Overview of the Refactoring Discovering Problem

Refactoring support is currently aimed at automatically finding and applying individual transformation steps. Some techniques exist to suggest wider design changes that can be made to a system in order to improve certain design characteristics. Therefore, the sequence of refactorings needed to perform the proposed changes must be found by the developer. Support for complex refactoring sequences generation and execution could help on high-level software systems redesigning, allowing the inclusion of very high level refactorings as a regular technique in the software development process. This paper presents which formalisms, techniques and tools we are exploring to address this problem, and shows the conclusions reached so far. Presented at ECOOP 2006 Doctoral Symposium and PhD Students Workshop (http://www.emn.fr/z-info/ecoop2006/phd.html).

Transcript

1. Overview of the Refactoring Discovering Problem Javier Pérez jperez@infor.uva.es Universidad de Valladolid ECOOP 2006, Doctoral Symposium and PhD Students Workshop Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 1 / 65
2. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 2 / 65
3. Introduction Introduction Refactorings are structural transformations that can be applied to a software system to perform design changes without modifying its behaviour. Current approaches to improve a system design with refactorings focus in: Individual refactoring steps. Detecting refactoring opportunities. Executing the refactoring with a tool. Some techniques can suggest wider design changes: Formal Concept Analysis to propose hierarchy reorganization. Metrics to detect Bad Smells. They require sequences of refactorings to perform the proposed change. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 3 / 65
4. Introduction Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 4 / 65
5. Introduction Main Goals 1 To automatically and dinamically generate a refactoring sequence (refactoring plan) that can transform a system, following a redesing proposal, and preserving the system’s behaviour. 2 To provide very high level (big) refactorings for design improvement, using refactoring plan generation altogether with analysis techniques that suggest redesign proposals. 3 To support this “refactoring plan” technique with tool prototypes to provide it as a regular design improving development technique. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 5 / 65
6. Introduction Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 6 / 65
7. Introduction Refactoring Sequences Problem Overview System Analysis Refactoring Plan Execution Refactoring Analysis Current System System Redesign Proposal Refactoring Plan Desirable System Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 7 / 65
8. Introduction Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 8 / 65
9. Refactoring Plans Defining Refactoring Plans Refactoring Plans We pretend to introduce a new concept: Refactoring Plans Definition A Refactoring Plan will be a specification of a refactoring sequence which matches a system redesign proposal, so that it can be automatically executed to modify the system in order to obtain that desirable system redesign without changing the system’s behaviour. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 9 / 65
10. Refactoring Plans Defining Refactoring Plans Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 10 / 65
11. Refactoring Plans The Problem of Automatic Generation of Refactoring Plans Refactoring Plan Questions Given a software system as the source of the transformation, a redesign proposal, and a set of refactorings that can be used as transformation operations: 1 Does a refactoring plan, which transforms the source, according to the redesign proposal, using the provided refactorings, exist? 2 When a refactoring plan exists, can it be generated and executed automatically? Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 11 / 65
12. Refactoring Plans The Problem of Automatic Generation of Refactoring Plans Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 12 / 65
13. Refactoring Plans The Problem of Automatic Generation of Refactoring Plans Subproblems Definition and formalization of the “Refactoring Plan” concept. Representation of Software Formalization of Refactorings Elaboration of a “refactoring sequences” discovering algorithm. Validation of proposed subproblem solutions through existing tools and prototype implementation. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 13 / 65
14. Refactoring Plans The Problem of Automatic Generation of Refactoring Plans Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 14 / 65
15. Refactoring Plans Research Strategy Research Strategy To reduce the problem complexity given by: differences between source and redesign proposal representation type, abstraction level, description language, . . . uncertainty about what kind of redesign proposal descriptions will we be capable to deal with. We have planned the research to progress through two stages: First, we will apply some restrictions to the problem and propose a restricted solution. Then, we will “open up” those restrictions to find a more general solution. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 15 / 65
16. Refactoring Plans Research Strategy Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 16 / 65
17. Refactoring Plans Research Strategy Research Strategy: First Stage We apply some restrictions to reduce complexity: The transformation source is the current system code. The redesign proposal is the modified system code. Goals of this stage are: To propose a solution to the restricted refactoring plan problem. To validate this solution. Results obtained so far can be applied to other problems: testing wether an evolved system preserves the behaviour of the original system. documenting refactoring changes performed in a redesigned system. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 17 / 65
18. Refactoring Plans Research Strategy Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 18 / 65
