1 \documentclass[USenglish,11pt]{ifimaster}
3 \usepackage[utf8]{inputenc}
4 \usepackage[T1]{fontenc,url}
5 \usepackage{lmodern} % using Latin Modern to be able to use bold typewriter font
12 \usepackage{tikz-qtree}
13 \usetikzlibrary{shapes,snakes,trees,arrows,shadows,positioning,calc}
14 \usepackage{babel,textcomp,csquotes,ifimasterforside}
17 \usepackage[hidelinks]{hyperref}
19 \usepackage[xindy]{glossaries}
21 \usepackage[style=alphabetic,backend=biber,doi=false,isbn=false]{biblatex}
23 \usepackage{mathtools}
25 % use 'disable' before printing:
26 \usepackage[]{todonotes}
34 \def\mintedframesep{11pt}
36 \usepackage{perpage} %the perpage package
37 \MakePerPage{footnote} %the perpage package command
39 \theoremstyle{definition}
40 \newtheorem*{wordDef}{Definition}
41 \newtheorem*{theorem}{Theorem}
43 \graphicspath{ {./figures/} }
45 \newcommand{\citing}[1]{~\cite{#1}}
46 %\newcommand{\myref}[1]{\cref{#1} on \cpageref{#1}}
47 \newcommand{\myref}[1]{\vref{#1}}
48 \newcommand{\Myref}[1]{\Vref{#1}}
49 \newcommand{\mysimpleref}[1]{\cref{#1}}
50 \newcommand{\Mysimpleref}[1]{\Cref{#1}}
52 %\newcommand{\glossref}[1]{\textsuperscript{(\glsrefentry{#1})}}
53 %\newcommand{\gloss}[1]{\gls{#1}\glossref{#1}}
54 %\newcommand{\glosspl}[1]{\glspl{#1}\glossref{#1}}
55 \newcommand{\gloss}[1]{\gls{#1}}
56 \newcommand{\glosspl}[1]{\glspl{#1}}
58 \newcommand{\definition}[1]{\begin{wordDef}#1\end{wordDef}}
59 \newcommand{\see}[1]{(see \myref{#1})}
60 \newcommand{\explanation}[3]{\noindent\textbf{\textit{#1}}\\*\emph{When:}
61 #2\\*\emph{How:} #3\\*[-7px]}
63 %\newcommand{\type}[1]{\lstinline{#1}}
64 \newcommand{\code}[1]{\texttt{\textbf{#1}}}
65 \newcommand{\type}[1]{\code{#1}}
66 \newcommand{\typeref}[1]{\footnote{\type{#1}}}
67 \newcommand{\typewithref}[2]{\type{#2}\typeref{#1.#2}}
68 \newcommand{\method}[1]{\type{#1}}
69 \newcommand{\methodref}[2]{\footnote{\type{#1}\method{\##2()}}}
70 \newcommand{\methodwithref}[2]{\method{#2}\footnote{\type{#1}\method{\##2()}}}
71 \newcommand{\var}[1]{\type{#1}}
73 \newcommand{\name}[1]{#1}
74 \newcommand{\tit}[1]{\emph{#1}}
75 \newcommand{\refa}[1]{\emph{#1}}
76 \newcommand{\pattern}[1]{\emph{#1}}
77 \newcommand{\metr}[1]{\emph{#1}}
78 \newcommand{\ExtractMethod}{\refa{Extract Method}\xspace}
79 \newcommand{\MoveMethod}{\refa{Move Method}\xspace}
80 \newcommand{\ExtractAndMoveMethod}{\refa{Extract and Move Method}\xspace}
82 \newcommand{\m}[1]{$#1$}
84 \newcommand\todoin[2][]{\todo[inline, caption={#2}, #1]{
85 \begin{minipage}{\textwidth-4pt}#2\end{minipage}}}
87 \title{Automated Composition of Refactorings}
88 \subtitle{Implementing and evaluating a search-based Extract and Move Method
90 \author{Erlend Kristiansen}
93 \newglossaryentry{profiling}
96 description={is to run a computer program through a profiler/with a profiler
99 \newglossaryentry{profiler}
102 description={A profiler is a program for analyzing performance within an
103 application. It is used to analyze memory consumption, processing time and
104 frequency of procedure calls and such}
106 \newglossaryentry{xUnit}
108 name={xUnit framework},
109 description={An xUnit framework is a framework for writing unit tests for a
110 computer program. It follows the patterns known from the JUnit framework for
111 Java\citing{fowlerXunit}
113 plural={xUnit frameworks}
115 \newglossaryentry{softwareObfuscation}
117 name={software obfuscation},
118 description={makes source code harder to read and analyze, while preserving
121 \newglossaryentry{extractClass}
123 name=\refa{Extract Class},
124 description={The \refa{Extract Class} refactoring works by creating a class,
125 for then to move members from another class to that class and access them from
126 the old class via a reference to the new class}
128 \newglossaryentry{designPattern}
130 name={design pattern},
131 description={A design pattern is a named abstraction that is meant to solve a
132 general design problem. It describes the key aspects of a common problem and
133 identifies its participators and how they collaborate},
134 plural={design patterns}
136 \newglossaryentry{enclosingClass}
138 name={enclosing class},
139 description={An enclosing class is the class that surrounds any specific piece
140 of code that is written in the inner scope of this class},
142 \newglossaryentry{mementoPattern}
144 name={memento pattern},
145 description={The memento pattern is a software design pattern that is used to
146 capture an object's internal state so that it can be restored to this state
147 later\citing{designPatterns}},
149 %\newglossaryentry{extractMethod}
151 % name=\refa{Extract Method},
152 % description={The \refa{Extract Method} refactoring is used to extract a
153 %fragment of code from its context and into a new method. A call to the new
154 %method is inlined where the fragment was before. It is used to break code into
155 %logical units, with names that explain their purpose}
157 %\newglossaryentry{moveMethod}
159 % name=\refa{Move Method},
160 % description={The \refa{Move Method} refactoring is used to move a method from
161 % one class to another. This is useful if the method is using more features of
162 % another class than of the class which it is currently defined. Then all calls
163 % to this method must be updated, or the method must be copied, with the old
164 %method delegating to the new method}
167 \bibliography{bibliography/master-thesis-erlenkr-bibliography}
168 \DefineBibliographyStrings{english}{%
169 bibliography = {References},
171 \newbibmacro{string+doi}[1]{%
172 \iffieldundef{doi}{#1}{\href{http://dx.doi.org/\thefield{doi}}{#1}}}
173 \DeclareFieldFormat{title}{\usebibmacro{string+doi}{\mkbibemph{#1}}}
174 \DeclareFieldFormat[article]{title}{\usebibmacro{string+doi}{\mkbibquote{#1}}}
176 % UML comment in TikZ:
177 % ref: https://tex.stackexchange.com/questions/103688/folded-paper-shape-tikz
179 \pgfdeclareshape{umlcomment}{
180 \inheritsavedanchors[from=rectangle] % this is nearly a rectangle
181 \inheritanchorborder[from=rectangle]
182 \inheritanchor[from=rectangle]{center}
183 \inheritanchor[from=rectangle]{north}
184 \inheritanchor[from=rectangle]{south}
185 \inheritanchor[from=rectangle]{west}
186 \inheritanchor[from=rectangle]{east}
187 % ... and possibly more
188 \backgroundpath{% this is new
189 % store lower right in xa/ya and upper right in xb/yb
190 \southwest \pgf@xa=\pgf@x \pgf@ya=\pgf@y
191 \northeast \pgf@xb=\pgf@x \pgf@yb=\pgf@y
192 % compute corner of ‘‘flipped page’’
193 \pgf@xc=\pgf@xb \advance\pgf@xc by-10pt % this should be a parameter
194 \pgf@yc=\pgf@yb \advance\pgf@yc by-10pt
195 % construct main path
196 \pgfpathmoveto{\pgfpoint{\pgf@xa}{\pgf@ya}}
197 \pgfpathlineto{\pgfpoint{\pgf@xa}{\pgf@yb}}
198 \pgfpathlineto{\pgfpoint{\pgf@xc}{\pgf@yb}}
199 \pgfpathlineto{\pgfpoint{\pgf@xb}{\pgf@yc}}
200 \pgfpathlineto{\pgfpoint{\pgf@xb}{\pgf@ya}}
203 \pgfpathmoveto{\pgfpoint{\pgf@xc}{\pgf@yb}}
204 \pgfpathlineto{\pgfpoint{\pgf@xc}{\pgf@yc}}
205 \pgfpathlineto{\pgfpoint{\pgf@xb}{\pgf@yc}}
206 \pgfpathlineto{\pgfpoint{\pgf@xc}{\pgf@yc}}
211 \tikzstyle{comment}=[%
224 %\interfootnotelinepenalty=10000
226 % Space between table rows
227 \renewcommand{\arraystretch}{1.3}
229 \newcommand{\spancols}[2]{\multicolumn{#1}{@{}l@{}}{#2}}
231 \newcolumntype{L}[1]{>{\hsize=#1\hsize\raggedright\arraybackslash}X}%
232 \newcolumntype{R}[1]{>{\hsize=#1\hsize\raggedleft\arraybackslash}X}%
236 %\pagenumbering{arabic}
241 %\setcounter{page}{3}
244 \todoin{\textbf{Remove all todos (including list) before delivery/printing!!!
245 Can be done by removing ``draft'' from documentclass.}}
246 \todoin{Write abstract}
254 %\setcounter{page}{13}
256 \chapter{Introduction}
258 \section{Motivation and structure}
260 For large software projects, complex program source code is an issue. It impacts
261 the cost of maintenance in a negative way. It often stalls the implementation of
262 new functionality and other program changes. The code may be difficult to
263 understand, the changes may introduce new bugs that are hard to find and its
264 complexity can simply keep people from doing code changes in fear of breaking
265 some dependent piece of code. All these problems are related, and often lead to
266 a vicious circle that slowly degrades the overall quality of a project.
268 More specifically, and in an object-oriented context, a class may depend on a
269 number of other classes. Sometimes these intimate relationships are appropriate,
270 and sometimes they are not. Inappropriate \emph{coupling} between classes can
271 make it difficult to know whether or not a change that is aimed at fixing a
272 specific problem also alters the behavior of another part of a program.
274 One of the tools that are used to fight complexity and coupling in program
275 source code is \emph{refactoring}. The intention for this master's thesis is
276 therefore to create an automated composite refactoring that reduces coupling
277 between classes. The refactoring shall be able to operate automatically in all
278 phases of a refactoring, from performing analysis to executing changes. It is
279 also a requirement that it should be able to process large quantities of source
280 code in a reasonable amount of time.
282 The current chapter proceeds in \mysimpleref{sec:refactoring} by describing what
283 refactoring is. Then the project is presented in \mysimpleref{sec:project},
284 before the chapter is concluded with a brief discussion of related work in
285 \mysimpleref{sec:relatedWork}.
287 \Mysimpleref{ch:extractAndMoveMethod} shows the workings of our refactoring
288 together with a simple example illustrating this.
290 \todoin{Structure. Write later\ldots}
293 \section{What is refactoring?}\label{sec:refactoring}
295 This question is best answered by first defining the concept of a
296 \emph{refactoring}, what it is to \emph{refactor}, and then discuss what aspects
297 of programming make people want to refactor their code.
299 \subsection{Defining refactoring}
300 Martin Fowler, in his classic book on refactoring\citing{refactoring}, defines a
301 refactoring like this:
304 \emph{Refactoring} (noun): a change made to the internal
305 structure\footnote{The structure observable by the programmer.} of software to
306 make it easier to understand and cheaper to modify without changing its
307 observable behavior.~\cite[p.~53]{refactoring}
310 \noindent This definition assigns additional meaning to the word
311 \emph{refactoring}, beyond the composition of the prefix \emph{re-}, usually
312 meaning something like ``again'' or ``anew'', and the word \emph{factoring},
313 which can mean to isolate the \emph{factors} of something. Here a \emph{factor}
314 would be close to the mathematical definition of something that divides a
315 quantity, without leaving a remainder. Fowler is mixing the \emph{motivation}
316 behind refactoring into his definition. Instead it could be more refined, formed
317 to only consider the \emph{mechanical} and \emph{behavioral} aspects of
318 refactoring. That is to factor the program again, putting it together in a
319 different way than before, while preserving the behavior of the program. An
320 alternative definition could then be:
322 \definition{A \emph{refactoring} is a transformation
323 done to a program without altering its external behavior.}
325 From this we can conclude that a refactoring primarily changes how the
326 \emph{code} of a program is perceived by the \emph{programmer}, and not the
327 \emph{behavior} experienced by any user of the program. Although the logical
328 meaning is preserved, such changes could potentially alter the program's
329 behavior when it comes to performance gain or -penalties. So any logic depending
330 on the performance of a program could make the program behave differently after
333 In the extreme case one could argue that \gloss{softwareObfuscation} is
334 refactoring. It is often used to protect proprietary software. It restrains
335 uninvited viewers, so they have a hard time analyzing code that they are not
336 supposed to know how works. This could be a problem when using a language that
337 is possible to decompile, such as Java.
339 Obfuscation could be done composing many, more or less randomly chosen,
340 refactorings. Then the question arises whether it can be called a
341 \emph{composite refactoring} or not \see{compositeRefactorings}? The answer is
342 not obvious. First, there is no way to describe the mechanics of software
343 obfuscation, because there are infinitely many ways to do that. Second,
344 obfuscation can be thought of as \emph{one operation}: Either the code is
345 obfuscated, or it is not. Third, it makes no sense to call software obfuscation
346 \emph{a refactoring}, since it holds different meaning to different people.
348 This last point is important, since one of the motivations behind defining
349 different refactorings, is to establish a \emph{vocabulary} for software
350 professionals to use when reasoning about and discussing programs, similar to
351 the motivation behind \glosspl{designPattern}\citing{designPatterns}.
353 So for describing \emph{software obfuscation}, it might be more appropriate to
354 define what you do when performing it rather than precisely defining its
355 mechanics in terms of other refactorings.
358 \subsection{The etymology of 'refactoring'}
359 It is a little difficult to pinpoint the exact origin of the word
360 ``refactoring'', as it seems to have evolved as part of a colloquial
361 terminology, more than a scientific term. There is no authoritative source for a
362 formal definition of it.
364 According to Martin Fowler\citing{etymology-refactoring}, there may also be more
365 than one origin of the word. The most well-known source, when it comes to the
366 origin of \emph{refactoring}, is the
367 Smalltalk\footnote{\label{footNote}Programming language} community and their
368 infamous \name{Refactoring
369 Browser}\footnote{\url{http://st-www.cs.illinois.edu/users/brant/Refactory/RefactoringBrowser.html}}
370 described in the article \tit{A Refactoring Tool for
371 Smalltalk}\citing{refactoringBrowser1997}, published in 1997.
372 Allegedly\citing{etymology-refactoring}, the metaphor of factoring programs was
373 also present in the Forth\textsuperscript{\ref{footNote}} community, and the
374 word ``refactoring'' is mentioned in a book by Leo Brodie, called \tit{Thinking
375 Forth}\citing{brodie2004}, first published in 1984\footnote{\tit{Thinking Forth}
376 was first published in 1984 by the \name{Forth Interest Group}. Then it was
377 reprinted in 1994 with minor typographical corrections, before it was
378 transcribed into an electronic edition typeset in \LaTeX\ and published under a
379 Creative Commons license in
380 2004. The edition cited here is the 2004 edition, but the content should
381 essentially be as in 1984.}. The exact word is only printed one
382 place~\cite[p.~232]{brodie2004}, but the term \emph{factoring} is prominent in
383 the book, which also contains a whole chapter dedicated to (re)factoring, and
384 how to keep the (Forth) code clean and maintainable.
387 \ldots good factoring technique is perhaps the most important skill for a
388 Forth programmer.~\cite[p.~172]{brodie2004}
391 \noindent Brodie also express what \emph{factoring} means to him:
394 Factoring means organizing code into useful fragments. To make a fragment
395 useful, you often must separate reusable parts from non-reusable parts. The
396 reusable parts become new definitions. The non-reusable parts become arguments
397 or parameters to the definitions.~\cite[p.~172]{brodie2004}
400 Fowler claims that the usage of the word \emph{refactoring} did not pass between
401 the \name{Forth} and \name{Smalltalk} communities, but that it emerged
402 independently in each of the communities.
404 \subsection{Reasons for refactoring}
405 There are many reasons why people want to refactor their programs. They can for
406 instance do it to remove duplication, break up long methods or to introduce
407 design patterns into their software systems. The shared trait for all these is
408 that peoples' intentions are to make their programs \emph{better}, in some
409 sense. But what aspects of their programs are becoming improved?
411 As just mentioned, people often refactor to get rid of duplication. They are
412 moving identical or similar code into methods, and are pushing methods up or
413 down in their class hierarchies. They are making template methods for
414 overlapping algorithms/functionality, and so on. It is all about gathering what
415 belongs together and putting it all in one place. The resulting code is then
416 easier to maintain. When removing the implicit coupling\footnote{When
417 duplicating code, the duplicate pieces of code might not be coupled, apart
418 from representing the same functionality. So if this functionality is going to
419 change, it might need to change in more than one place, thus creating an
420 implicit coupling between multiple pieces of code.} between code snippets, the
421 location of a bug is limited to only one place, and new functionality need only
422 to be added to this one place, instead of a number of places people might not
425 A problem you often encounter when programming, is that a program contains a lot
426 of long and hard-to-grasp methods. It can then help to break the methods into
427 smaller ones, using the \ExtractMethod refactoring\citing{refactoring}. Then
428 you may discover something about a program that you were not aware of before;
429 revealing bugs you did not know about or could not find due to the complex
430 structure of your program. Making the methods smaller and giving good names to
431 the new ones clarifies the algorithms and enhances the \emph{understandability}
432 of the program \see{magic_number_seven}. This makes refactoring an excellent
433 method for exploring unknown program code, or code that you had forgotten that
436 Most primitive refactorings are simple, and usually involves moving code
437 around\citing{kerievsky2005}. The motivation behind them may first be revealed
438 when they are combined into larger --- higher level --- refactorings, called
439 \emph{composite refactorings} \see{compositeRefactorings}. Often the goal of
440 such a series of refactorings is a design pattern. Thus the design can
441 \emph{evolve} throughout the lifetime of a program, as opposed to designing
442 up-front. It is all about being structured and taking small steps to improve a
445 Many software design pattern are aimed at lowering the coupling between
446 different classes and different layers of logic. One of the most famous is
447 perhaps the \pattern{Model-View-Controller}\citing{designPatterns} pattern. It
448 is aimed at lowering the coupling between the user interface, the business logic
449 and the data representation of a program. This also has the added benefit that
450 the business logic could much easier be the target of automated tests, thus
451 increasing the productivity in the software development process.
