in the package >>> and asking them what reference they make. >>> >>> And we have >>> FAMIXPackage>>methods >>> ^ self localMethods union: self extensionMethods >>> >>> So when asking a package what external references it makes, we get >>> references stemming from methods that extend classes of other packages. >>> >>> We found this while computing a coupling metric for packages. >>> >>> So the solution would be easy to implement, but we are not sure what is >>> the best solution ? >>> >>> - we could change FAMIXPackage>>methods (methods belong to the package >>> that owns their class) >>> - we could change queryOutgoingReferences (references/invocation/accesses >>> stem from the methods in the classes in the package, not from the methods in >>> the package) >>> - or we could change our metric >>> >>> nicolas >>> _______________________________________________ >>> Moose-dev mailing list >>> [hidden email] >>> https://www.iam.unibe.ch/mailman/listinfo/moose-dev >> >> -- >> www.tudorgirba.com >> >> "Yesterday is a fact. >> Tomorrow is a possibility. >> Today is a challenge." >> >> >> >> >> _______________________________________________ >> Moose-dev mailing list >> [hidden email] >> https://www.iam.unibe.ch/mailman/listinfo/moose-dev > > > _______________________________________________ > Moose-dev mailing list > [hidden email] > https://www.iam.unibe.ch/mailman/listinfo/moose-dev -- www.tudorgirba.com "Every thing has its own flow" _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
Re: MooseChef: to what package do methods belong?
|
|
A simple example.
If you create a model with the package Fame-Util, then, inspect such package: "1" (self classes flatCollectAsSet: [ :each | each queryOutgoingDependencies atPackageScope ]) size " = 7"
"2" (self queryOutgoingDependencies atPackageScope asSet) size " = 12" What in a first moment is strange. At least when we are implementing metrics..
We would expect the same outgoing references (same size) bor both class level (1) and package level (2). I know that the second one is bigger because it checks all methods, including extensions....but....
On Thu, Aug 2, 2012 at 9:32 AM, Tudor Girba <[hidden email]> wrote: Hi, Andre Hora _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
Re: MooseChef: to what package do methods belong?
|
|
Great example. I think it is now clear that your expectations do not
match the meaning of FAMIX, at least not when you have a Smalltalk system. Am I right? Cheers, Doru On Thu, Aug 2, 2012 at 11:57 AM, Andre Hora <[hidden email]> wrote: > A simple example. > > If you create a model with the package Fame-Util, then, inspect such > package: > > "1" > (self classes flatCollectAsSet: [ :each | each queryOutgoingDependencies > atPackageScope ]) size " = 7" > > "2" > (self queryOutgoingDependencies atPackageScope asSet) size " = 12" > > What in a first moment is strange. At least when we are implementing > metrics.. > We would expect the same outgoing references (same size) bor both class > level (1) and package level (2). > I know that the second one is bigger because it checks all methods, > including extensions....but.... > > > On Thu, Aug 2, 2012 at 9:32 AM, Tudor Girba <[hidden email]> wrote: >> >> Hi, >> >> I still do not understand :). >> >> Let's be more specific. As I understand we could use the following >> example: >> - package A = FAMIX-Core >> - package B = Mondrian >> - package C = Moose-Finder >> >> The Moose-Finder extends FAMIX entities with visualization methods >> that depend on Mondrian. This decision was taken exactly so that FAMIX >> does not depend on Mondrian. So, why would you want to see FAMIX-Core >> depend on Mondrian? >> >> Cheers, >> Doru >> >> >> On Thu, Aug 2, 2012 at 9:24 AM, Nicolas Anquetil >> <[hidden email]> wrote: >> > >> > we work on coupling metrics that measure how much a package depends on >> > other >> > packages (can you reuse it without the others). >> > >> > Even if it is trough a method added by extension, if package A use >> > something >> > in package B, then A needs B (or if you prefer A+C needs B) >> > The current semantic does not show