19. Refactoring Plans Research Strategy Research Strategy: Second Stage After validating the first stage restricted solution, we will reduce restrictions to generalise the problem: The transformation source will still be the current system code. The redesign proposal will be allowed to be of different type, abstraction level, description language, . . . , than the source. Redesign proposals in the second stage will contain less information than redesign proposal in the first stage (modified system code). Goals of this stage are: To specify which kind of descriptions will we be capable to deal with and would be allowed as redesign proposals. To propose a general solution to the problem of automatic elaboration of refactoring plans. To develop tool prototypes. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 19 / 65
20. Refactoring Plans Research Strategy Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 20 / 65
21. Refactoring Plans Research Strategy Research Strategy: Progress so far We are at the first stage of the research, and working on these subproblems: Refactoring formalization Software representation “refactoring sequences discovering” algorithm I will present the current progress of each one. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 21 / 65
22. Refactoring Plans Research Strategy Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 22 / 65
23. Problem Formalization Refactoring Formalization Any refactoring formalization method must allow: to deal with system structure. to check behaviour preserving conditions. We will use Graph Transformations because: Representing and managing structural information is straightforward with graphs. This approach has already been validated (Mens et al.). With Graph Transformation: Software is represented as graphs Refactorings are represented as graph transformation rules. Other refactoring formalization approaches: Köch, Kniesel: First Order Logic. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 23 / 65
24. Problem Formalization Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 24 / 65
25. Problem Formalization Refactoring and Graph Transformation We have found two main directions within the field of graph transformations, that can be useful: Rule Driven Systems Programmed Graph Rewriting Systems Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 25 / 65
26. Problem Formalization Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 26 / 65
27. Problem Formalization Rule Driven Systems Rule Driven Systems Rule Driven Systems are graph rewriting systems where: transformation rules are described by a graph grammar transformations follow known derivation sequences rules are randomly selected to automatically transform a graph The problem was modeled as a formal language problem: Current system ⇒ starting node Refactorings ⇒ production rules Desirable system ⇒ final node Does the plan exist? ⇒ membership problem Refactoring plan ⇒ derivation sequence Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 27 / 65
28. Problem Formalization Rule Driven Systems Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 28 / 65
29. Problem Formalization Rule Driven Systems Rule Driven Systems We explored this approach and found some issues: This approach deals well with problems for which derivation paths are well defined (visual languages parsing). A well known derivation tree for refactoring sequences should be needed The number of refactorings which can be applied for each derivation step is unpredictable We used the AGG graph transformation tool. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 29 / 65
30. Problem Formalization Rule Driven Systems Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 30 / 65
31. Problem Formalization Programmed Graph Rewriting Systems Programmed Graph Rewriting Systems PGR Systems present a more general graph rewriting approach: graph grammars are also used they include structured programming transformations can be programmed and organized in modules The problem can be seen as a state space search problem: Current system ⇒ starting state Refactorings ⇒ state changing operations Desirable System ⇒ a desired state Does the plan exist? ⇒ reachability of the desired state Refactoring Plan ⇒ path to the final state Issues found so far: the combinatorial explosion the need to develop heuristics We will use PROGRES, a Programmed Graph Rewriting tool. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 31 / 65
32. Problem Formalization Programmed Graph Rewriting Systems Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 32 / 65
33. Problem Formalization Programmed Graph Rewriting Systems Refactoring Formalization with PGR Systems After exploring both possibilities, we have choose Programmed Graph Rewriting Systems because: it offers progarmmable control over the graph transformation process it offers more expressiveness than the grammar based systems refactorings which take multiple transformation steps are very difficult to describe with Rule Driven Systems, that refactorings can be described easily with PGR Systems In the first stage of the research we are abording the already mentioned subproblems according to the PGR systems paradigm. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 33 / 65
34. Problem Formalization Programmed Graph Rewriting Systems Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 34 / 65
35. Software Representation Software Representation: Program Graphs A graph representation for Object-Oriented Software is needed. We must represent: elements of OO paradigm (classes, fields, methods, ...) structural relationships method bodies We have choose the software representation part from the refactoring formalization of Mens. This representation: uses directed type graphs is language independent, lacking specific language constructions has been simplified to be as flexible as possible Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 35 / 65
36. Software Representation Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 36 / 65
37. Software Representation Software Representation: Java Program Graphs For real systems, it is necessary to extend the graph format, adding: elements for specific languages more detailed representation of method bodies We have extended program graphs for Java: Java Program Graphs. Our graph representation format adds: Java concepts such as visibility, interfaces, packages, . . . More detailed representation of method bodies, with new node types, attributes and relationships. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 37 / 65