453 Another effect of refactoring is that with the increased separation of concerns
454 coming out of many refactorings, the \emph{performance} can be improved. When
455 profiling programs, the problematic parts are narrowed down to smaller parts of
456 the code, which are easier to tune, and optimization can be performed only where
457 needed and in a more effective way\citing{refactoring}.
459 Last, but not least, and this should probably be the best reason to refactor, is
460 to refactor to \emph{facilitate a program change}. If one has managed to keep
461 one's code clean and tidy, and the code is not bloated with design patterns that
462 are not ever going to be needed, then some refactoring might be needed to
463 introduce a design pattern that is appropriate for the change that is going to
466 Refactoring program code --- with a goal in mind --- can give the code itself
467 more value. That is in the form of robustness to bugs, understandability and
468 maintainability. Having robust code is an obvious advantage, but
469 understandability and maintainability are both very important aspects of
470 software development. By incorporating refactoring in the development process,
471 bugs are found faster, new functionality is added more easily and code is easier
472 to understand by the next person exposed to it, which might as well be the
473 person who wrote it. The consequence of this, is that refactoring can increase
474 the average productivity of the development process, and thus also add to the
475 monetary value of a business in the long run. The perspective on productivity
476 and money should also be able to open the eyes of the many nearsighted managers
477 that seldom see beyond the next milestone.
479 \subsection{The magical number seven}\label{magic_number_seven}
480 The article \tit{The magical number seven, plus or minus two: some limits on our
481 capacity for processing information}\citing{miller1956} by George A. Miller,
482 was published in the journal \name{Psychological Review} in 1956. It presents
483 evidence that support that the capacity of the number of objects a human being
484 can hold in its working memory is roughly seven, plus or minus two objects. This
485 number varies a bit depending on the nature and complexity of the objects, but
486 is according to Miller ``\ldots never changing so much as to be
489 Miller's article culminates in the section called \emph{Recoding}, a term he
490 borrows from communication theory. The central result in this section is that by
491 recoding information, the capacity of the amount of information that a human can
492 process at a time is increased. By \emph{recoding}, Miller means to group
493 objects together in chunks, and give each chunk a new name that it can be
497 \ldots recoding is an extremely powerful weapon for increasing the amount of
498 information that we can deal with.~\cite[p.~95]{miller1956}
501 By organizing objects into patterns of ever growing depth, one can memorize and
502 process a much larger amount of data than if it were to be represented as its
503 basic pieces. This grouping and renaming is analogous to how many refactorings
504 work, by grouping pieces of code and give them a new name. Examples are the
505 fundamental \ExtractMethod and \refa{Extract Class}
506 refactorings\citing{refactoring}.
508 An example from the article addresses the problem of memorizing a sequence of
509 binary digits. The example presented here is a slightly modified version of the
510 one presented in the original article\citing{miller1956}, but it preserves the
511 essence of it. Let us say we have the following sequence of
512 16 binary digits: ``1010001001110011''. Most of us will have a hard time
513 memorizing this sequence by only reading it once or twice. Imagine if we instead
514 translate it to this sequence: ``A273''. If you have a background from computer
515 science, it will be obvious that the latter sequence is the first sequence
516 recoded to be represented by digits in base 16. Most people should be able to
517 memorize this last sequence by only looking at it once.
519 Another result from the Miller article is that when the amount of information a
520 human must interpret increases, it is crucial that the translation from one code
521 to another must be almost automatic for the subject to be able to remember the
522 translation, before \heshe is presented with new information to recode. Thus
523 learning and understanding how to best organize certain kinds of data is
524 essential to efficiently handle that kind of data in the future. This is much
525 like when humans learn to read. First they must learn how to recognize letters.
526 Then they can learn distinct words, and later read sequences of words that form
527 whole sentences. Eventually, most of them will be able to read whole books and
528 briefly retell the important parts of its content. This suggests that the use of
529 design patterns is a good idea when reasoning about computer programs. With
530 extensive use of design patterns when creating complex program structures, one
531 does not always have to read whole classes of code to comprehend how they
532 function, it may be sufficient to only see the name of a class to almost fully
533 understand its responsibilities.
536 Our language is tremendously useful for repackaging material into a few chunks
537 rich in information.~\cite[p.~95]{miller1956}
540 Without further evidence, these results at least indicate that refactoring
541 source code into smaller units with higher cohesion and, when needed,
542 introducing appropriate design patterns, should aid in the cause of creating
543 computer programs that are easier to maintain and have code that is easier (and
546 \subsection{Notable contributions to the refactoring literature}
549 \item[1992] William F. Opdyke submits his doctoral dissertation called
550 \tit{Refactoring Object-Oriented Frameworks}\citing{opdyke1992}. This work
551 defines a set of refactorings that are behavior-preserving given that their
552 preconditions are met. The dissertation is focused on the automation of
554 \item[1999] Martin Fowler et al.: \tit{Refactoring: Improving the Design of
555 Existing Code}\citing{refactoring}. This is maybe the most influential text
556 on refactoring. It bares similarities with Opdykes thesis\citing{opdyke1992}
557 in the way that it provides a catalog of refactorings. But Fowler's book is
558 more about the craft of refactoring, as he focuses on establishing a
559 vocabulary for refactoring, together with the mechanics of different
560 refactorings and when to perform them. His methodology is also founded on
561 the principles of test-driven development.
562 \item[2005] Joshua Kerievsky: \tit{Refactoring to
563 Patterns}\citing{kerievsky2005}. This book is heavily influenced by Fowler's
564 \tit{Refactoring}\citing{refactoring} and the ``Gang of Four'' \tit{Design
565 Patterns}\citing{designPatterns}. It is building on the refactoring
566 catalogue from Fowler's book, but is trying to bridge the gap between
567 \emph{refactoring} and \emph{design patterns} by providing a series of
568 higher-level composite refactorings, that makes code evolve toward or away
569 from certain design patterns. The book is trying to build up the reader's
570 intuition around \emph{why} one would want to use a particular design
571 pattern, and not just \emph{how}. The book is encouraging evolutionary
572 design \see{relationToDesignPatterns}.
575 \subsection{Tool support (for Java)}\label{toolSupport}
576 This section will briefly compare the refactoring support of the three IDEs
577 \name{Eclipse}\footnote{\url{http://www.eclipse.org/}}, \name{IntelliJ
578 IDEA}\footnote{The IDE under comparison is the \name{Community Edition},
579 \url{http://www.jetbrains.com/idea/}} and
580 \name{NetBeans}\footnote{\url{https://netbeans.org/}}. These are the most
581 popular Java IDEs\citing{javaReport2011}.
583 All three IDEs provide support for the most useful refactorings, like the
584 different extract, move and rename refactorings. In fact, Java-targeted IDEs are
585 known for their good refactoring support, so this did not appear as a big
588 The IDEs seem to have excellent support for the \ExtractMethod refactoring, so
589 at least they have all passed the first ``refactoring
590 rubicon''\citing{fowlerRubicon2001,secondRubicon2012}.
592 Regarding the \MoveMethod refactoring, the \name{Eclipse} and \name{IntelliJ}
593 IDEs do the job in very similar manners. In most situations they both do a
594 satisfying job by producing the expected outcome. But they do nothing to check
595 that the result does not break the semantics of the program
596 \see{sec:correctness}.
597 The \name{NetBeans} IDE implements this refactoring in a somewhat
598 unsophisticated way. For starters, the refactoring's default destination for the
599 move, is the same class as the method already resides in, although it refuses to
600 perform the refactoring if chosen. But the worst part is, that if moving the
601 method \method{f} of the class \type{C} to the class \type{X}, it will break the
602 code. The result is shown in \myref{lst:moveMethod_NetBeans}.
606 \begin{minted}[samepage]{java}
619 \begin{minted}[samepage]{java}
629 \caption{Moving method \method{f} from \type{C} to \type{X}.}
630 \label{lst:moveMethod_NetBeans}
633 \name{NetBeans} will try to create code that call the methods \method{m} and \method{n}
634 of \type{X} by accessing them through \var{c.x}, where \var{c} is a parameter of
635 type \type{C} that is added the method \method{f} when it is moved. (This is
636 seldom the desired outcome of this refactoring, but ironically, this ``feature''
637 keeps \name{NetBeans} from breaking the code in the example from
638 \myref{sec:correctness}.) If \var{c.x} for some reason is inaccessible to
639 \type{X}, as in this case, the refactoring breaks the code, and it will not
640 compile. \name{NetBeans} presents a preview of the refactoring outcome, but the
641 preview does not catch it if the IDE is about break the program.
643 The IDEs under investigation seem to have fairly good support for primitive
644 refactorings, but what about more complex ones, such as
645 \gloss{extractClass}\citing{refactoring}? \name{IntelliJ} handles this in a
646 fairly good manner, although, in the case of private methods, it leaves unused
647 methods behind. These are methods that delegate to a field with the type of the
648 new class, but are not used anywhere. \name{Eclipse} has added its own quirk to
649 the \refa{Extract Class} refactoring, and only allows for \emph{fields} to be
650 moved to a new class, \emph{not methods}. This makes it effectively only
651 extracting a data structure, and calling it \refa{Extract Class} is a little
652 misleading. One would often be better off with textual extract and paste than
653 using the \refa{Extract Class} refactoring in \name{Eclipse}. When it comes to
654 \name{NetBeans}, it does not even show an attempt on providing this refactoring.
656 \subsection{The relation to design patterns}\label{relationToDesignPatterns}
658 Refactoring and design patterns have at least one thing in common, they are both
659 promoted by advocates of \emph{clean code}\citing{cleanCode} as fundamental
660 tools on the road to more maintainable and extendable source code.
663 Design patterns help you determine how to reorganize a design, and they can
664 reduce the amount of refactoring you need to do
665 later.~\cite[p.~353]{designPatterns}
668 Although sometimes associated with
669 over-engineering\citing{kerievsky2005,refactoring}, design patterns are in
670 general assumed to be good for maintainability of source code. That may be
671 because many of them are designed to support the \emph{open/closed principle} of
672 object-oriented programming. The principle was first formulated by Bertrand
673 Meyer, the creator of the Eiffel programming language, like this: ``Modules
674 should be both open and closed.''\citing{meyer1988} It has been popularized,
675 with this as a common version:
678 Software entities (classes, modules, functions, etc.) should be open for
679 extension, but closed for modification.\footnote{See
680 \url{http://c2.com/cgi/wiki?OpenClosedPrinciple} or
681 \url{https://en.wikipedia.org/wiki/Open/closed_principle}}
684 Maintainability is often thought of as the ability to be able to introduce new
685 functionality without having to change too much of the old code. When
686 refactoring, the motivation is often to facilitate adding new functionality. It
687 is about factoring the old code in a way that makes the new functionality being
688 able to benefit from the functionality already residing in a software system,
689 without having to copy old code into new. Then, next time someone shall add new
690 functionality, it is less likely that the old code has to change. Assuming that
691 a design pattern is the best way to get rid of duplication and assist in
692 implementing new functionality, it is reasonable to conclude that a design
693 pattern often is the target of a series of refactorings. Having a repertoire of
694 design patterns can also help in knowing when and how to refactor a program to
695 make it reflect certain desired characteristics.
698 There is a natural relation between patterns and refactorings. Patterns are
699 where you want to be; refactorings are ways to get there from somewhere
700 else.~\cite[p.~107]{refactoring}
703 This quote is wise in many contexts, but it is not always appropriate to say
704 ``Patterns are where you want to be\ldots''. \emph{Sometimes}, patterns are
705 where you want to be, but only because it will benefit your design. It is not
706 true that one should always try to incorporate as many design patterns as
707 possible into a program. It is not like they have intrinsic value. They only add
708 value to a system when they support its design. Otherwise, the use of design
709 patterns may only lead to a program that is more complex than necessary.
712 The overuse of patterns tends to result from being patterns happy. We are
713 \emph{patterns happy} when we become so enamored of patterns that we simply
714 must use them in our code.~\cite[p.~24]{kerievsky2005}
717 This can easily happen when relying largely on up-front design. Then it is
718 natural, in the very beginning, to try to build in all the flexibility that one
719 believes will be necessary throughout the lifetime of a software system.
720 According to Joshua Kerievsky ``That sounds reasonable --- if you happen to be
721 psychic.''~\cite[p.~1]{kerievsky2005} He is advocating what he believes is a
722 better approach: To let software continually evolve. To start with a simple
723 design that meets today's needs, and tackle future needs by refactoring to
724 satisfy them. He believes that this is a more economic approach than investing
725 time and money into a design that inevitably is going to change. By relying on
726 continuously refactoring a system, its design can be made simpler without
727 sacrificing flexibility. To be able to fully rely on this approach, it is of
728 utter importance to have a reliable suit of tests to lean on \see{testing}. This
729 makes the design process more natural and less characterized by difficult
730 decisions that has to be made before proceeding in the process, and that is
731 going to define a project for all of its unforeseeable future.
733 \subsection{The impact on software quality}
735 \subsubsection{What is software quality?}
736 The term \emph{software quality} has many meanings. It all depends on the
737 context we put it in. If we look at it with the eyes of a software developer, it
738 usually means that the software is easily maintainable and testable, or in other
739 words, that it is \emph{well designed}. This often correlates with the
740 management scale, where \emph{keeping the schedule} and \emph{customer
741 satisfaction} is at the center. From the customers point of view, in addition to
742 good usability, \emph{performance} and \emph{lack of bugs} is always
743 appreciated, measurements that are also shared by the software developer. (In
744 addition, such things as good documentation could be measured, but this is out
745 of the scope of this document.)
747 \subsubsection{The impact on performance}
749 Refactoring certainly will make software go more slowly\footnote{With today's
750 compiler optimization techniques and performance tuning of e.g. the Java
751 virtual machine, the penalties of object creation and method calls are
752 debatable.}, but it also makes the software more amenable to performance
753 tuning.~\cite[p.~69]{refactoring}
756 \noindent There is a common belief that refactoring compromises performance, due
757 to increased degree of indirection and that polymorphism is slower than
760 In a survey, Demeyer\citing{demeyer2002} disproves this view in the case of
761 polymorphism. He did an experiment on, what he calls, ``Transform Self Type
762 Checks'' where you introduce a new polymorphic method and a new class hierarchy
763 to get rid of a class' type checking of a ``type attribute``. He uses this kind
764 of transformation to represent other ways of replacing conditionals with
765 polymorphism as well. The experiment is performed on the C++ programming
766 language and with three different compilers and platforms. Demeyer concludes
767 that, with compiler optimization turned on, polymorphism beats middle to large
768 sized if-statements and does as well as case-statements. (In accordance with
769 his hypothesis, due to similarities between the way C++ handles polymorphism and
773 The interesting thing about performance is that if you analyze most programs,
774 you find that they waste most of their time in a small fraction of the
775 code.~\cite[p.~70]{refactoring}
778 \noindent So, although an increased amount of method calls could potentially
779 slow down programs, one should avoid premature optimization and sacrificing good
780 design, leaving the performance tuning until after \gloss{profiling} the
781 software and having isolated the actual problem areas.
783 \subsection{Composite refactorings}\label{compositeRefactorings}
784 Generally, when thinking about refactoring, at the mechanical level, there are
785 essentially two kinds of refactorings. There are the \emph{primitive}
786 refactorings, and the \emph{composite} refactorings.
788 \definition{A \emph{primitive refactoring} is a refactoring that cannot be
789 expressed in terms of other refactorings.}
791 \noindent Examples are the \refa{Pull Up Field} and \refa{Pull Up
792 Method} refactorings\citing{refactoring}, that move members up in their class
795 \definition{A \emph{composite refactoring} is a refactoring that can be
796 expressed in terms of two or more other refactorings.}
798 \noindent An example of a composite refactoring is the \refa{Extract
799 Superclass} refactoring\citing{refactoring}. In its simplest form, it is composed
800 of the previously described primitive refactorings, in addition to the
801 \refa{Pull Up Constructor Body} refactoring\citing{refactoring}. It works
802 by creating an abstract superclass that the target class(es) inherits from, then
803 by applying \refa{Pull Up Field}, \refa{Pull Up Method} and
804 \refa{Pull Up Constructor Body} on the members that are to be members of
805 the new superclass. If there are multiple classes in play, their interfaces may
806 need to be united with the help of some rename refactorings, before extracting
807 the superclass. For an overview of the \refa{Extract Superclass}
808 refactoring, see \myref{fig:extractSuperclass}.
812 \includegraphics[angle=270,width=\linewidth]{extractSuperclassItalic.pdf}
813 \caption{The Extract Superclass refactoring, with united interfaces.}
814 \label{fig:extractSuperclass}
817 \subsection{Manual vs. automated refactorings}
818 Refactoring is something every programmer does, even if \heshe does not known
819 the term \emph{refactoring}. Every refinement of source code that does not alter
820 the program's behavior is a refactoring. For small refactorings, such as
821 \ExtractMethod, executing it manually is a manageable task, but is still prone
822 to errors. Getting it right the first time is not easy, considering the method
823 signature and all the other aspects of the refactoring that has to be in place.
825 Consider the renaming of classes, methods and fields. For complex programs these
826 refactorings are almost impossible to get right. Attacking them with textual
827 search and replace, or even regular expressions, will fall short on these tasks.
828 Then it is crucial to have proper tool support that can perform them
829 automatically. Tools that can parse source code and thus have semantic knowledge
830 about which occurrences of which names belong to what construct in the program.