this because the extending method >> > will be >> > part of package C (the one that introduces the extension) and therefore >> > C >> > will depend on B (which seem correct to me), but A will not depend on B >> > ... >> > >> > So with extensions, we can consider that both A and C depend on B. >> > This is something that the coupling metrics do not consider. >> > So we have to make a choice. >> > For coherence with the same experiment on Java systems, we prefer to say >> > that only A depends on B. >> > >> > But tis is not how famix and moose chef see things ... >> > >> > Additionaly, we have a metric whose results are useless in the presence >> > of >> > extensions. >> > >> > nicolas >> > >> > >> > On 01/08/12 20:39, Tudor Girba wrote: >> >> >> >> Hi Nicolas, >> >> >> >> The current semantics for both methods and queryOutgoingReferences are >> >> quite fine. >> >> >> >> As I understand you want to count the amount of references that the >> >> code >> >> inside a package induces. So, this includes both the extension methods >> >> and >> >> the defined methods. Why do you see this as a problem? >> >> >> >> Cheers, >> >> Doru >> >> >> >> >> >> On 31 Jul 2012, at 16:21, Nicolas Anquetil wrote: >> >> >> >>> Another MooseChef decision: >> >>> >> >>> aPackage queryOutgoingReferences >> >>> >> >>> returns all the FamixReferences made from aPackage. >> >>> >> >>> this is implemented by taking all methods (self methods) in the >> >>> package >> >>> and asking them what reference they make. >> >>> >> >>> And we have >> >>> FAMIXPackage>>methods >> >>> ^ self localMethods union: self extensionMethods >> >>> >> >>> So when asking a package what external references it makes, we get >> >>> references stemming from methods that extend classes of other >> >>> packages. >> >>> >> >>> We found this while computing a coupling metric for packages. >> >>> >> >>> So the solution would be easy to implement, but we are not sure what >> >>> is >> >>> the best solution ? >> >>> >> >>> - we could change FAMIXPackage>>methods (methods belong to the package >> >>> that owns their class) >> >>> - we could change queryOutgoingReferences >> >>> (references/invocation/accesses >> >>> stem from the methods in the classes in the package, not from the >> >>> methods in >> >>> the package) >> >>> - or we could change our metric >> >>> >> >>> nicolas >> >>> _______________________________________________ >> >>> Moose-dev mailing list >> >>> [hidden email] >> >>> https://www.iam.unibe.ch/mailman/listinfo/moose-dev >> >> >> >> -- >> >> www.tudorgirba.com >> >> >> >> "Yesterday is a fact. >> >> Tomorrow is a possibility. >> >> Today is a challenge." >> >> >> >> >> >> >> >> >> >> _______________________________________________ >> >> Moose-dev mailing list >> >> [hidden email] >> >> https://www.iam.unibe.ch/mailman/listinfo/moose-dev >> > >> > >> > _______________________________________________ >> > Moose-dev mailing list >> > [hidden email] >> > https://www.iam.unibe.ch/mailman/listinfo/moose-dev >> >> >> >> -- >> www.tudorgirba.com >> >> "Every thing has its own flow" >> _______________________________________________ >> Moose-dev mailing list >> [hidden email] >> https://www.iam.unibe.ch/mailman/listinfo/moose-dev > > > > > -- > Andre Hora > > > _______________________________________________ > Moose-dev mailing list > [hidden email] > https://www.iam.unibe.ch/mailman/listinfo/moose-dev > -- www.tudorgirba.com "Every thing has its own flow" _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
Re: MooseChef: to what package do methods belong?
|
|
Yes. And that's why I have said "in a first moment", because for this behavior is quite ok for a Smalltalk model.
So, I think we need to adapt our metric and not to touch in MooseChef..:)
On Thu, Aug 2, 2012 at 3:08 PM, Tudor Girba <[hidden email]> wrote: Great example. I think it is now clear that your expectations do not Andre Hora _______________________________________________ Moose-dev mailing list [hidden email] https://www.iam.unibe.ch/mailman/listinfo/moose-dev |
«
Return to Moose
|
1 view|%1 views
| Free forum by Nabble | Edit this page |