38. Software Representation Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 38 / 65
39. Refactoring Rules and Refactoring Discovering Algorithm Refactoring Rules Description (MODIFICAR) Refactoring rules descriptions should be decomposable for the algorithm can use each part it needs. Programmed Graph Rewriting allows to: Split refactoring descriptions in multiple parts: pre and postconditions transformation process transformation relationships between pre and post conditions. Describe structured conditions: if...then...else Algorithmically describe refactorings that take multiple transformation steps: Pull Up Method We are describing refactoring rules that way. Example: Change Interface Type Parameter. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 40 / 65
40. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 41 / 65
41. Refactoring Rules and Refactoring Discovering Algorithm Change Interface Type Parameter M 1 Preconditions: Postconditions: Transformation: 1) 2) 5) 6) 4) 3) 5) 4) I C I P 2 3 M 1 p t C P 2 4 M 1 p t I 3 C 4 i* I P 2 3 M 1 p t X I 3 C 4 i* C P 2 4 M 1 p t I 3 i* C P 2 4 M 1 p t I 3 i* M 1 1) 2) 3) I C 3 3 4 4 5 4,5 1 2 3 1 2 3 4 6 pre post Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 42 / 65
42. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 43 / 65
43. Refactoring Rules and Refactoring Discovering Algorithm Refactoring Sequences Discovering Algorithm A basic state space search algorithm is being developed. It needs refactoring descriptions to be expressed in terms of preconditions, transformation and postconditions. It needs the expressiveness and execution control of programmed graph rewriting. Main ideas of the algorithm are: The Algorithm is guided by pre and postconditions. To look for preconditions in the source graph: will reveal which refactorings can be applied. To look for postconditions in the target graph: will find which refactorings are more likely to be part of the sequence. The source graph gets transformed progressively into the target graph. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 44 / 65
44. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 45 / 65
45. Refactoring Rules and Refactoring Discovering Algorithm Refactoring Sequences Discovering Algorithm (MODIFICAR) Iterate select a refactoring (backtracking point) apply it to the source graph Until Source and Target graphs are isomorphic Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 47 / 65
46. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 48 / 65
47. Refactoring Rules and Refactoring Discovering Algorithm Refactoring Sequences Discovering Algorithm (MODIFICAR) Select Refactoring Look for preconditions in the source graph Select refactorings which preconditions hold Look for postconditions in the target graph Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 50 / 65
48. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 51 / 65
49. Refactoring Rules and Refactoring Discovering Algorithm Refactoring Sequences Discovering Algorithm (MODIFICAR) Apply Refactoring If postconditions hold in the target graph apply refactoring else apply refactoring save postconditions not held as awaiting postonditions If awaiting postconditions are no longer being held on Remove awaiting postcondition Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 53 / 65
50. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 54 / 65
51. Refactoring Rules and Refactoring Discovering Algorithm Refactoring Sequences Discovering Algorithm (MODIFICAR) Check Termination Source and Target Graphs must be isomorphic There must be no awaiting postconditions Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 56 / 65
52. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 57 / 65
53. Refactoring Rules and Refactoring Discovering Algorithm Heuristics Possible heuristics to use in the refactoring sequences discovering algorithm: Prioritize refactorings which postconditions hold on the target graph. Prioritize refactorings which make that previous selected. refactorings could hold their awaiting postconditions Use metrics to ponderate refactorings? Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 58 / 65
54. Refactoring Rules and Refactoring Discovering Algorithm Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 59 / 65
55. Conclusions and Future Work Conclusions Automatic generation of refactoring plans will provide very high level refactorings to improve design of existing code. Main subproblems and research strategy have been introduced Graph transformation can be used as the underlying formalism, specifically the programmed graph rewriting approach. Representing Java programs with Java Program Graphs. Describing refactoring rules with programmed graph transformation rules in terms of pre, postconditions and transformations. The problem can be modeled as a state space search problem. Using a “refactoring sequences discovering” algorithm for searching. Using heuristics to guide the algorithm. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 60 / 65
56. Conclusions and Future Work Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 61 / 65
57. Conclusions and Future Work Doubts and known limitations/problems “Rename” Refactorings: We are afraid that rename refactorings could not be detected with graph transformation. Correction and completeness of the algorithm: Heuristics could not be enough to prevent the algorithm getting lost in the search. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 62 / 65
58. Conclusions and Future Work Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 63 / 65
59. Conclusions and Future Work Future Work Main future tasks will be directed to: Further definition of the “Refactoring Plan” concept Extend the available set of formalized refactorings. Validate proposals implementing the algorithm in the PROGRES tool. Analyse termination and correctness conditions of the refactoring discovering algorithm. Analyse, improve and extend heuristics of the state space search algorithm. Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 64 / 65
60. Conclusions and Future Work Javier Pérez (Universidad de Valladolid) The Refactoring Discovering Problem ECOOP’06 PhD S. Workshop 65 / 65