831 For even trying to perform one of these complex tasks manually, one would have
832 to be very confident on the existing test suite \see{testing}.
834 \subsection{Correctness of refactorings}\label{sec:correctness}
835 For automated refactorings to be truly useful, they must show a high degree of
836 behavior preservation. This last sentence might seem obvious, but there are
837 examples of refactorings in existing tools that break programs. In an ideal
838 world, every automated refactoring would be ``complete'', in the sense that it
839 would never break a program. In an ideal world, every program would also be free
840 from bugs. In modern IDEs the implemented automated refactorings are working for
841 \emph{most} cases, which is enough for making them useful.
843 I will now present an example of a \emph{corner case} where a program breaks
844 when a refactoring is applied. The example shows an \ExtractMethod refactoring
845 followed by a \MoveMethod refactoring that breaks a program in both the
846 \name{Eclipse} and \name{IntelliJ} IDEs\footnote{The \name{NetBeans} IDE handles this
847 particular situation without altering the program's behavior, mainly because
848 its \refa{Move Method} refactoring implementation is a bit flawed in other ways
849 \see{toolSupport}.}. The target and the destination for the composed
850 refactoring are shown in \myref{lst:correctnessExtractAndMove}. Note that the
851 method \method{m(C c)} of class \type{X} assigns to the field \var{x} of the
852 argument \var{c} that has type \type{C}.
856 \begin{minted}[linenos,frame=topline,label={Refactoring
857 target},framesep=\mintedframesep]{java}
859 public X x = new X();
871 \begin{minted}[frame=topline,label={Method
872 destination},framesep=\mintedframesep]{java}
876 // If m is called from
877 // c, then c.x no longer
884 \caption{The target and the destination for the composition of the Extract
885 Method and \refa{Move Method} refactorings.}
886 \label{lst:correctnessExtractAndMove}
890 The refactoring sequence works by extracting line 6 through 8 from the original
891 class \type{C} into a method \method{f} with the statements from those lines as
892 its method body (but with the comment left out, since it will no longer hold any
893 meaning). The method is then moved to the class \type{X}. The result is shown
894 in \myref{lst:correctnessExtractAndMoveResult}.
896 Before the refactoring, the methods \method{m} and \method{n} of class \type{X}
897 are called on different object instances (see line 6 and 8 of the original class
898 \type{C} in \cref{lst:correctnessExtractAndMove}). After the refactoring, they
899 are called on the same object, and the statement on line
900 3 of class \type{X} (in \cref{lst:correctnessExtractAndMoveResult}) no longer
901 has the desired effect in our example. The method \method{f} of class \type{C}
902 is now calling the method \method{f} of class \type{X} (see line 5 of class
903 \type{C} in \cref{lst:correctnessExtractAndMoveResult}), and the program now
904 behaves different than before.
908 \begin{minted}[linenos]{java}
910 public X x = new X();
920 \begin{minted}[linenos]{java}
935 \caption{The result of the composed refactoring.}
936 \label{lst:correctnessExtractAndMoveResult}
939 The bug introduced in the previous example is of such a nature\footnote{Caused
940 by aliasing. See \url{https://en.wikipedia.org/wiki/Aliasing_(computing)}}
941 that it is very difficult to spot if the refactored code is not covered by
942 tests. It does not generate compilation errors, and will thus only result in
943 a runtime error or corrupted data, which might be hard to detect.
945 \subsection{Refactoring and the importance of testing}\label{testing}
947 If you want to refactor, the essential precondition is having solid
948 tests.\citing{refactoring}
951 When refactoring, there are roughly three classes of errors that can be made.
952 The first class of errors is the one that makes the code unable to compile.
953 These \emph{compile-time} errors are of the nicer kind. They flash up at the
954 moment they are made (at least when using an IDE), and are usually easy to fix.
955 The second class is the \emph{runtime} errors. Although these errors take a bit
956 longer to surface, they usually manifest after some time in an illegal argument
957 exception, null pointer exception or similar during the program execution.
958 These kinds of errors are a bit harder to handle, but at least they will show,
959 eventually. Then there are the \emph{behavior-changing} errors. These errors are
960 of the worst kind. They do not show up during compilation and they do not turn
961 on a blinking red light during runtime either. The program can seem to work
962 perfectly fine with them in play, but the business logic can be damaged in ways
963 that will only show up over time.
965 For discovering runtime errors and behavior changes when refactoring, it is
966 essential to have good test coverage. Testing in this context means writing
967 automated tests. Manual testing may have its uses, but when refactoring, it is
968 automated unit testing that dominate. For discovering behavior changes it is
969 especially important to have tests that cover potential problems, since these
970 kinds of errors do not reveal themselves.
972 Unit testing is not a way to \emph{prove} that a program is correct, but it is a
973 way to make you confident that it \emph{probably} works as desired. In the
974 context of test-driven development (commonly known as TDD), the tests are even a
975 way to define how the program is \emph{supposed} to work. It is then, by
976 definition, working if the tests are passing.
978 If the test coverage for a code base is perfect, then it should, theoretically,
979 be risk-free to perform refactorings on it. This is why automated tests and
980 refactoring is such a great match.
982 \subsubsection{Testing the code from correctness section}
983 The worst thing that can happen when refactoring is to introduce changes to the
984 behavior of a program, as in the example on \myref{sec:correctness}. This
985 example may be trivial, but the essence is clear. The only problem with the
986 example is that it is not clear how to create automated tests for it, without
987 changing it in intrusive ways.
989 Unit tests, as they are known from the different \glosspl{xUnit} around, are
990 only suitable to test the \emph{result} of isolated operations. They can not
991 easily (if at all) observe the \emph{history} of a program.
993 This problem is still open.
997 Assuming a sequential (non-concurrent) program:
1000 tracematch (C c, X x) {
1002 call(* X.m(C)) && args(c) && cflow(within(C));
1004 call(* X.n()) && target(x) && cflow(within(C));
1006 set(C.x) && target(c) && !cflow(m);
1010 { assert x == c.x; }
1014 %\begin{minted}{java}
1015 %tracematch (X x1, X x2) {
1017 % call(* X.m(C)) && target(x1);
1019 % call(* X.n()) && target(x2);
1021 % set(C.x) && !cflow(m) && !cflow(n);
1025 % { assert x1 != x2; }
1031 \section{The Project}\label{sec:project}
1032 In this section we look at the work that shall be done for this project, its
1033 building stones and some of the methodologies used.
1035 \subsection{Project description}
1036 The aim of this master's project will be to explore the relationship between the
1037 \ExtractMethod and the \MoveMethod refactorings. This will be done by composing
1038 the two into a composite refactoring. The refactoring will be called the
1039 \ExtractAndMoveMethod refactoring.
1041 The two primitive \ExtractMethod and \MoveMethod refactorings must already be
1042 implemented in a tool, so the \ExtractAndMoveMethod refactoring is going to be
1043 built on top of those.
1045 The composition of the \ExtractMethod and \MoveMethod refactorings springs
1046 naturally out of the need to move procedures closer to the data they manipulate.
1047 This composed refactoring is not well described in the literature, but it is
1048 implemented in at least one tool called
1049 \name{CodeRush}\footnote{\url{https://help.devexpress.com/\#CodeRush/CustomDocument3519}},
1050 which is an extension for \name{MS Visual
1051 Studio}\footnote{\url{http://www.visualstudio.com/}}. In CodeRush it is called
1052 \refa{Extract Method to
1053 Type}\footnote{\url{https://help.devexpress.com/\#CodeRush/CustomDocument6710}},
1054 but I choose to call it \ExtractAndMoveMethod, since I feel this better
1055 communicates which primitive refactorings it is composed of.
1057 The project will consist of implementing the \ExtractAndMoveMethod refactoring,
1058 as well as executing it over a larger code base, as a case study. To be able to
1059 execute the refactoring automatically, I have to make it analyze code to
1060 determine the best selections to extract into new methods.
1062 \subsection{The premises}
1063 Before we can start manipulating source code and write a tool for doing so, we
1064 need to decide on a programming language for the code we are going to
1065 manipulate. Also, since we do not want to start from scratch by implementing
1066 primitive refactorings ourselves, we need to choose an existing tool that
1067 provides the needed refactorings. In addition to be able to perform changes, we
1068 need a framework for analyzing source code for the language we select.
1070 \subsubsection{Choosing the target language}
1071 Choosing which programming language the code that shall be manipulated shall be
1072 written in, is not a very difficult task. We choose to limit the possible
1073 languages to the object-oriented programming languages, since most of the
1074 terminology and literature regarding refactoring comes from the world of
1075 object-oriented programming. In addition, the language must have existing tool
1076 support for refactoring.
1078 The \name{Java} programming language\footnote{\url{https://www.java.com/}} is
1079 the dominating language when it comes to example code in the literature of
1080 refactoring, and is thus a natural choice. Java is perhaps, currently the most
1081 influential programming language in the world, with its \name{Java Virtual
1082 Machine} that runs on all of the most popular architectures and also supports
1083 dozens of other programming languages\footnote{They compile to Java bytecode.},
1084 with \name{Scala}, \name{Clojure} and \name{Groovy} as the most prominent ones.
1085 Java is currently the language that every other programming language is compared
1086 against. It is also the primary programming language for the author of this
1089 \subsubsection{Choosing the tools}
1090 When choosing a tool for manipulating Java, there are certain criteria that
1091 have to be met. First of all, the tool should have some existing refactoring
1092 support that this thesis can build upon. Secondly it should provide some kind of
1093 framework for parsing and analyzing Java source code. Third, it should itself be
1094 open source. This is both because of the need to be able to browse the code for
1095 the existing refactorings that is contained in the tool, and also because open
1096 source projects hold value in them selves. Another important aspect to consider
1097 is that open source projects of a certain size, usually has large communities of
1098 people connected to them, that are committed to answering questions regarding the
1099 use and misuse of the products, that to a large degree is made by the community
1102 There is a certain class of tools that meet these criteria, namely the class of
1103 \emph{IDEs}\footnote{\emph{Integrated Development Environment}}. These are
1104 programs that are meant to support the whole production cycle of a computer
1105 program, and the most popular IDEs that support Java, generally have quite good
1106 refactoring support.
1108 The main contenders for this thesis is the \name{Eclipse IDE}, with the
1109 \name{Java development tools} (JDT), the \name{IntelliJ IDEA Community Edition}
1110 and the \name{NetBeans IDE} \see{toolSupport}. \name{Eclipse} and
1111 \name{NetBeans} are both free, open source and community driven, while the
1112 \name{IntelliJ IDEA} has an open sourced community edition that is free of
1113 charge, but also offer an \name{Ultimate Edition} with an extended set of
1114 features, at additional cost. All three IDEs supports adding plugins to extend
1115 their functionality and tools that can be used to parse and analyze Java source
1116 code. But one of the IDEs stand out as a favorite, and that is the \name{Eclipse
1117 IDE}. This is the most popular\citing{javaReport2011} among them and seems to be
1118 de facto standard IDE for Java development regardless of platform.
1121 \subsection{The primitive refactorings}
1122 The refactorings presented here are the primitive refactorings used in this
1123 project. They are the abstract building blocks used by the \ExtractAndMoveMethod
1126 \paragraph{The Extract Method refactoring}
1127 The \refa{Extract Method} refactoring is used to extract a fragment of code
1128 from its context and into a new method. A call to the new method is inlined
1129 where the fragment was before. It is used to break code into logical units, with
1130 names that explain their purpose.
1132 An example of an \ExtractMethod refactoring is shown in
1133 \myref{lst:extractMethodRefactoring}. It shows a method containing calls to the
1134 methods \method{foo} and \method{bar} of a type \type{X}. These statements are
1135 then extracted into the new method \method{fooBar}.
1138 \begin{multicols}{2}
1139 \begin{minted}[samepage,frame=topline,label={Before},framesep=\mintedframesep]{java}
1150 \begin{minted}[samepage,frame=topline,label={After},framesep=\mintedframesep]{java}
1162 \caption{An example of an \ExtractMethod refactoring.}
1163 \label{lst:extractMethodRefactoring}
1166 \paragraph{The Move Method refactoring}
1167 The \refa{Move Method} refactoring is used to move a method from one class to
1168 another. This can be appropriate if the method is using more features of another
1169 class than of the class which it is currently defined.
1171 \Myref{lst:moveMethodRefactoring} shows an example of this refactoring. Here a
1172 method \method{fooBar} is moved from the class \type{C} to the class \type{X}.
1175 \begin{multicols}{2}
1176 \begin{minted}[samepage,frame=topline,label={Before},framesep=\mintedframesep]{java}
1195 \begin{minted}[samepage,frame=topline,label={After},framesep=\mintedframesep]{java}
1212 \caption{An example of a \MoveMethod refactoring.}
1213 \label{lst:moveMethodRefactoring}
1216 \subsection{The Extract and Move Method refactoring}
1217 The \ExtractAndMoveMethod refactoring is a composite refactoring composed of the
1218 primitive \ExtractMethod and \MoveMethod refactorings. The effect of this
1219 refactoring on source code is the same as when extracting a method and moving it
1220 to another class. Conceptually, this is done without an intermediate step. In
1221 practice, as we shall see later, an intermediate step may be necessary.
1223 An example of this composite refactoring is shown in
1224 \myref{lst:extractAndMoveMethodRefactoring}. The example joins the examples from
1225 \cref{lst:extractMethodRefactoring} and \cref{lst:moveMethodRefactoring}. This
1226 means that the selection consisting of the consecutive calls to the methods
1227 \method{foo} and \method{bar}, is extracted into a new method \method{fooBar}
1228 located in the class \type{X}.
1231 \begin{multicols}{2}
1232 \begin{minted}[samepage,frame=topline,label={Before},framesep=\mintedframesep]{java}
1248 \begin{minted}[samepage,frame=topline,label={After},framesep=\mintedframesep]{java}
1265 \caption{An example of the \ExtractAndMoveMethod refactoring.}
1266 \label{lst:extractAndMoveMethodRefactoring}
1269 \subsection{The Coupling Between Object Classes metric}\label{sec:CBO}
1270 The best known metric for measuring coupling between classes in object-oriented
1271 software is called \metr{Coupling Between Object Classes}, usually abbreviated
1272 as CBO. The metric is defined in the article \tit{A Metrics Suite for Object
1273 Oriented Design}\citing{metricsSuite1994} by Chidamber and Kemerer, published in
1276 \definition{\emph{CBO} for a class is a count of the number of other classes to
1277 which it is coupled.}
1279 An object is coupled to another object if one of them acts on the other by using
1280 methods or instance variables of the other object. This relation goes both ways,
1281 so both outgoing and incoming uses are counted. Each coupling relationship is
1282 only considered once when measuring CBO for a class.
1284 \paragraph{How can the Extract and Move Method refactoring improve CBO?}
1285 \Myref{lst:CBOExample} shows how CBO changes for a class when it is refactored
1286 with the \ExtractAndMoveMethod refactoring. In the example we consider only the
1287 CBO value of class \type{C}.
1290 \begin{multicols}{2}
1291 \begin{minted}[linenos,samepage,frame=topline,label={Before},framesep=\mintedframesep]{java}
1322 \begin{minted}[linenos,samepage,frame=topline,label={After},framesep=\mintedframesep]{java}
1354 \caption{An example of improving CBO. Class \type{C} has a CBO value of 4
1355 before refactoring it, and 3 after.}
1356 \label{lst:CBOExample}
1359 Before refactoring the class \type{C} with the \ExtractAndMoveMethod
1360 refactoring, it has a CBO value of 4. The class uses members of the classes
1361 \type{A} and \type{B}, which accounts for 2 of the coupling relationships of
1362 class \type{C}. In addition to this, it uses its variable \var{x} with type
1363 \type{X} and also the methods \method{foo} and \method{bar} declared in class
1364 \type{Y}, giving it a total CBO value of 4.
1366 The after-part of the example code in \mysimpleref{lst:CBOExample} shows the
1367 result of extracting the lines
1368 5 and 6 of class \type{C} into a new method \method{fooBar}, with a subsequent
1369 move of it to class \type{X}.
1371 With respect to the CBO metric, the refactoring action accomplishes something
1372 important: It eliminates the uses of class \type{Y} from class \type{C}. This
1373 means that the class \type{C} is no longer coupled to \type{Y}, only the classes
1374 \type{A}, \type{B} and \type{X}. The CBO value of class \type{C} is therefore 3
1375 after refactoring, while no other class have received any increase in CBO.
1377 The example shown here is an ideal situation. Coupling is reduced for one class
1378 without any increase of coupling for another class. There is also another point
1379 that is important. It is the fact that to reduce the CBO value for a class, we
1380 need to remove \emph{all} its uses of another class. This is done for the class
1381 \type{C} in \myref{lst:CBOExample}, where all uses of class \type{Y} is removed
1382 by the \ExtractAndMoveMethod refactoring.
1383 \todoin{Highlight code}
1386 \subsection{Research questions}\label{sec:researchQuestions}
1387 The main question that I seek an answer to in this thesis is:
1390 Is it possible to automate the analysis and execution of the
1391 \ExtractAndMoveMethod refactoring, and do so for all of the code of a larger
1395 \noindent The secondary questions will then be:
1397 \paragraph{Can we do this efficiently?} Can we automate the analysis and
1398 execution of the refactoring so it can be run in a reasonable amount of time?
1400 \paragraph{Can we perform changes safely?} Can we take actions to prevent the
1401 refactoring from breaking the code? By breaking the code we mean to either do
1402 changes that do not compile, or make changes that alter the behavior of the
1405 \paragraph{Can we improve the quality of source code?} Assuming that the
1406 refactoring is safe: Is it feasible to assure that the code we refactor actually
1407 gets better in terms of coupling?
1409 \paragraph{How can the automation of the refactoring be helpful?} Assuming the
1410 refactoring does in fact improve the quality of source code and is safe to use:
1411 What is the usefulness of the refactoring in a software development setting? In
1412 what parts of the development process can the refactoring play a role?
