Using Candidate - Moose-dev

Simon Denier

4 May 4 May

9:58 p.m.

On 4 mai 2011, at 22:22, Alexandre Bergel wrote:

Hi! Having a proper list of candidates when importing Smalltalk code is crucial. Without this, it is not possible to seriously analyze Smalltalk code. How to use the Roel Typer? I guess I have to tick "Compute type of attributes" in the wizard. But which strategy to compute invocation candidates should I use? I tried when I get an error from the typer.

These are two different things: the strategy to compute candidates is selected in the last combo box in the wizard (below importing context). Currently there are different strategies, from no candidate to sure candidates and some more which are not clearly defined (they lack tests). 'Compute type of attributes' is clearly different since it will compute the FAMIXAttribute>>declaredType property (which is often nil in our Smalltalk models). Now, using the inferred type to refine candidates is of course the next step. They were some tentative in this direction but I don't think we went very far (because tests are seriously needed in this area to gain confidence: we try to do some heuristics on top of other heuristics). Could you review AbstractCandidateListOperator and its subclasses and tell what direction would be good?

...

We absolutely need to fix this. Cyrille, did you manage to get this to work? Cheers, Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

-- Simon Denier

Reply

Stéphane Ducasse

5 May 5 May

8 a.m.

On May 4, 2011, at 11:56 PM, Alexandre Bergel wrote:

...

Okay, I see better now. AbstractCandidateListOperator seems to use RoelTyper in some methods (e.g., computeInstanceVariableCandidateListFor:) We first need to know whether these methods work or not. Cyrille wrote them, back in 2009. I also wrote a profiler that tells me what are the signature of methods by profiling the unit test execution. Maybe Cyrille can have a look at it?

not really because he has other tasks. We have to get the demo up and running on of these months.

...

Cheers, Alexandre On 4 May 2011, at 15:58, Simon Denier wrote:

On 4 mai 2011, at 22:22, Alexandre Bergel wrote:

Hi! Having a proper list of candidates when importing Smalltalk code is crucial. Without this, it is not possible to seriously analyze Smalltalk code. How to use the Roel Typer? I guess I have to tick "Compute type of attributes" in the wizard. But which strategy to compute invocation candidates should I use? I tried when I get an error from the typer.

These are two different things: the strategy to compute candidates is selected in the last combo box in the wizard (below importing context). Currently there are different strategies, from no candidate to sure candidates and some more which are not clearly defined (they lack tests). 'Compute type of attributes' is clearly different since it will compute the FAMIXAttribute>>declaredType property (which is often nil in our Smalltalk models). Now, using the inferred type to refine candidates is of course the next step. They were some tentative in this direction but I don't think we went very far (because tests are seriously needed in this area to gain confidence: we try to do some heuristics on top of other heuristics). Could you review AbstractCandidateListOperator and its subclasses and tell what direction would be good?

We absolutely need to fix this. Cyrille, did you manage to get this to work? Cheers, Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

-- Simon Denier _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

-- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

Reply

Tudor Girba

11:35 a.m.

Please make sure to write tests for these cases. It is crucial to be able to rely on this importer. Cheers, Doru On 5 May 2011, at 11:20, Cyrille Delaunay wrote:

...

If you get an error when computing candidates , you can report your specific case: http://code.google.com/p/moose-technology/issues/detail?id=586 Because I got some errors too but I managed to make it work now (at least my specific case works). 2011/5/5 Stéphane Ducasse <stephane.ducasse(a)inria.fr> On May 4, 2011, at 11:56 PM, Alexandre Bergel wrote:

Okay, I see better now. AbstractCandidateListOperator seems to use RoelTyper in some methods (e.g., computeInstanceVariableCandidateListFor:) We first need to know whether these methods work or not. Cyrille wrote them, back in 2009. I also wrote a profiler that tells me what are the signature of methods by profiling the unit test execution. Maybe Cyrille can have a look at it?

not really because he has other tasks. We have to get the demo up and running on of these months.

Cheers, Alexandre On 4 May 2011, at 15:58, Simon Denier wrote:

On 4 mai 2011, at 22:22, Alexandre Bergel wrote:

Hi! Having a proper list of candidates when importing Smalltalk code is crucial. Without this, it is not possible to seriously analyze Smalltalk code. How to use the Roel Typer? I guess I have to tick "Compute type of attributes" in the wizard. But which strategy to compute invocation candidates should I use? I tried when I get an error from the typer.

These are two different things: the strategy to compute candidates is selected in the last combo box in the wizard (below importing context). Currently there are different strategies, from no candidate to sure candidates and some more which are not clearly defined (they lack tests). 'Compute type of attributes' is clearly different since it will compute the FAMIXAttribute>>declaredType property (which is often nil in our Smalltalk models). Now, using the inferred type to refine candidates is of course the next step. They were some tentative in this direction but I don't think we went very far (because tests are seriously needed in this area to gain confidence: we try to do some heuristics on top of other heuristics). Could you review AbstractCandidateListOperator and its subclasses and tell what direction would be good?

We absolutely need to fix this. Cyrille, did you manage to get this to work? Cheers, Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

-- Simon Denier _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

-- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;. _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

_______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev

-- www.tudorgirba.com "Presenting is storytelling."

Reply