1414 \subsection{Methodology}
1415 This section will present some of the methods used during the work of this
1418 \subsubsection{Evolutionary design}
1419 In the programming work for this project, I have tried using a design strategy
1420 called evolutionary design, also known as continuous or incremental
1421 design\citing{wiki_continuous_2014}. It is a software design strategy advocated
1422 by the Extreme Programming community. The essence of the strategy is that you
1423 should let the design of your program evolve naturally as your requirements
1424 change. This is seen in contrast with up-front design, where design decisions
1425 are made early in the process.
1427 The motivation behind evolutionary design is to keep the design of software as
1428 simple as possible. This means not introducing unneeded functionality into a
1429 program. You should defer introducing flexibility into your software, until it
1430 is needed to be able to add functionality in a clean way.
1432 Holding up design decisions, implies that the time will eventually come when
1433 decisions have to be made. The flexibility of the design then relies on the
1434 programmer's abilities to perform the necessary refactoring, and \his confidence
1435 in those abilities. From my experience working on this project, I can say that
1436 this confidence is greatly enhanced by having automated tests to rely on
1439 The choice of going for evolutionary design developed naturally. As Fowler
1440 points out in his article \tit{Is Design Dead?}, evolutionary design much
1441 resembles the ``code and fix'' development strategy\citing{fowler_design_2004}.
1442 A strategy that most of us have practiced in school. This was also the case when
1443 I first started this work. I had to learn the inner workings of Eclipse and its
1444 refactoring-related plugins. That meant a lot of fumbling around with code I did
1445 not know, in a trial and error fashion. Eventually I started writing tests for
1446 my code, and my design began to evolve.
1448 \subsubsection{Test-driven development}\label{tdd}
1449 As mentioned before, the project started out as a classic code and fix
1450 development process. My focus was aimed at getting something to work, rather
1451 than doing so according to best practice. This resulted in a project that got
1452 out of its starting blocks, but it was not accompanied by any tests. Hence it
1453 was soon difficult to make any code changes with the confidence that the program
1454 was still correct afterwards (assuming it was so before changing it). I always
1455 knew that I had to introduce some tests at one point, but this experience
1456 accelerated the process of leading me onto the path of testing.
1458 I then wrote tests for the core functionality of the plugin, and thus gained
1459 more confidence in the correctness of my code. I could now perform quite drastic
1460 changes without ``wetting my pants``. After this, nearly all of the semantic
1461 changes done to the business logic of the project, or the addition of new
1462 functionality, were made in a test-driven manner. This means that before
1463 performing any changes, I would define the desired functionality through a set
1464 of tests. I would then run the tests to check that they were run and that they
1465 did not pass. Then I would do any code changes necessary to make the tests
1466 pass. The definition of how the program is supposed to operate is then captured
1467 by the tests. However, this does not prove the correctness of the analysis
1468 leading to the test definitions.
1470 \subsection{Case study}
1473 \subsection{Dogfooding}
1476 \section{Related Work}\label{sec:relatedWork}
1478 \subsection{Refactoring safety}
1479 This section presents a couple of approaches to improving the safety of
1480 performing refactorings. In these approaches, the problems that are addressed
1481 are not compilation problems, but behavior-altering problems that are not easily
1482 discovered during static analysis of source code. An example of such a problem
1483 is presented in \myref{sec:correctness}.
1485 \subsubsection{Project ``Safer Refactorings''}
1486 \tit{Safer Refactorings}\citing{stolzSaferRefactorings} is a proposal for a
1487 master's thesis. The proposer is my supervisor, Volker Stolz from the University
1490 The proposed solution for making refactorings safer, is to insert assertions
1491 into source code when refactoring it. For the example in
1492 \myref{lst:correctnessExtractAndMoveResult}, which is the result of a
1493 refactoring, it is suggested that we insert an assert statement between lines 9
1494 and 10. In this example, the assert statement
1495 would be \mint{java}|assert c.x == this;| which would discover the aliasing
1496 problems of this example.
1498 \subsubsection{``Making Program Refactoring Safer''}
1499 This is the name of an article\citing{soaresSafer2010} about providing a way to
1500 improve safety during refactoring. Soares et al. approaches the problem of
1501 preserving behavior during refactoring by analyzing a transformation and then
1502 generate a test suite for it, using static analysis. These tests are then run
1503 for both the before- and after-code, and is compared to assure that they are
1506 \subsection{Search-based refactoring}
1507 \tit{Search-Based Refactoring: an
1508 empirical study}\citing{okeeffeSearchBased2008} is a paper by Mark O'Keeffe and
1509 Mel Ó Cinnéide published in 2008. The authors present an empirical study of
1510 different algorithmic approaches to search-based refactoring.
1512 The common approach for all these algorithms is to generate a set of changes to
1513 a program for then to use a ``fitness function'' to evaluate if they improve its
1514 design or not. The fitness function consists of a weighted sum of different
1515 object-oriented metrics.
1517 Among other things, the authors conclude that even with small input programs,
1518 their solution representation is memory-intensive, at least for some algorithms.
1519 The programs they refactor on have in average 4,000 lines of code, spread over
1520 57 classes. I.e. considerably smaller than one of the programs that will be
1521 subject to refactoring in this project.
1524 \subsection{The compositional paradigm of refactoring}
1525 This paradigm builds upon the observation of Vakilian et
1526 al.\citing{vakilian2012}, that of the many automated refactorings existing in
1527 modern IDEs, the simplest ones are dominating the usage statistics. The report
1528 mainly focuses on \name{Eclipse} as the tool under investigation.
1530 The paradigm is described almost as the opposite of automated composition of
1531 refactorings \see{compositeRefactorings}. It works by providing the programmer
1532 with easily accessible primitive refactorings. These refactorings shall be
1533 accessed via keyboard shortcuts or quick-assist menus\footnote{Think
1534 quick-assist with Ctrl+1 in \name{Eclipse}} and be promptly executed, opposed to in the
1535 currently dominating wizard-based refactoring paradigm. They are meant to
1536 stimulate composing smaller refactorings into more complex changes, rather than
1537 doing a large upfront configuration of a wizard-based refactoring, before
1538 previewing and executing it. The compositional paradigm of refactoring is
1539 supposed to give control back to the programmer, by supporting \himher with an
1540 option of performing small rapid changes instead of large changes with a lesser
1541 degree of control. The report authors hope this will lead to fewer unsuccessful
1542 refactorings. It also could lower the bar for understanding the steps of a
1543 larger composite refactoring and thus also help in figuring out what goes wrong
1544 if one should choose to op in on a wizard-based refactoring.
1546 Vakilian and his associates have performed a survey of the effectiveness of the
1547 compositional paradigm versus the wizard-based one. They claim to have found
1548 evidence of that the \emph{compositional paradigm} outperforms the
1549 \emph{wizard-based}. It does so by reducing automation, which seems
1550 counterintuitive. Therefore they ask the question ``What is an appropriate level
1551 of automation?'', and thus questions what they feel is a rush toward more
1552 automation in the software engineering community.
1556 \chapter{The search-based Extract and Move Method
1557 refactoring}\label{ch:extractAndMoveMethod}
1558 In this chapter I will delve into the workings of the search-based
1559 \ExtractAndMoveMethod refactoring. We will see the choices it must make along
1560 the way and why it chooses a text selection as a candidate for refactoring or
1563 After defining some concepts, I will introduce an example that will be used
1564 throughout the chapter to illustrate how the refactoring works in some simple
1567 \section{The inputs to the refactoring}
1568 For executing an \ExtractAndMoveMethod refactoring, there are two simple
1569 requirements. The first thing the refactoring needs is a text selection, telling
1570 it what to extract. Its second requirement is a target for the subsequent move
1573 The extracted method must be called instead of the selection that makes up its
1574 body. Also, the method call has to be performed via a variable, since the method
1575 is not static. Therefore, the move target must be a variable in the scope of the
1576 extracted selection. The actual new location for the extracted method will be
1577 the class representing the type of the move target variable. But, since the
1578 method also must be called through a variable, it makes sense to define the move
1579 target to be either a local variable or a field in the scope of the text
1582 \section{Defining a text selection}
1583 A text selection, in our context, is very similar to what you think of when
1584 selecting a bit of text in your editor or other text processing tool with your
1585 mouse or keyboard. It is an abstract construct that is meant to capture which
1586 specific portion of text we are about to deal with.
1588 To be able to clearly reason about a text selection done to a portion of text in
1589 a computer file, which consists of pure text, we put up the following
1592 \definition{A \emph{text selection} in a text file is defined by two
1593 non-negative integers, in addition to a reference to the file itself. The first
1594 integer is an offset into the file, while the second reference is the length of
1595 the text selection.}
1597 This means that the selected text consist of a number of characters equal to the
1598 length of the selection, where the first character is found at the specified
1601 \section{Where we look for text selections}
1603 \subsection{Text selections are found in methods}
1604 The text selections we are interested in are those that surround program
1605 statements. Therefore, the place we look for selections that can form candidates
1606 for an execution of the \ExtractAndMoveMethod refactoring, is within the body of
1609 \paragraph{On ignoring static methods}
1610 In this project we are not analyzing static methods for candidates to the
1611 \ExtractAndMoveMethod refactoring. The reason for this is that in the cases
1612 where we want to perform the refactoring for a selection within a static method,
1613 the first step is to extract the selection into a new method. Hence this method
1614 also becomes static, since it must be possible to call it from a static context.
1615 It would then be difficult to move the method to another class, make it
1616 non-static and calling it through a variable. To avoid these obstacles, we
1617 simply ignore static methods.
1619 \begin{listing}[htb]
1620 \def\charwidth{5.8pt}
1621 \def\indent{2*\charwidth}
1622 \def\lineheight{\baselineskip}
1623 \def\mintedtop{2*\lineheight+5.8pt}
1625 \begin{tikzpicture}[overlay, yscale=-1, xshift=3.8pt+\charwidth*31]
1626 \tikzstyle{overlaybox}=[fill=lightgray,opacity=0.2]
1628 \draw[overlaybox] (\indent,\mintedtop+\lineheight*4) rectangle
1629 +(23*\charwidth,17*\lineheight);
1632 \draw[overlaybox] (2*\indent,\mintedtop+5*\lineheight) rectangle
1633 +(15*\charwidth,3*\lineheight);
1634 \draw[overlaybox] (2*\indent,\mintedtop+15*\lineheight) rectangle
1635 +(15*\charwidth,3*\lineheight);
1636 \draw[overlaybox] (2*\indent,\mintedtop+19*\lineheight) rectangle
1637 +(15*\charwidth,\lineheight);
1639 \begin{multicols}{2}
1640 \begin{minted}[linenos,frame=topline,label=Clean,framesep=\mintedframesep]{java}
1642 A a; B b; boolean bool;
1644 void method(int val) {
1668 \begin{minted}[frame=topline,label={With statement
1669 sequences},framesep=\mintedframesep]{java}
1671 A a; B b; boolean bool;
1673 void method(int val) {
1696 \caption{Classes \type{A} and \type{B} are both public. The methods
1697 \method{foo} and \method{bar} are public members of class \type{A}.}
1698 \label{lst:grandExample}
1701 \subsection{The possible text selections of a method body}
1702 The number of possible text selections that can be made from the text in a
1703 method body, are equal to all the sub-sequences of characters within it. For our
1704 purposes, analyzing program source code, we must define what it means for a text
1705 selection to be valid.
1707 \definition{A \emph{valid text selection} is a text selection that contains all
1708 of one or more consecutive program statements.}
1710 For a sequence of statements, the text selections that can be made from it, are
1711 equal to all its sub-sequences. \Myref{lst:textSelectionsExample} show an
1712 example of all the text selections that can be made from the code in
1713 \myref{lst:grandExample}, lines 16-18. For convenience and the clarity of this
1714 example, the text selections are represented as tuples with the start and end
1715 line of all selections: $\{(16), (17), (18), (16,17), (16,18), (17,18)\}$.
1717 \begin{listing}[htb]
1718 \def\charwidth{5.7pt}
1719 \def\indent{4*\charwidth}
1720 \def\lineheight{\baselineskip}
1721 \def\mintedtop{\lineheight-1pt}
1723 \begin{tikzpicture}[overlay, yscale=-1]
1724 \tikzstyle{overlaybox}=[fill=lightgray,opacity=0.2]
1727 \draw[overlaybox] (2*\charwidth,\mintedtop) rectangle
1728 +(16*\charwidth,\lineheight);
1731 \draw[overlaybox] (2*\charwidth,\mintedtop+\lineheight) rectangle
1732 +(16*\charwidth,\lineheight);
1735 \draw[overlaybox] (2*\charwidth,\mintedtop+2*\lineheight) rectangle
1736 +(16*\charwidth,\lineheight);
1738 \draw[overlaybox] (\indent-3*\charwidth,\mintedtop) rectangle
1739 +(18*\charwidth,2*\lineheight);
1741 \draw[overlaybox] (3*\charwidth,\mintedtop+\lineheight) rectangle
1742 +(14*\charwidth,2*\lineheight);
1745 \draw[overlaybox] (\indent,\mintedtop) rectangle
1746 +(12*\charwidth,3*\lineheight);
1748 % indent should be 5 spaces
1749 \begin{minted}[linenos,firstnumber=16]{java}
1754 \caption{Example of how the text selections generator would generate text
1755 selections based on a lists of statements. Each highlighted rectangle
1756 represents a text selection.}
1757 \label{lst:textSelectionsExample}
1760 Each nesting level of a method body can have many such sequences of statements.
1761 The outermost nesting level has one such sequence, and each branch contains
1762 its own sequence of statements. \Myref{lst:grandExample} has a version of some
1763 code where all such sequences of statements are highlighted for a method body.
1765 To complete our example of possible text selections, I will now list all
1766 possible text selections for the method in \myref{lst:grandExample}, by nesting
1767 level. There are 23 of them in total.
1770 \item[Level 1 (10 selections)] \hfill \\
1771 $\{(5,9), (11), (12), (14,21), (5,11), (5,12), (5,21), (11,12),
1772 (11,21), \\(12,21)\}$
1774 \item[Level 2 (13 selections)] \hfill \\
1775 $\{(6), (7), (8), (6,7), (6,8), (7,8), (16), (17), (18), (16,17), (16,18), \\
1779 \subsubsection{The complexity}\label{sec:complexity}
1780 The complexity of how many text selections that need to be analyzed for a body
1781 of in total $n$ statements, is bounded by $O(n^2)$. A body of statements is here
1782 all the statements in all nesting levels of a sequence of statements. A method
1783 body (or a block) is a body of statements. To prove that the complexity is
1784 bounded by $O(n^2)$, I present a couple of theorems and prove them.
1787 The number of text selections that need to be analyzed for each list of
1788 statements of length $n$, is exactly
1791 \sum_{i=1}^{n} i = \frac{n(n+1)}{2}
1792 \label{eq:complexityStatementList}
1794 \label{thm:numberOfTextSelection}
1798 For $n=1$ this is trivial: $\frac{1(1+1)}{2} = \frac{2}{2} = 1$. One statement
1799 equals one selection.
1801 For $n=2$, you get one text selection for the first statement, one selection
1802 for the second statement, and one selection for the two of them combined.
1803 This equals three selections. $\frac{2(2+1)}{2} = \frac{6}{2} = 3$.
1805 For $n=3$, you get 3 selections for the two first statements, as in the case
1806 where $n=2$. In addition you get one selection for the third statement itself,
1807 and two more statements for the combinations of it with the two previous
1808 statements. This equals six selections. $\frac{3(3+1)}{2} = \frac{12}{2} = 6$.
1810 Assume that for $n=k$ there exists $\frac{k(k+1)}{2}$ text selections. Then we
1811 want to add selections for another statement, following the previous $k$
1812 statements. So, for $n=k+1$, we get one additional selection for the statement
1813 itself. Then we get one selection for each pair of the new selection and the
1814 previous $k$ statements. So the total number of selections will be the number
1815 of already generated selections, plus $k$ for every pair, plus one for the
1816 statement itself: $\frac{k(k+1)}{2} + k +
1817 1 = \frac{k(k+1)+2k+2}{2} = \frac{k(k+1)+2(k+1)}{2} = \frac{(k+1)(k+2)}{2} =
1818 \frac{(k+1)((k+1)+1)}{2} = \sum_{i=1}^{k+1} i$
1821 %\definition{A \emph{body of statements} is a sequence of statements where every
1822 %statement may have sub-statements.}
1825 The number of text selections for a body of statements is maximized if all the
1826 statements are at the same level.
1827 \label{thm:textSelectionsMaximized}
1831 Assume we have a body of, in total, $k$ statements. Then, the sum of the
1832 lengths of all the lists of statements in the body, is also $k$. Let
1833 $\{l,\ldots,m,(k-l-\ldots-m)\}$ be the lengths of the lists of statements in
1834 the body, with $l+\ldots+m<k \Rightarrow \forall i \in \{l,\ldots,m\} : i < k$.
1836 Then, the number of text selections that are generated for the $k$ statements
1842 \frac{l(l+1)}{2} + \ldots + \frac{m(m+1)}{2} +
1843 \frac{(k-l-\ldots-m)((k-l-\ldots-m)+ 1)}{2} = \\
1844 \frac{l^2+l}{2} + \ldots + \frac{m^2+m}{2} + \frac{k^2 - 2kl - \ldots - 2km +
1845 l^2 + \ldots + m^2 + k - l - \ldots - m}{2} = \\
1846 \frac{2l^2 - 2kl + \ldots + 2m^2 - 2km + k^2 + k}{2}
1850 \noindent It then remains to show that this inequality holds:
1853 \frac{2l^2 - 2kl + \ldots + 2m^2 - 2km + k^2 + k}{2} < \frac{k(k+1)}{2} =
1857 \noindent By multiplication by $2$ on both sides, and by removing the equal
1861 2l^2 - 2kl + \ldots + 2m^2 - 2km < 0
1864 Since $\forall i \in \{l,\ldots,m\} : i < k$, we have that $\forall i \in
1865 \{l,\ldots,m\} : 2ki > 2i^2$, so all the pairs of parts on the form $2i^2-2ki$
1866 are negative. In sum, the inequality holds.
1870 Therefore, the complexity for the number of selections that need to be analyzed
1871 for a body of $n$ statements is $O\bigl(\frac{n(n+1)}{2}\bigr) = O(n^2)$.
1873 \section{Disqualifying a selection}
1874 Certain text selections would lead to broken code if used as input to the
1875 \ExtractAndMoveMethod refactoring. To avoid this, we have to check all text
1876 selections for such conditions before they are further analyzed. This section
1877 is therefore going to present some properties that make a selection unsuitable
1878 for our refactoring.
1880 \subsection{A call to a protected or package-private method}
1881 If a text selection contains a call to a protected or package-private method, it
1882 would not be safe to move it to another class. The reason for this, is that we
1883 cannot know if the called method is being overridden by some subclass of the
1884 \gloss{enclosingClass}, or not.
1886 Imagine that the protected method \method{foo} is declared in class \m{A},
1887 and overridden in class \m{B}. The method \method{foo} is called from within a
1888 selection done to a method in \m{A}. We want to extract and move this selection
1889 to another class. The method \method{foo} is not public, so the \MoveMethod
1890 refactoring must make it public, making the extracted method able to call it
1891 from the extracted method's new location. The problem is, that the now public
1892 method \method{foo} is overridden in a subclass, where it has a protected
1893 status. This makes the compiler complain that the subclass \m{B} is trying to
1894 reduce the visibility of a method declared in its superclass \m{A}. This is not
1895 allowed in Java, and for good reasons. It would make it possible to make a
1896 subclass that could not be a substitute for its superclass.
1898 The problem this check helps to avoid, is a little subtle. The problem does not
1899 arise in the class where the change is done, but in a class derived from it.
1900 This shows that classes acting as superclasses are especially fragile to
1901 introducing errors in the context of automated refactoring.
1903 This is also shown in bug\ldots \todoin{File Eclipse bug report}
1906 \subsection{A double class instance creation}
1907 The following is a problem caused solely by the underlying \MoveMethod
1908 refactoring. The problem occurs if two classes are instantiated such that the
1909 first constructor invocation is an argument to a second, and that the first
1910 constructor invocation takes an argument that is built up using a field. As an
1911 example, say that \var{name} is a field of the enclosing class, and we have the
1912 expression \code{new A(new B(name))}. If this expression is located in a
1913 selection that is moved to another class, \var{name} will be left untouched,
1914 instead of being prefixed with a variable of the same type as it is declared in.
1915 If \var{name} is the destination for the move, it is not replaced by
1916 \code{this}, or removed if it is a prefix to a member access
1917 (\code{name.member}), but it is still left by itself.
1919 Situations like this would lead to code that will not compile. Therefore, we
1920 have to avoid them by not allowing selections to contain such double class
1921 instance creations that also contain references to fields.
1923 \todoin{File Eclipse bug report}
1926 \subsection{Instantiation of non-static inner class}
1927 When a non-static inner class is instantiated, this must happen in the scope of
1928 its declaring class. This is because it must have access to the members of the
1929 declaring class. If the inner class is public, it is possible to instantiate it
1930 through an instance of its declaring class, but this is not handled by the
1931 underlying \MoveMethod refactoring.
1933 Performing a move on a method that instantiates a non-static inner class, will
1934 break the code if the instantiation is not handled properly. For this reason,
1935 selections that contain instantiations of non-static inner classes are deemed
1936 unsuitable for the \ExtractAndMoveMethod refactoring.
1938 \subsection{References to enclosing instances of the enclosing class}
1939 The title of this section may be a little hard to grasp at first. What it means
1940 is that there is a (non-static) class \m{C} that is declared in the scope of
1941 possibly multiple other classes. And there is a statement in the body of a
1942 method declared in class \m{C}, that contains a reference to one or more
1943 instances of these enclosing classes of \m{C}.
1945 The problem with this, is that these references may not be valid if they are
1946 moved to another class. Theoretically, some situations could easily be solved by
1947 passing, to the moved method, a reference to the instance where the problematic
1948 referenced member is declared. This should work in the case where this member is
1949 publicly accessible. This is not done in the underlying \MoveMethod refactoring,
1950 so it cannot be allowed in the \ExtractAndMoveMethod refactoring either.
1952 \subsection{Inconsistent return statements}
1953 To verify that a text selection is consistent with respect to return statements,
1954 we must check that if a selection contains a return statement, then every
1955 possible execution path within the selection ends in either a return or a throw
1956 statement. This property is important regarding the \ExtractMethod refactoring.
1957 If it holds, it means that a method could be extracted from the selection, and a
1958 call to it could be substituted for the selection. If the method has a non-void
1959 return type, then a call to it would also be a valid return point for the
1960 calling method. If its return value is of the void type, then the \ExtractMethod
1961 refactoring will append an empty return statement to the back of the method
1962 call. Therefore, the analysis does not discriminate on either kind of return
1963 statements, with or without a return value.
1965 A throw statement is accepted anywhere a return statement is required. This is
1966 because a throw statement causes an immediate exit from the current block,
1967 together with all outer blocks in its control flow that does not catch the
1970 Return statements can be either explicit or implicit. An \emph{explicit} return
1971 statement is formed by using the \code{return} keyword, while an \emph{implicit}
1972 return statement is a statement that is not formed using \code{return}, but must
1973 be the last statement of a method that can have any side effects. This can
1974 happen in methods with a void return type. An example is a statement that is
1975 inside one or more blocks. The last statement of a method could for instance be
1976 a synchronized statement, but the last statement that is executed in the method,
1977 and that can have any side effects, may be located inside the body of the
1978 synchronized statement.
1980 We can start the check for this property by looking at the last statement of a
1981 selection to see if it is a return statement (explicit or implicit) or a throw
1982 statement. If this is the case, then the property holds, assuming the selected
1983 code do not contain any compilation errors. All execution paths within the
1984 selection should end in either this, or another, return or throw statement.
1985 \todoin{State somewhere that we assume no compilation errors?}
1987 If the last statement of the selection is not a return or throw, the execution
1988 of it must eventually end in one for the selection to be legal. This means that
1989 all branches of the last statement of every branch must end in a return or
1990 throw. Given this recursive definition, there are only five types of statements
1991 that are guaranteed to end in a return or throw if their child branches do. All
1992 other statements would have to be considered illegal. The first three:
1993 Block-statements, labeled statements and do-statements are all kinds of
1994 fall-through statements that always get their body executed. Do-statements would
1995 not make much sense if written such that they
1996 always ends after the first round of execution of their body, but that is not
1997 our concern. The remaining two statements that can end in a return or throw are
1998 if-statements and try-statements.
2000 For an if-statement, the rule is that if its then-part does not contain any
2001 return or throw statements, this is considered illegal. If the then-part does
2002 contain a return or throw, the else-part is checked. If its else-part is
2003 non-existent, or it does not contain any return or throw statements, the
2004 statement is considered illegal. If an if-statement is not considered illegal,
2005 the bodies of its two parts must be checked.
2007 Try-statements are handled much the same way as if-statements. The body of a
2008 try-statement must contain a return or throw. The same applies to its catch
2009 clauses and finally body.
2011 \subsection{Ambiguous return values}
2012 The problem with ambiguous return values arises when a selection is chosen to be
2013 extracted into a new method, but it needs to return more than one value from
2016 This problem occurs in two situations. The first situation arises when there is
2017 more than one local variable that is both assigned to within a selection and
2018 also referenced after the selection. The other situation occurs when there is
2019 only one such assignment, but the selection also contain return statements.
2021 Therefore we must examine the selection for assignments to local variables that
2022 are referenced after the text selection. Then we must verify that not more than
2023 one such reference is done, or zero if any return statements are found.
2025 \subsection{Illegal statements}
2026 An illegal statement may be a statement that is of a type that is never allowed,
2027 or it may be a statement of a type that is only allowed if certain conditions
2030 Any use of the \var{super} keyword is prohibited, since its meaning is altered
2031 when moving a method to another class.
2033 For a \emph{break} statement, there are two situations to consider: A break
2034 statement with or without a label. If the break statement has a label, it is
2035 checked that whole of the labeled statement is inside the selection. If the
2036 break statement does not have a label attached to it, it is checked that its
2037 innermost enclosing loop or switch statement also is inside the selection.
2039 The situation for a \emph{continue} statement is the same as for a break
2040 statement, except that it is not allowed inside switch statements.
2042 Regarding \emph{assignments}, two types of assignments are allowed: Assignments
2043 to non-final variables and assignments to array access. All other assignments
2044 are regarded illegal.
2046 \todoin{Expand with more illegal statements and/or conclude that I did not have
2047 time to analyze all statement types.}
2049 \section{Disqualifying selections from the
2050 example}\label{sec:disqualifyingExample}
2051 Among the selections we found for the code in \myref{lst:grandExample}, not many
2052 of them must be disqualified on the basis of containing something illegal. The
2053 only statement causing trouble is the break statement in line 18. None of the
2054 selections on nesting level 2 can contain this break statement, since the
2055 innermost switch statement is not inside any of these selections.
2057 This means that the text selections $(18)$, $(16,18)$ and $(17,18)$ can be
2058 excluded from further consideration, and we are left with the following
2062 \item[Level 1 (10 selections)] \hfill \\
2063 $\{(5,9), (11), (12), (14,21), (5,11), (5,12), (5,21), (11,12),
2064 (11,21), \\(12,21)\}$
2066 \item[Level 2 (10 selections)] \hfill \\
2067 $\{(6), (7), (8), (6,7), (6,8), (7,8), (16), (17), (16,17), (20)\}$
2070 \section{Finding a move target}
2071 In the analysis needed to perform the \ExtractAndMoveMethod refactoring
2072 automatically, the selection we choose is found among all the selections that
2073 have a possible move target. Therefore, the best possible move target must be
2074 found for all the candidate selections, so that we are able to sort out the
2075 selection that is best suited for the refactoring.
2077 To find the best move target for a specific text selection, we first need to
2078 find all the possible targets. Since the target must be a local variable or a
2079 field, we are basically looking for names within the selection; names that
2080 represents references to variables.
2082 The names we are looking for, we call prefixes. This is because we are not
2083 interested in names that occur in the middle of a dot-separated sequence of
2084 names. We are only interested in names constituting prefixes of other names, and
2085 possibly themselves. The reason for this, is that two lexically equal names need
2086 not be referencing the same variable, if they themselves are not referenced via
2087 the same prefix. Consider the two method calls \code{a.x.foo()} and
2088 \code{b.x.foo()}. Here, the two references to \code{x}, in the middle of the
2089 qualified names both preceding \code{foo()}, are not referencing the same
2090 variable. Even though the variables may share the type, and the method
2091 \method{foo} thus is the same for both, we would not know through which of the
2092 variables \var{a} or \var{b} we should call the extracted method.
2094 The possible move targets are then the prefixes that are not among a subset of
2095 the prefixes that are not valid move targets \see{s:unfixes}. Also, prefixes
2096 that are just simple names, and have only one occurrence, are left out. This is
2097 because they are not going to have any positive effect on coupling between
2098 classes, and are only going to increase the complexity of the code.
2100 For finding the best move target among these safe prefixes, a simple heuristic
2101 is used. It is as simple as choosing the prefix that is most frequently
2102 referenced within the selection.
2104 \section{Unfixes}\label{s:unfixes}
2105 The prefixes that are not valid as move targets are called unfixes.
2107 A name that is assigned to within a selection can be an unfix. The reason for
2108 this is that the result would be an assignment to the \type{this} keyword, which
2109 is not valid in Java \see{eclipse_bug_420726}.
2111 Prefixes that originate from variable declarations within the same selection are
2112 also considered unfixes. This is because when a method is moved, it needs to be
2113 called through a variable. If this variable is also declared within the method
2114 that is to be moved, this obviously cannot be done.
2116 Also considered as unfixes are variable references that are of types that are
2117 not suitable for moving methods to. This can either be because it is not
2118 physically possible to move a method to the desired class or that it will cause
2119 compilation errors by doing so.
2121 If the type binding for a name is not resolved it is considered and unfix. The
2122 same applies to types that are only found in compiled code, so they have no
2123 underlying source that is accessible to us. (E.g. the \type{java.lang.String}
2126 Interfaces types are not suitable as targets. This is simply because interfaces
2127 in Java cannot contain methods with bodies. (This thesis does not deal with
2128 features of Java versions later than Java 7. Java 8 has interfaces with default
2129 implementations of methods.)
2131 Neither is local types allowed. This accounts for both local and anonymous
2132 classes. Anonymous classes are effectively the same as interface types with
2133 respect to unfixes. Local classes could in theory be used as targets, but this
2134 is not possible due to limitations of the way the \refa{Extract and Move Method}
2135 refactoring has to be implemented. The problem is that the refactoring is done
2136 in two steps, so the intermediate state between the two refactorings would not
2137 be legal Java code. In the intermediate step for the case where a local class is
2138 the move target, the extracted method would need to take the local class as a
2139 parameter. This new method would need to live in the scope of the declaring
2140 class of the originating method. The local class would then not be in the scope
2141 of the extracted method, thus bringing the source code into an illegal state.
2142 One could imagine that the method was extracted and moved in one operation,
2143 without an intermediate state. Then it would make sense to include variables
2144 with types of local classes in the set of legal targets, since the local classes
2145 would then be in the scopes of the method calls. If this makes any difference
2146 for software metrics that measure coupling would be a different discussion.
2149 \begin{listing}[htb]
2150 \begin{multicols}{2}
2151 \begin{minted}[frame=topline,label=Before,framesep=\mintedframesep]{java}
2152 void declaresLocalClass() {
2167 \begin{minted}[frame=topline,label={After Extract
2168 Method},framesep=\mintedframesep]{java}
2169 void declaresLocalClass() {
2180 // Intermediate step
2181 void fooBar(LocalClass inst) {
2187 \caption{When the \refa{Extract and Move Method} tries to use a variable with a
2188 local type as the move target, an intermediate step is performed that is not
2189 allowed. Here: \type{LocalClass} is not in the scope of \method{fooBar} in its
2190 intermediate location.}
2191 \label{lst:extractMethod_LocalClass}
2194 The last class of names that are considered unfixes are names used in null
2195 tests. These are tests that read like this: if \code{<name>} equals \var{null}
2196 then do something. If allowing variables used in those kinds of expressions as
2197 targets for moving methods, we would end up with code containing boolean
2198 expressions like \code{this == null}, which would not be meaningful, since
2199 \var{this} would never be \var{null}.
2201 \section{Finding the example selections that have possible targets}
2202 We now pick up the thread from \myref{sec:disqualifyingExample} where we have a
2203 set of text selections that need to be analyzed to find out if some of them are
2204 suitable targets for the \ExtractAndMoveMethod refactoring.
2206 We start by analyzing the text selections for nesting level 2, because these
2207 results can be used to reason about the selections for nesting level 1. First we
2208 have all the single-statement selections.
2211 \item[Selections $(6)$, $(8)$ and $(20)$.] \hfill \\
2212 All these selections have a prefix that contains a possible target, namely
2213 the variable \var{a}. The problem is that the prefixes are only one segment
2214 long, and their frequency counts are only 1 as well. None of these
2215 selections are therefore considered as suitable candidates for the
2218 \item[Selection $(7)$.] \hfill \\
2219 This selection contains the unfix \var{a}, and no other possible targets.
2220 The reason for \var{a} being an unfix is that it is assigned to within the
2221 selection. Selection $(7)$ is therefore unsuited as a refactoring candidate.
2223 \item[Selections $(16)$ and $(17)$.] \hfill \\
2224 These selections both have a possible target. The target for both selections
2225 is the variable \var{b}. Both the prefixes have frequency 1. We denote this
2226 with the new tuples $((16), \texttt{b.a}, f(1))$ and $((17), \texttt{b.a},
2227 f(1))$. They contain the selection, the prefix with the target and the
2228 frequency for this prefix.
2232 Then we have all the text selections from level 2 that are composed of multiple
2236 \item[Selections $(6,7)$, $(6,8)$ and $(7,8)$.] \hfill \\
2237 All these selections are disqualified for the reason that they contain the
2238 unfix \var{a}, due to the assignment, and no other possible move targets.
2240 \item[Selection $(16,17)$.] \hfill \\
2241 This selection is the last selection that is not analyzed on nesting level
2242 2. It contains only one possible move target, and that is the variable
2243 \var{b}. It also contains only one prefix \var{b.a}, with frequency count
2244 2. Therefore we have a new candidate $((16,17), \texttt{b.a}, f(2))$.
2248 Moving on to the text selections for nesting level 1, starting with the
2249 single-statement selections:
2252 \item[Selection $(5,9)$.] \hfill \\
2253 This selection contains two variable references that must be analyzed to see
2254 if they are possible move candidates. The first one is the variable
2255 \var{bool}. This variable is of type \type{boolean}, which is a primary type
2256 and therefore not possible to make any changes to. The second variable is
2257 \var{a}. The variable \var{a} is an unfix in $(5,9)$, for the same reason as
2258 in the selections $(6,7)$, $(7,8)$ and $(6,8)$. So selection $(5,9)$
2259 contains no possible move targets.
2261 \item[Selections $(11)$ and $(12)$.] \hfill \\
2262 These selections are disqualified for the same reasons as selections $(6)$
2263 and $(8)$. Their prefixes are one segment long and are referenced only one
2266 \item[Selection $(14,21)$] \hfill \\
2267 This is the switch statement from \myref{lst:grandExample}. It contains the
2268 relevant variable references \var{val}, \var{a} and \var{b}. The variable
2269 \var{val} is a primary type, just as \var{bool}. The variable \var{a} is
2270 only found in one statement, and in a prefix with only one segment, so it is
2271 not considered to be a possible move target. The only variable left is
2272 \var{b}. Just as in the selection $(16,17)$, \var{b} is part of the prefix
2273 \code{b.a}, which has 2 appearances. We have found a new candidate
2274 $((14,21), \texttt{b.a}, f(2))$.
2278 It remains to see if we get any new candidates by analyzing the multi-statement
2279 text selections for nesting level 1:
2282 \item[Selections $(5,11)$ and $(5,12)$.] \hfill \\
2283 These selections are disqualified for the same reason as $(5,9)$. The only
2284 possible move target \var{a} is an unfix.
2286 \item[Selection $(5,21)$.] \hfill \\
2287 This is whole of the method body in \myref{lst:grandExample}. The variables
2288 \var{a}, \var{bool} and \var{val} are either an unfix or primary types. The
2289 variable \var{b} is the only possible move target, and we have, again, the
2290 prefix \code{b.a} as the center of attention. The refactoring candidate is
2291 $((5,21), \texttt{b.a}, f(2))$.
2293 \item[Selection $(11,12)$.] \hfill \\
2294 This small selection contains the prefix \code{a} with frequency 2, and no
2295 unfixes. The candidate is $((11,12), \texttt{a}, f(2))$.
2297 \item[Selection $(11,21)$] \hfill \\
2298 This selection contains the selection $(11,12)$ in addition to the switch
2299 statement. The selection has two possible move targets. The first one is
2300 \var{b}, in a prefix with frequency 2. The second is \var{a}, although it
2301 is in a simple prefix, it is referenced 3 times, and is therefore chosen
2302 as the target for the selection. The new candidate is $((11,21),
2305 \item[Selection $(12,21)$.] \hfill \\
2306 This selection is similar to the previous $(11,21)$, only that \var{a} now
2307 has frequency count 2. This means that the prefixes \code{a} and
2308 \code{b.a} both have the count 2. Of the two, \code{b.a} is preferred,
2309 since it has more segments than \code{a}. Thus the candidate for this
2310 selection is $((12,21), \texttt{b.a}, f(2))$.
2314 For the method in \myref{lst:grandExample} we therefore have the following 8
2315 candidates for the \ExtractAndMoveMethod refactoring: $\{((16), \texttt{b.a},
2316 f(1)), ((17), \texttt{b.a}, f(1)), ((16,17), \texttt{b.a}, f(2)), ((14,21),
2317 \texttt{b.a}, f(2)), \\ ((5,21), \texttt{b.a}, f(2)), ((11,12), \texttt{a},
2318 f(2)), ((11,21), \texttt{a}, f(3)), ((12,21), \texttt{b.a}, f(2))\}$.
2320 It now only remains to specify an order for these candidates, so we can choose
2321 the most suitable one to refactor.
2324 \section{Choosing the selection}\label{sec:choosingSelection}
2325 When choosing a selection between the text selections that have possible move
2326 targets, the selections need to be ordered. The criteria below are presented in
2327 the order they are prioritized. If not one selection is favored over the other
2328 for a concrete criterion, the selections are evaluated by the next criterion.
2331 \item The first criterion that is evaluated is whether a selection contains
2332 any unfixes or not. If selection \m{A} contains no unfixes, while selection
2333 \m{B} does, selection \m{A} is favored over selection \m{B}. This is
2334 because, if we can, we want to avoid moving such as assignments and variable
2335 declarations. This is done under the assumption that, if possible, avoiding
2336 selections containing unfixes will make the code moved a little cleaner.
2338 \item The second criterion that is evaluated is whether a selection contains
2339 multiple possible targets or not. If selection \m{A} has only one possible
2340 target, and selection \m{B} has multiple, selection \m{A} is favored. If
2341 both selections have multiple possible targets, they are considered equal
2342 with respect to this criterion. The rationale for this heuristic is that we
2343 would prefer not to introduce new couplings between classes when performing
2344 the \ExtractAndMoveMethod refactoring.
2346 \item When evaluating this criterion, this is with the knowledge that
2347 selection \m{A} and \m{B} both have one possible target, or multiple
2348 possible targets. Then, if the move target candidate of selection \m{A} has
2349 a higher reference count than the target candidate of selection \m{B},
2350 selection \m{A} is favored. The reason for this is that we would like to
2351 move the selection that gets rid of the most references to another class.
2353 \item The last criterion is that if the frequencies of the targets chosen for
2354 both selections are equal, the selection with the target that is part of the
2355 prefix with highest number of segments is favored. This is done to favor
2360 If none of the above mentioned criteria favors one selection over another, the
2361 selections are considered to be equally good candidates for the
2362 \ExtractAndMoveMethod refactoring.
2364 \section{Concluding the example}
2365 For choosing one of the remaining selections, we need to order our candidates
2366 after the criteria in the previous section. Below we have the candidates ordered
2367 in descending order, with the ``best'' candidate first:
2369 \begin{multicols}{2}
2371 \item $((16,17), \texttt{b.a}, f(2))$
2372 \item $((11,12), \texttt{a}, f(2))$
2373 \item $((16), \texttt{b.a}, f(1))$
2374 \item $((17), \texttt{b.a}, f(1))$
2377 % Many possible targets
2378 \item $((11,21), \texttt{a}, f(3))$
2379 \item $((5,21), \texttt{b.a}, f(2))$
2380 \item $((12,21), \texttt{b.a}, f(2))$
2381 \item $((14,21), \texttt{b.a}, f(2))$
2406 The candidates ordered 5-8 all have unfixes in them, therefore they are ordered
2407 last. None of the candidates ordered 1-4 have multiple possible targets. The
2408 only interesting discussion is now why $(16,17)$ was favored over $(11,12)$.
2409 This is because the prefix \code{b.a} enclosing the move target of selection
2410 $(16,17)$ has one more segment than the prefix \code{a} of $(11,12)$.
2412 The selection is now extracted into a new method \method{gen\_123} and then
2413 moved to class \type{B}, since that is the type of the variable \var{b} that is
2414 the target from the chosen refactoring candidate. The name of the method has a
2415 randomly generated suffix. In the refactoring implementation, the extracted
2416 methods follow the pattern \code{generated\_<long>}, where \code{<long>} is a
2417 pseudo-random long value. This is shortened here to make the example readable.
2418 The result after the refactoring is shown in \myref{lst:grandExampleResult}.
2420 \begin{listing}[htb]
2421 \begin{multicols}{2}
2422 \begin{minted}[linenos]{java}
2424 A a; B b; boolean bool;
2426 void method(int val) {
2449 \begin{minted}[]{java}
2453 public void gen_123(C c) {
2461 \caption{The result after refactoring the class \type{C} of
2462 \myref{lst:grandExample} with the \ExtractAndMoveMethod refactoring with
2463 $((16,17), \texttt{b.a}, f(2))$ as input.}
2464 \label{lst:grandExampleResult}
2469 \todoin{Short description of changing code and error handling?}
2471 \todoin{Pointing to implementation chapter}
2473 \chapter{Refactorings in Eclipse JDT: Design and
2474 Shortcomings}\label{ch:jdt_refactorings}
2476 This chapter will deal with some of the design behind refactoring support in
2477 \name{Eclipse}, and the JDT in specific. After which it will follow a section
2478 about shortcomings of the refactoring API in terms of composition of
2482 The refactoring world of \name{Eclipse} can in general be separated into two parts: The
2483 language independent part and the part written for a specific programming
2484 language -- the language that is the target of the supported refactorings.
2485 \todo{What about the language specific part?}
2487 \subsection{The Language Toolkit}
2488 The Language Toolkit\footnote{The content of this section is a mixture of
2489 written material from
2490 \url{https://www.eclipse.org/articles/Article-LTK/ltk.html} and
2491 \url{http://www.eclipse.org/articles/article.php?file=Article-Unleashing-the-Power-of-Refactoring/index.html},
2492 the LTK source code and my own memory.}, or LTK for short, is the framework that
2493 is used to implement refactorings in \name{Eclipse}. It is language independent and
2494 provides the abstractions of a refactoring and the change it generates, in the
2495 form of the classes \typewithref{org.eclipse.ltk.core.refactoring}{Refactoring}
2496 and \typewithref{org.eclipse.ltk.core.refactoring}{Change}.
2498 There are also parts of the LTK that is concerned with user interaction, but
2499 they will not be discussed here, since they are of little value to us and our
2500 use of the framework. We are primarily interested in the parts that can be
2503 \subsubsection{The Refactoring Class}
2504 The abstract class \type{Refactoring} is the core of the LTK framework. Every
2505 refactoring that is going to be supported by the LTK has to end up creating an
2506 instance of one of its subclasses. The main responsibilities of subclasses of
2507 \type{Refactoring} are to implement template methods for condition checking
2508 (\methodwithref{org.eclipse.ltk.core.refactoring.Refactoring}{checkInitialConditions}
2510 \methodwithref{org.eclipse.ltk.core.refactoring.Refactoring}{checkFinalConditions}),
2512 \methodwithref{org.eclipse.ltk.core.refactoring.Refactoring}{createChange}
2513 method that creates and returns an instance of the \type{Change} class.
2515 If the refactoring shall support that others participate in it when it is
2516 executed, the refactoring has to be a processor-based
2517 refactoring\typeref{org.eclipse.ltk.core.refactoring.participants.ProcessorBasedRefactoring}.
2518 It then delegates to its given
2519 \typewithref{org.eclipse.ltk.core.refactoring.participants}{RefactoringProcessor}
2520 for condition checking and change creation. Participating in a refactoring can
2521 be useful in cases where the changes done to programming source code affect
2522 other related resources in the workspace. This can be names or paths in
2523 configuration files, or maybe one would like to perform additional logging of
2524 changes done in the workspace.
2526 \subsubsection{The Change Class}
2527 This class is the base class for objects that is responsible for performing the
2528 actual workspace transformations in a refactoring. The main responsibilities for
2529 its subclasses are to implement the
2530 \methodwithref{org.eclipse.ltk.core.refactoring.Change}{perform} and
2531 \methodwithref{org.eclipse.ltk.core.refactoring.Change}{isValid} methods. The
2532 \method{isValid} method verifies that the change object is valid and thus can be
2533 executed by calling its \method{perform} method. The \method{perform} method
2534 performs the desired change and returns an undo change that can be executed to
2535 reverse the effect of the transformation done by its originating change object.
2537 \subsubsection{Executing a Refactoring}\label{executing_refactoring}
2538 The life cycle of a refactoring generally follows two steps after creation:
2539 condition checking and change creation. By letting the refactoring object be
2541 \typewithref{org.eclipse.ltk.core.refactoring}{CheckConditionsOperation} that
2542 in turn is handled by a
2543 \typewithref{org.eclipse.ltk.core.refactoring}{CreateChangeOperation}, it is
2544 assured that the change creation process is managed in a proper manner.
2546 The actual execution of a change object has to follow a detailed life cycle.
2547 This life cycle is honored if the \type{CreateChangeOperation} is handled by a
2548 \typewithref{org.eclipse.ltk.core.refactoring}{PerformChangeOperation}. If also
2549 an undo manager\typeref{org.eclipse.ltk.core.refactoring.IUndoManager} is set
2550 for the \type{PerformChangeOperation}, the undo change is added into the undo
2553 \section{Shortcomings}
2554 This section is introduced naturally with a conclusion: The JDT refactoring
2555 implementation does not facilitate composition of refactorings.
2556 \todo{refine}This section will try to explain why, and also identify other
2557 shortcomings of both the usability and the readability of the JDT refactoring
2560 I will begin at the end and work my way toward the composition part of this
2563 \subsection{Absence of Generics in Eclipse Source Code}
2564 This section is not only concerning the JDT refactoring API, but also large
2565 quantities of the \name{Eclipse} source code. The code shows a striking absence of the
2566 Java language feature of generics. It is hard to read a class' interface when
2567 methods return objects or takes parameters of raw types such as \type{List} or
2568 \type{Map}. This sometimes results in having to read a lot of source code to
2569 understand what is going on, instead of relying on the available interfaces. In
2570 addition, it results in a lot of ugly code, making the use of typecasting more
2571 of a rule than an exception.
2573 \subsection{Composite Refactorings Will Not Appear as Atomic Actions}
2575 \subsubsection{Missing Flexibility from JDT Refactorings}
2576 The JDT refactorings are not made with composition of refactorings in mind. When
2577 a JDT refactoring is executed, it assumes that all conditions for it to be
2578 applied successfully can be found by reading source files that have been
2579 persisted to disk. They can only operate on the actual source material, and not
2580 (in-memory) copies thereof. This constitutes a major disadvantage when trying to
2581 compose refactorings, since if an exception occurs in the middle of a sequence
2582 of refactorings, it can leave the project in a state where the composite
2583 refactoring was only partially executed. It makes it hard to discard the changes
2584 done without monitoring and consulting the undo manager, an approach that is not
2587 \subsubsection{Broken Undo History}
2588 When designing a composed refactoring that is to be performed as a sequence of
2589 refactorings, you would like it to appear as a single change to the workspace.
2590 This implies that you would also like to be able to undo all the changes done by
2591 the refactoring in a single step. This is not the way it appears when a sequence
2592 of JDT refactorings is executed. It leaves the undo history filled up with
2593 individual undo actions corresponding to every single JDT refactoring in the
2594 sequence. This problem is not trivial to handle in \name{Eclipse}
2595 \see{hacking_undo_history}.
2599 \chapter{Composite Refactorings in Eclipse}
2601 \section{A Simple Ad Hoc Model}
2602 As pointed out in \myref{ch:jdt_refactorings}, the \name{Eclipse} JDT refactoring model
2603 is not very well suited for making composite refactorings. Therefore a simple
2604 model using changer objects (of type \type{RefaktorChanger}) is used as an
2605 abstraction layer on top of the existing \name{Eclipse} refactorings, instead of
2606 extending the \typewithref{org.eclipse.ltk.core.refactoring}{Refactoring} class.
2608 The use of an additional abstraction layer is a deliberate choice. It is due to
2609 the problem of creating a composite
2610 \typewithref{org.eclipse.ltk.core.refactoring}{Change} that can handle text
2611 changes that interfere with each other. Thus, a \type{RefaktorChanger} may, or
2612 may not, take advantage of one or more existing refactorings, but it is always
2613 intended to make a change to the workspace.
2615 \subsection{A typical \type{RefaktorChanger}}
2616 The typical refaktor changer class has two responsibilities, checking
2617 preconditions and executing the requested changes. This is not too different
2618 from the responsibilities of an LTK refactoring, with the distinction that a
2619 refaktor changer also executes the change, while an LTK refactoring is only
2620 responsible for creating the object that can later be used to do the job.
2622 Checking of preconditions is typically done by an
2623 \typewithref{no.uio.ifi.refaktor.analyze.analyzers}{Analyzer}. If the
2624 preconditions validate, the upcoming changes are executed by an
2625 \typewithref{no.uio.ifi.refaktor.change.executors}{Executor}.
2627 \section{The Extract and Move Method Refactoring}
2628 %The Extract and Move Method Refactoring is implemented mainly using these
2631 % \item \type{ExtractAndMoveMethodChanger}
2632 % \item \type{ExtractAndMoveMethodPrefixesExtractor}
2633 % \item \type{Prefix}
2634 % \item \type{PrefixSet}
2637 \subsection{The Building Blocks}
2638 This is a composite refactoring, and hence is built up using several primitive
2639 refactorings. These basic building blocks are, as its name implies, the
2640 \ExtractMethod refactoring\citing{refactoring} and the \MoveMethod
2641 refactoring\citing{refactoring}. In \name{Eclipse}, the implementations of these
2642 refactorings are found in the classes
2643 \typewithref{org.eclipse.jdt.internal.corext.refactoring.code}{ExtractMethodRefactoring}
2645 \typewithref{org.eclipse.jdt.internal.corext.refactoring.structure}{MoveInstanceMethodProcessor},
2646 where the last class is designed to be used together with the processor-based
2647 \typewithref{org.eclipse.ltk.core.refactoring.participants}{MoveRefactoring}.
2649 \subsubsection{The ExtractMethodRefactoring Class}
2650 This class is quite simple in its use. The only parameters it requires for
2651 construction is a compilation
2652 unit\typeref{org.eclipse.jdt.core.ICompilationUnit}, the offset into the source
2653 code where the extraction shall start, and the length of the source to be
2654 extracted. Then you have to set the method name for the new method together with
2655 its visibility and some not so interesting parameters.
2657 \subsubsection{The MoveInstanceMethodProcessor Class}
2658 For the \refa{Move Method}, the processor requires a little more advanced input than
2659 the class for the \refa{Extract Method}. For construction it requires a method
2660 handle\typeref{org.eclipse.jdt.core.IMethod} for the method that is to be moved.
2661 Then the target for the move has to be supplied as the variable binding from a
2662 chosen variable declaration. In addition to this, some parameters have to be set
2663 regarding setters/getters, as well as delegation.
2665 To make the processor a working refactoring, a \type{MoveRefactoring} must be
2666 created with it as a parameter.
2668 \subsection{The ExtractAndMoveMethodChanger}
2670 The \typewithref{no.uio.ifi.refaktor.changers}{ExtractAndMoveMethodChanger}
2671 class is a subclass of the class
2672 \typewithref{no.uio.ifi.refaktor.changers}{RefaktorChanger}. It is responsible
2673 for analyzing and finding the best target for, and also executing, a composition
2674 of the \refa{Extract Method} and \refa{Move Method} refactorings. This particular changer is
2675 the one of my changers that is closest to being a true LTK refactoring. It can
2676 be reworked to be one if the problems with overlapping changes are resolved. The
2677 changer requires a text selection and the name of the new method, or else a
2678 method name will be generated. The selection has to be of the type
2679 \typewithref{no.uio.ifi.refaktor.utils}{CompilationUnitTextSelection}. This
2680 class is a custom extension to
2681 \typewithref{org.eclipse.jface.text}{TextSelection}, that in addition to the
2682 basic offset, length and similar methods, also carry an instance of the
2683 underlying compilation unit handle for the selection.
2686 \type{ExtractAndMoveMethodAnalyzer}}\label{extractAndMoveMethodAnalyzer}
2687 The analysis and precondition checking is done by the
2688 \typewithref{no.uio.ifi.refaktor.analyze.analyzers}{ExtractAnd\-MoveMethodAnalyzer}.
2689 First is check whether the selection is a valid selection or not, with respect
2690 to statement boundaries and that it actually contains any selections. Then it
2691 checks the legality of both extracting the selection and also moving it to
2692 another class. This checking of is performed by a range of checkers
2693 \see{checkers}. If the selection is approved as legal, it is analyzed to find
2694 the presumably best target to move the extracted method to.
2696 For finding the best suitable target the analyzer is using a
2697 \typewithref{no.uio.ifi.refaktor.analyze.collectors}{PrefixesCollector} that
2698 collects all the possible candidate targets for the refactoring. All the
2699 non-candidates are found by an
2700 \typewithref{no.uio.ifi.refaktor.analyze.collectors}{UnfixesCollector} that
2701 collects all the targets that will give some kind of error if used. (For
2702 details about the property collectors, see \myref{propertyCollectors}.) All
2703 prefixes (and unfixes) are represented by a
2704 \typewithref{no.uio.ifi.refaktor.extractors}{Prefix}, and they are collected
2705 into sets of prefixes. The safe prefixes are found by subtracting from the set
2706 of candidate prefixes the prefixes that is enclosing any of the unfixes. A
2707 prefix is enclosing an unfix if the unfix is in the set of its sub-prefixes. As
2708 an example, \code{``a.b''} is enclosing \code{``a''}, as is \code{``a''}. The
2709 safe prefixes is unified in a \type{PrefixSet}. If a prefix has only one
2710 occurrence, and is a simple expression, it is considered unsuitable as a move
2711 target. This occurs in statements such as \code{``a.foo()''}. For such
2712 statements it bares no meaning to extract and move them. It only generates an
2713 extra method and the calling of it.
2715 The most suitable target for the refactoring is found by finding the prefix with
2716 the most occurrences. If two prefixes have the same occurrence count, but they
2717 differ in the number of segments, the one with most segments is chosen.
2720 \type{ExtractAndMoveMethodExecutor}}\label{extractAndMoveMethodExecutor}
2721 If the analysis finds a possible target for the composite refactoring, it is
2723 \typewithref{no.uio.ifi.refaktor.change.executors}{ExtractAndMoveMethodExecutor}.
2724 It is composed of the two executors known as
2725 \typewithref{no.uio.ifi.refaktor.change.executors}{ExtractMethodRefactoringExecutor}
2727 \typewithref{no.uio.ifi.refaktor.change.executors}{MoveMethodRefactoringExecutor}.
2728 The \type{ExtractAndMoveMethodExecutor} is responsible for gluing the two
2729 together by feeding the \type{MoveMethod\-RefactoringExecutor} with the
2730 resources needed after executing the extract method refactoring.
2732 \subsubsection{The \type{ExtractMethodRefactoringExecutor}}
2733 This executor is responsible for creating and executing an instance of the
2734 \type{ExtractMethodRefactoring} class. It is also responsible for collecting
2735 some post execution resources that can be used to find the method handle for the
2736 extracted method, as well as information about its parameters, including the
2737 variable they originated from.
2739 \subsubsection{The \type{MoveMethodRefactoringExecutor}}
2740 This executor is responsible for creating and executing an instance of the
2741 \type{MoveRefactoring}. The move refactoring is a processor-based refactoring,
2742 and for the \refa{Move Method} refactoring it is the \type{MoveInstanceMethodProcessor}
2745 The handle for the method to be moved is found on the basis of the information
2746 gathered after the execution of the \refa{Extract Method} refactoring. The only
2747 information the \type{ExtractMethodRefactoring} is sharing after its execution,
2748 regarding find the method handle, is the textual representation of the new
2749 method signature. Therefore it must be parsed, the strings for types of the
2750 parameters must be found and translated to a form that can be used to look up
2751 the method handle from its type handle. They have to be on the unresolved
2752 form.\todo{Elaborate?} The name for the type is found from the original
2753 selection, since an extracted method must end up in the same type as the
2756 When analyzing a selection prior to performing the \refa{Extract Method} refactoring, a
2757 target is chosen. It has to be a variable binding, so it is either a field or a
2758 local variable/parameter. If the target is a field, it can be used with the
2759 \type{MoveInstanceMethodProcessor} as it is, since the extracted method still is
2760 in its scope. But if the target is local to the originating method, the target
2761 that is to be used for the processor must be among its parameters. Thus the
2762 target must be found among the extracted method's parameters. This is done by
2763 finding the parameter information object that corresponds to the parameter that
2764 was declared on basis of the original target's variable when the method was
2765 extracted. (The extracted method must take one such parameter for each local
2766 variable that is declared outside the selection that is extracted.) To match the
2767 original target with the correct parameter information object, the key for the
2768 information object is compared to the key from the original target's binding.
2769 The source code must then be parsed to find the method declaration for the
2770 extracted method. The new target must be found by searching through the
2771 parameters of the declaration and choose the one that has the same type as the
2772 old binding from the parameter information object, as well as the same name that
2773 is provided by the parameter information object.
2777 SearchBasedExtractAndMoveMethodChanger}\label{searchBasedExtractAndMoveMethodChanger}
2779 \typewithref{no.uio.ifi.refaktor.change.changers}{SearchBasedExtractAndMoveMethodChanger}
2780 is a changer whose purpose is to automatically analyze a method, and execute the
2781 \ExtractAndMoveMethod refactoring on it if it is a suitable candidate for the
2784 First, the \typewithref{no.uio.ifi.refaktor.analyze.analyzers}{SearchBasedExtractAndMoveMethodAnalyzer} is used
2785 to analyze the method. If the method is found to be a candidate, the result from
2786 the analysis is fed to the \type{ExtractAndMoveMethodExecutor}, whose job is to
2787 execute the refactoring \see{extractAndMoveMethodExecutor}.
2789 \subsubsection{The SearchBasedExtractAndMoveMethodAnalyzer}
2790 This analyzer is responsible for analyzing all the possible text selections of a
2791 method and then to choose the best result out of the analysis results that are,
2792 by the analyzer, considered to be the potential candidates for the
2793 \ExtractAndMoveMethod refactoring.
2795 Before the analyzer is able to work with the text selections of a method, it
2796 needs to generate them. To do this, it parses the method to obtain a
2797 \type{MethodDeclaration} for it \see{astEclipse}. Then there is a statement
2798 lists creator that creates statements lists of the different groups of
2799 statements in the body of the method declaration. A text selections generator
2800 generates text selections of all the statement lists for the analyzer to work
2803 \paragraph{The statement lists creator}
2804 is responsible for generating lists of statements for all the possible nesting
2805 levels of statements in the method. The statement lists creator is implemented
2806 as an AST visitor \see{astVisitor}. It generates lists of statements by visiting
2807 all the blocks in the method declaration and stores their statements in a
2808 collection of statement lists. In addition, it visits all of the other
2809 statements that can have a statement as a child, such as the different control
2810 structures and the labeled statement.
2812 The switch statement is the only kind of statement that is not straight forward
2813 to obtain the child statements from. It stores all of its children in a flat
2814 list. Its switch case statements are included in this list. This means that
2815 there are potential statement lists between all of these case statements. The
2816 list of statements from a switch statement is therefore traversed, and the
2817 statements between the case statements are grouped as separate lists.
2819 The highlighted part of \myref{lst:grandExample} shows an example of how the
2820 statement lists creator would separate a method body into lists of statements.
2822 \paragraph{The text selections generator} generates text selections for each
2823 list of statements from the statement lists creator. The generator generates a
2824 text selection for every sub-sequence of statements in a list. For a sequence of
2825 statements, the first statement and the last statement span out a text
2828 In practice, the text selections are calculated by only one traversal of the
2829 statement list. There is a set of generated text selections. For each statement,
2830 there is created a temporary set of selections, in addition to a text selection
2831 based on the offset and length of the statement. This text selection is added to
2832 the temporary set. Then the new selection is added with every selection from the
2833 set of generated text selections. These new selections are added to the
2834 temporary set. Then the temporary set of selections is added to the set of
2835 generated text selections. The result of adding two text selections is a new
2836 text selection spanned out by the two addends.
2840 \def\charwidth{5.7pt}
2841 \def\indent{4*\charwidth}
2842 \def\lineheight{\baselineskip}
2843 \def\mintedtop{\lineheight}
2845 \begin{tikzpicture}[overlay, yscale=-1]
2846 \tikzstyle{overlaybox}=[fill=lightgray,opacity=0.2]
2848 \draw[overlaybox] (2*\charwidth,\mintedtop) rectangle
2849 +(18*\charwidth,\lineheight);
2851 \draw[overlaybox] (2*\charwidth,\mintedtop+\lineheight) rectangle
2852 +(18*\charwidth,\lineheight);
2854 \draw[overlaybox] (2*\charwidth,\mintedtop+3*\lineheight) rectangle
2855 +(18*\charwidth,\lineheight);
2857 \draw[overlaybox] (\indent-3*\charwidth,\mintedtop) rectangle
2858 +(20*\charwidth,2*\lineheight);
2860 \draw[overlaybox] (3*\charwidth,\mintedtop+\lineheight) rectangle
2861 +(16*\charwidth,3*\lineheight);
2863 \draw[overlaybox] (\indent,\mintedtop) rectangle
2864 +(14*\charwidth,4*\lineheight);
2866 \begin{minted}{java}
2872 \caption{Example of how the text selections generator would generate text
2873 selections based on a lists of statements. Each highlighted rectangle
2874 represents a text selection.}
2875 \label{lst:textSelectionsExample}
2877 \todoin{fix \myref{lst:textSelectionsExample}? Text only? All
2878 sub-sequences\ldots}
2881 \paragraph{Finding the candidate} for the refactoring is done by analyzing all
2882 the generated text selection with the \type{ExtractAndMoveMethodAnalyzer}
2883 \see{extractAndMoveMethodAnalyzer}. If the analyzer generates a useful result,
2884 an \type{ExtractAndMoveMethodCandidate} is created from it, which is kept in a
2885 list of potential candidates. If no candidates are found, the
2886 \type{NoTargetFoundException} is thrown.
2888 Since only one of the candidates can be chosen, the analyzer must sort out which
2889 candidate to choose. The sorting is done by the static \method{sort} method of
2890 \type{Collections}. The comparison in this sorting is done by an
2891 \type{ExtractAndMoveMethodCandidateComparator}.
2892 \todoin{Write about the
2893 ExtractAndMoveMethodCandidateComparator/FavorNoUnfixesCandidateComparator}
2896 \subsection{The Prefix Class}
2897 This class exists mainly for holding data about a prefix, such as the expression
2898 that the prefix represents and the occurrence count of the prefix within a
2899 selection. In addition to this, it has some functionality such as calculating
2900 its sub-prefixes and intersecting it with another prefix. The definition of the
2901 intersection between two prefixes is a prefix representing the longest common
2902 expression between the two.
2904 \subsection{The PrefixSet Class}
2905 A prefix set holds elements of type \type{Prefix}. It is implemented with the
2906 help of a \typewithref{java.util}{HashMap} and contains some typical set
2907 operations, but it does not implement the \typewithref{java.util}{Set}
2908 interface, since the prefix set does not need all of the functionality a
2909 \type{Set} requires to be implemented. In addition It needs some other
2910 functionality not found in the \type{Set} interface. So due to the relatively
2911 limited use of prefix sets, and that it almost always needs to be referenced as
2912 such, and not a \type{Set<Prefix>}, it remains as an ad hoc solution to a
2915 There are two ways adding prefixes to a \type{PrefixSet}. The first is through
2916 its \method{add} method. This works like one would expect from a set. It adds
2917 the prefix to the set if it does not already contain the prefix. The other way
2918 is to \emph{register} the prefix with the set. When registering a prefix, if the
2919 set does not contain the prefix, it is just added. If the set contains the
2920 prefix, its count gets incremented. This is how the occurrence count is handled.
2922 The prefix set also computes the set of prefixes that is not enclosing any
2923 prefixes of another set. This is kind of a set difference operation only for
2926 \subsection{Hacking the Refactoring Undo
2927 History}\label{hacking_undo_history}
2928 \todoin{Where to put this section?}
2930 As an attempt to make multiple subsequent changes to the workspace appear as a
2931 single action (i.e. make the undo changes appear as such), I tried to alter
2932 the undo changes\typeref{org.eclipse.ltk.core.refactoring.Change} in the history
2933 of the refactorings.
2935 My first impulse was to remove the, in this case, last two undo changes from the
2936 undo manager\typeref{org.eclipse.ltk.core.refactoring.IUndoManager} for the
2937 \name{Eclipse} refactorings, and then add them to a composite
2938 change\typeref{org.eclipse.ltk.core.refactoring.CompositeChange} that could be
2939 added back to the manager. The interface of the undo manager does not offer a
2940 way to remove/pop the last added undo change, so a possible solution could be to
2941 decorate\citing{designPatterns} the undo manager, to intercept and collect the
2942 undo changes before delegating to the \method{addUndo}
2943 method\methodref{org.eclipse.ltk.core.refactoring.IUndoManager}{addUndo} of the
2944 manager. Instead of giving it the intended undo change, a null change could be
2945 given to prevent it from making any changes if run. Then one could let the
2946 collected undo changes form a composite change to be added to the manager.
2948 There is a technical challenge with this approach, and it relates to the undo
2949 manager, and the concrete implementation
2950 \typewithref{org.eclipse.ltk.internal.core.refactoring}{UndoManager2}. This
2951 implementation is designed in a way that it is not possible to just add an undo
2952 change, you have to do it in the context of an active
2953 operation\typeref{org.eclipse.core.commands.operations.TriggeredOperations}.
2954 One could imagine that it might be possible to trick the undo manager into
2955 believing that you are doing a real change, by executing a refactoring that is
2956 returning a kind of null change that is returning our composite change of undo
2957 refactorings when it is performed. But this is not the way to go.
2959 Apart from the technical problems with this solution, there is a functional
2960 problem: If it all had worked out as planned, this would leave the undo history
2961 in a dirty state, with multiple empty undo operations corresponding to each of
2962 the sequentially executed refactoring operations, followed by a composite undo
2963 change corresponding to an empty change of the workspace for rounding of our
2964 composite refactoring. The solution to this particular problem could be to
2965 intercept the registration of the intermediate changes in the undo manager, and
2966 only register the last empty change.
2968 Unfortunately, not everything works as desired with this solution. The grouping
2969 of the undo changes into the composite change does not make the undo operation
2970 appear as an atomic operation. The undo operation is still split up into
2971 separate undo actions, corresponding to the changes done by their originating
2972 refactorings. And in addition, the undo actions have to be performed separate in
2973 all the editors involved. This makes it no solution at all, but a step toward
2976 There might be a solution to this problem, but it remains to be found. The
2977 design of the refactoring undo management is partly to be blamed for this, as
2978 it is too complex to be easily manipulated.
2981 \chapter{Analyzing Source Code in Eclipse}
2983 \section{The Java model}\label{javaModel}
2984 The Java model of \name{Eclipse} is its internal representation of a Java project. It
2985 is light-weight, and has only limited possibilities for manipulating source
2986 code. It is typically used as a basis for the Package Explorer in \name{Eclipse}.
2988 The elements of the Java model are only handles to the underlying elements. This
2989 means that the underlying element of a handle does not need to actually exist.
2990 Hence the user of a handle must always check that it exist by calling the
2991 \method{exists} method of the handle.
2993 The handles with descriptions are listed in \myref{tab:javaModel}, while the
2994 hierarchy of the Java Model is shown in \myref{fig:javaModel}.
2997 \caption{The elements of the Java Model\citing{vogelEclipseJDT2012}.}
2998 \label{tab:javaModel}
3000 % sum must equal number of columns (3)
3001 \begin{tabularx}{\textwidth}{@{} L{0.7} L{1.1} L{1.2} @{}}
3003 \textbf{Project Element} & \textbf{Java Model element} &
3004 \textbf{Description} \\
3006 Java project & \type{IJavaProject} & The Java project which contains all other objects. \\
3008 Source folder /\linebreak[2] binary folder /\linebreak[3] external library &
3009 \type{IPackageFragmentRoot} & Hold source or binary files, can be a folder
3010 or a library (zip / jar file). \\
3012 Each package & \type{IPackageFragment} & Each package is below the
3013 \type{IPackageFragmentRoot}, sub-packages are not leaves of the package,
3014 they are listed directed under \type{IPackageFragmentRoot}. \\
3016 Java Source file & \type{ICompilationUnit} & The Source file is always below
3017 the package node. \\
3019 Types / Fields /\linebreak[3] Methods & \type{IType} / \type{IField}
3020 /\linebreak[3] \type{IMethod} & Types, fields and methods. \\
3028 \begin{tikzpicture}[%
3029 grow via three points={one child at (0,-0.7) and
3030 two children at (0,-0.7) and (0,-1.4)},
3031 edge from parent path={(\tikzparentnode.south west)+(0.5,0) |-
3032 (\tikzchildnode.west)}]
3033 \tikzstyle{every node}=[draw=black,thick,anchor=west]
3034 \tikzstyle{selected}=[draw=red,fill=red!30]
3035 \tikzstyle{optional}=[dashed,fill=gray!50]
3036 \node {\type{IJavaProject}}
3037 child { node {\type{IPackageFragmentRoot}}
3038 child { node {\type{IPackageFragment}}
3039 child { node {\type{ICompilationUnit}}
3040 child { node {\type{IType}}
3041 child { node {\type{\{ IType \}*}}
3042 child { node {\type{\ldots}}}
3045 child { node {\type{\{ IField \}*}}}
3046 child { node {\type{IMethod}}
3047 child { node {\type{\{ IType \}*}}
3048 child { node {\type{\ldots}}}
3053 child { node {\type{\{ IMethod \}*}}}
3062 child { node {\type{\{ IType \}*}}}
3073 child { node {\type{\{ ICompilationUnit \}*}}}
3086 child { node {\type{\{ IPackageFragment \}*}}}
3101 child { node {\type{\{ IPackageFragmentRoot \}*}}}
3104 \caption{The Java model of \name{Eclipse}. ``\type{\{ SomeElement \}*}'' means
3105 ``\type{SomeElement} zero or more times``. For recursive structures,
3106 ``\type{\ldots}'' is used.}
3107 \label{fig:javaModel}
3110 \section{The Abstract Syntax Tree}
3111 \name{Eclipse} is following the common paradigm of using an abstract syntax tree for
3112 source code analysis and manipulation.
3114 When parsing program source code into something that can be used as a foundation
3115 for analysis, the start of the process follows the same steps as in a compiler.
3116 This is all natural, because the way a compiler analyzes code is no different
3117 from how source manipulation programs would do it, except for some properties of
3118 code that is analyzed in the parser, and that they may be differing in what
3119 kinds of properties they analyze. Thus the process of translation source code
3120 into a structure that is suitable for analyzing, can be seen as a kind of
3121 interrupted compilation process \see{fig:interruptedCompilationProcess}.
3126 base/.style={anchor=north, align=center, rectangle, minimum height=1.4cm},
3127 basewithshadow/.style={base, drop shadow, fill=white},
3128 outlined/.style={basewithshadow, draw, rounded corners, minimum
3130 primary/.style={outlined, font=\bfseries},
3131 dashedbox/.style={outlined, dashed},
3132 arrowpath/.style={black, align=center, font=\small},
3133 processarrow/.style={arrowpath, ->, >=angle 90, shorten >=1pt},
3135 \begin{tikzpicture}[node distance=1.3cm and 3cm, scale=1, every
3136 node/.style={transform shape}]
3137 \node[base](AuxNode1){\small source code};
3138 \node[primary, right=of AuxNode1, xshift=-2.5cm](Scanner){Scanner};
3139 \node[primary, right=of Scanner, xshift=0.5cm](Parser){Parser};
3140 \node[dashedbox, below=of Parser](SemanticAnalyzer){Semantic\\Analyzer};
3141 \node[dashedbox, left=of SemanticAnalyzer](SourceCodeOptimizer){Source
3143 \node[dashedbox, below=of SourceCodeOptimizer
3144 ](CodeGenerator){Code\\Generator};
3145 \node[dashedbox, right=of CodeGenerator](TargetCodeOptimizer){Target
3147 \node[base, right=of TargetCodeOptimizer](AuxNode2){};
3149 \draw[processarrow](AuxNode1) -- (Scanner);
3151 \path[arrowpath] (Scanner) -- node [sloped](tokens){tokens}(Parser);
3152 \draw[processarrow](Scanner) -- (tokens) -- (Parser);
3154 \path[arrowpath] (Parser) -- node (syntax){syntax
3155 tree}(SemanticAnalyzer);
3156 \draw[processarrow](Parser) -- (syntax) -- (SemanticAnalyzer);
3158 \path[arrowpath] (SemanticAnalyzer) -- node
3159 [sloped](annotated){annotated\\tree}(SourceCodeOptimizer);
3160 \draw[processarrow, dashed](SemanticAnalyzer) -- (annotated) --
3161 (SourceCodeOptimizer);
3163 \path[arrowpath] (SourceCodeOptimizer) -- node
3164 (intermediate){intermediate code}(CodeGenerator);
3165 \draw[processarrow, dashed](SourceCodeOptimizer) -- (intermediate) --
3168 \path[arrowpath] (CodeGenerator) -- node [sloped](target1){target
3169 code}(TargetCodeOptimizer);
3170 \draw[processarrow, dashed](CodeGenerator) -- (target1) --
3171 (TargetCodeOptimizer);
3173 \path[arrowpath](TargetCodeOptimizer) -- node [sloped](target2){target
3175 \draw[processarrow, dashed](TargetCodeOptimizer) -- (target2) (AuxNode2);
3177 \caption{Interrupted compilation process. {\footnotesize (Full compilation
3178 process borrowed from \emph{Compiler construction: principles and practice}
3179 by Kenneth C. Louden\citing{louden1997}.)}}
3180 \label{fig:interruptedCompilationProcess}
3183 The process starts with a \emph{scanner}, or lexer. The job of the scanner is to
3184 read the source code and divide it into tokens for the parser. Therefore, it is
3185 also sometimes called a tokenizer. A token is a logical unit, defined in the
3186 language specification, consisting of one or more consecutive characters. In
3187 the Java language the tokens can for instance be the \var{this} keyword, a curly
3188 bracket \var{\{} or a \var{nameToken}. It is recognized by the scanner on the
3189 basis of something equivalent of a regular expression. This part of the process
3190 is often implemented with the use of a finite automata. In fact, it is common to
3191 specify the tokens in regular expressions, which in turn are translated into a
3192 finite automata lexer. This process can be automated.
3194 The program component used to translate a stream of tokens into something
3195 meaningful, is called a parser. A parser is fed tokens from the scanner and
3196 performs an analysis of the structure of a program. It verifies that the syntax
3197 is correct according to the grammar rules of a language, that are usually
3198 specified in a context-free grammar, and often in a variant of the
3200 Form}\footnote{\url{https://en.wikipedia.org/wiki/Backus-Naur\_Form}}. The
3201 result coming from the parser is in the form of an \emph{Abstract Syntax Tree},
3202 AST for short. It is called \emph{abstract}, because the structure does not
3203 contain all of the tokens produced by the scanner. It only contains logical
3204 constructs, and because it forms a tree, all kinds of parentheses and brackets
3205 are implicit in the structure. It is this AST that is used when performing the
3206 semantic analysis of the code.
3208 As an example we can think of the expression \code{(5 + 7) * 2}. The root of
3209 this tree would in \name{Eclipse} be an \type{InfixExpression} with the operator
3210 \var{TIMES}, and a left operand, which is also an \type{InfixExpression} with
3211 the operator \var{PLUS}. The left operand \type{InfixExpression}, has in turn a
3212 left operand of type \type{NumberLiteral} with the value \var{``5''} and a right
3213 operand \type{NumberLiteral} with the value \var{``7''}. The root will have a
3214 right operand of type \type{NumberLiteral} and value \var{``2''}. The AST for
3215 this expression is illustrated in \myref{fig:astInfixExpression}.
3217 Contrary to the Java Model, an abstract syntax tree is a heavy-weight
3218 representation of source code. It contains information about properties like
3219 type bindings for variables and variable bindings for names.
3224 \begin{tikzpicture}[scale=0.8]
3225 \tikzset{level distance=40pt}
3226 \tikzset{sibling distance=5pt}
3227 \tikzstyle{thescale}=[scale=0.8]
3228 \tikzset{every tree node/.style={align=center}}
3229 \tikzset{edge from parent/.append style={thick}}
3230 \tikzstyle{inode}=[rectangle,rounded corners,draw,fill=lightgray,drop
3231 shadow,align=center]
3232 \tikzset{every internal node/.style={inode}}
3233 \tikzset{every leaf node/.style={draw=none,fill=none}}
3235 \Tree [.\type{InfixExpression} [.\type{InfixExpression}
3236 [.\type{NumberLiteral} \var{``5''} ] [.\type{Operator} \var{PLUS} ]
3237 [.\type{NumberLiteral} \var{``7''} ] ]
3238 [.\type{Operator} \var{TIMES} ]
3239 [.\type{NumberLiteral} \var{``2''} ]
3242 \caption{The abstract syntax tree for the expression \code{(5 + 7) * 2}.}
3243 \label{fig:astInfixExpression}
3246 \subsection{The AST in Eclipse}\label{astEclipse}
3247 In \name{Eclipse}, every node in the AST is a child of the abstract superclass
3248 \typewithref{org.eclipse.jdt.core.dom}{ASTNode}. Every \type{ASTNode}, among a
3249 lot of other things, provides information about its position and length in the
3250 source code, as well as a reference to its parent and to the root of the tree.
3252 The root of the AST is always of type \type{CompilationUnit}. It is not the same
3253 as an instance of an \type{ICompilationUnit}, which is the compilation unit
3254 handle of the Java model. The children of a \type{CompilationUnit} is an
3255 optional \type{PackageDeclaration}, zero or more nodes of type
3256 \type{ImportDecaration} and all its top-level type declarations that has node
3257 types \type{AbstractTypeDeclaration}.
3259 An \type{AbstractType\-Declaration} can be one of the types
3260 \type{AnnotationType\-Declaration}, \type{Enum\-Declaration} or
3261 \type{Type\-Declaration}. The children of an \type{AbstractType\-Declaration}
3262 must be a subtype of a \type{BodyDeclaration}. These subtypes are:
3263 \type{AnnotationTypeMember\-Declaration}, \type{EnumConstant\-Declaration},
3264 \type{Field\-Declaration}, \type{Initializer} and \type{Method\-Declaration}.
3266 Of the body declarations, the \type{Method\-Declaration} is the most interesting
3267 one. Its children include lists of modifiers, type parameters, parameters and
3268 exceptions. It has a return type node and a body node. The body, if present, is
3269 of type \type{Block}. A \type{Block} is itself a \type{Statement}, and its
3270 children is a list of \type{Statement} nodes.
3272 There are too many types of the abstract type \type{Statement} to list up, but
3273 there exists a subtype of \type{Statement} for every statement type of Java, as
3274 one would expect. This also applies to the abstract type \type{Expression}.
3275 However, the expression \type{Name} is a little special, since it is both used
3276 as an operand in compound expressions, as well as for names in type declarations
3279 There is an overview of some of the structure of an \name{Eclipse} AST in
3280 \myref{fig:astEclipse}.
3284 \begin{tikzpicture}[scale=0.8]
3285 \tikzset{level distance=50pt}
3286 \tikzset{sibling distance=5pt}
3287 \tikzstyle{thescale}=[scale=0.8]
3288 \tikzset{every tree node/.style={align=center}}
3289 \tikzset{edge from parent/.append style={thick}}
3290 \tikzstyle{inode}=[rectangle,rounded corners,draw,fill=lightgray,drop
3291 shadow,align=center]
3292 \tikzset{every internal node/.style={inode}}
3293 \tikzset{every leaf node/.style={draw=none,fill=none}}
3295 \Tree [.\type{CompilationUnit} [.\type{[ PackageDeclaration ]} [.\type{Name} ]
3296 [.\type{\{ Annotation \}*} ] ]
3297 [.\type{\{ ImportDeclaration \}*} [.\type{Name} ] ]
3298 [.\type{\{ AbstractTypeDeclaration \}+} [.\node(site){\type{\{
3299 BodyDeclaration \}*}}; ] [.\type{SimpleName} ] ]
3301 \begin{scope}[shift={(0.5,-6)}]
3302 \node[inode,thescale](root){\type{MethodDeclaration}};
3303 \node[inode,thescale](modifiers) at (4.5,-5){\type{\{ IExtendedModifier \}*}
3304 \\ {\footnotesize (Of type \type{Modifier} or \type{Annotation})}};
3305 \node[inode,thescale](typeParameters) at (-6,-3.5){\type{\{ TypeParameter
3307 \node[inode,thescale](parameters) at (-5,-5){\type{\{
3308 SingleVariableDeclaration \}*} \\ {\footnotesize (Parameters)}};
3309 \node[inode,thescale](exceptions) at (5,-3){\type{\{ Name \}*} \\
3310 {\footnotesize (Exceptions)}};
3311 \node[inode,thescale](return) at (-6.5,-2){\type{Type} \\ {\footnotesize
3313 \begin{scope}[shift={(0,-5)}]
3314 \Tree [.\node(body){\type{[ Block ]} \\ {\footnotesize (Body)}};
3315 [.\type{\{ Statement \}*} [.\type{\{ Expression \}*} ]
3316 [.\type{\{ Statement \}*} [.\type{\ldots} ]]
3321 \draw[->,>=triangle 90,shorten >=1pt](root.east)..controls +(east:2) and
3322 +(south:1)..(site.south);
3324 \draw (root.south) -- (modifiers);
3325 \draw (root.south) -- (typeParameters);
3326 \draw (root.south) -- ($ (parameters.north) + (2,0) $);
3327 \draw (root.south) -- (exceptions);
3328 \draw (root.south) -- (return);
3329 \draw (root.south) -- (body);
3332 \caption{The format of the abstract syntax tree in \name{Eclipse}.}
3333 \label{fig:astEclipse}
3335 \todoin{Add more to the AST format tree? \myref{fig:astEclipse}}
3337 \section{The ASTVisitor}\label{astVisitor}
3338 So far, the only thing that has been addressed is how the data that is going to
3339 be the basis for our analysis is structured. Another aspect of it is how we are
3340 going to traverse the AST to gather the information we need, so we can conclude
3341 about the properties we are analyzing. It is of course possible to start at the
3342 top of the tree, and manually search through its nodes for the ones we are
3343 looking for, but that is a bit inconvenient. To be able to efficiently utilize
3344 such an approach, we would need to make our own framework for traversing the
3345 tree and visiting only the types of nodes we are after. Luckily, this
3346 functionality is already provided in \name{Eclipse}, by its
3347 \typewithref{org.eclipse.jdt.core.dom}{ASTVisitor}.
3349 The \name{Eclipse} AST, together with its \type{ASTVisitor}, follows the
3350 \pattern{Visitor} pattern\citing{designPatterns}. The intent of this design
3351 pattern is to facilitate extending the functionality of classes without touching
3352 the classes themselves.
3354 Let us say that there is a class hierarchy of elements. These elements all have
3355 a method \method{accept(Visitor visitor)}. In its simplest form, the
3356 \method{accept} method just calls the \method{visit} method of the visitor with
3357 itself as an argument, like this: \code{visitor.visit(this)}. For the visitors
3358 to be able to extend the functionality of all the classes in the elements
3359 hierarchy, each \type{Visitor} must have one visit method for each concrete
3360 class in the hierarchy. Say the hierarchy consists of the concrete classes
3361 \type{ConcreteElementA} and \type{ConcreteElementB}. Then each visitor must have
3362 the (possibly empty) methods \method{visit(ConcreteElementA element)} and
3363 \method{visit(ConcreteElementB element)}. This scenario is depicted in
3364 \myref{fig:visitorPattern}.
3368 \tikzstyle{abstract}=[rectangle, draw=black, fill=white, drop shadow, text
3369 centered, anchor=north, text=black, text width=6cm, every one node
3370 part/.style={align=center, font=\bfseries\itshape}]
3371 \tikzstyle{concrete}=[rectangle, draw=black, fill=white, drop shadow, text
3372 centered, anchor=north, text=black, text width=6cm]
3373 \tikzstyle{inheritarrow}=[->, >=open triangle 90, thick]
3374 \tikzstyle{commentarrow}=[->, >=angle 90, dashed]
3375 \tikzstyle{line}=[-, thick]
3376 \tikzset{every one node part/.style={align=center, font=\bfseries}}
3377 \tikzset{every second node part/.style={align=center, font=\ttfamily}}
3379 \begin{tikzpicture}[node distance=1cm, scale=0.8, every node/.style={transform
3381 \node (Element) [abstract, rectangle split, rectangle split parts=2]
3383 \nodepart{one}{Element}
3384 \nodepart{second}{+accept(visitor: Visitor)}
3386 \node (AuxNode01) [text width=0, minimum height=2cm, below=of Element] {};
3387 \node (ConcreteElementA) [concrete, rectangle split, rectangle split
3388 parts=2, left=of AuxNode01]
3390 \nodepart{one}{ConcreteElementA}
3391 \nodepart{second}{+accept(visitor: Visitor)}
3393 \node (ConcreteElementB) [concrete, rectangle split, rectangle split
3394 parts=2, right=of AuxNode01]
3396 \nodepart{one}{ConcreteElementB}
3397 \nodepart{second}{+accept(visitor: Visitor)}
3400 \node[comment, below=of ConcreteElementA] (CommentA) {visitor.visit(this)};
3402 \node[comment, below=of ConcreteElementB] (CommentB) {visitor.visit(this)};
3404 \node (AuxNodeX) [text width=0, minimum height=1cm, below=of AuxNode01] {};
3406 \node (Visitor) [abstract, rectangle split, rectangle split parts=2,
3409 \nodepart{one}{Visitor}
3410 \nodepart{second}{+visit(ConcreteElementA)\\+visit(ConcreteElementB)}
3412 \node (AuxNode02) [text width=0, minimum height=2cm, below=of Visitor] {};
3413 \node (ConcreteVisitor1) [concrete, rectangle split, rectangle split
3414 parts=2, left=of AuxNode02]
3416 \nodepart{one}{ConcreteVisitor1}
3417 \nodepart{second}{+visit(ConcreteElementA)\\+visit(ConcreteElementB)}
3419 \node (ConcreteVisitor2) [concrete, rectangle split, rectangle split
3420 parts=2, right=of AuxNode02]
3422 \nodepart{one}{ConcreteVisitor2}
3423 \nodepart{second}{+visit(ConcreteElementA)\\+visit(ConcreteElementB)}
3427 \draw[inheritarrow] (ConcreteElementA.north) -- ++(0,0.7) -|
3429 \draw[line] (ConcreteElementA.north) -- ++(0,0.7) -|
3430 (ConcreteElementB.north);
3432 \draw[inheritarrow] (ConcreteVisitor1.north) -- ++(0,0.7) -|
3434 \draw[line] (ConcreteVisitor1.north) -- ++(0,0.7) -|
3435 (ConcreteVisitor2.north);
3437 \draw[commentarrow] (CommentA.north) -- (ConcreteElementA.south);
3438 \draw[commentarrow] (CommentB.north) -- (ConcreteElementB.south);
3442 \caption{The Visitor Pattern.}
3443 \label{fig:visitorPattern}
3446 The use of the visitor pattern can be appropriate when the hierarchy of elements
3447 is mostly stable, but the family of operations over its elements is constantly
3448 growing. This is clearly the case for the \name{Eclipse} AST, since the
3449 hierarchy for the type \type{ASTNode} is very stable, but the functionality of
3450 its elements is extended every time someone need to operate on the AST. Another
3451 aspect of the \name{Eclipse} implementation is that it is a public API, and the
3452 visitor pattern is an easy way to provide access to the nodes in the tree.
3454 The version of the visitor pattern implemented for the AST nodes in \name{Eclipse} also
3455 provides an elegant way to traverse the tree. It does so by following the
3456 convention that every node in the tree first let the visitor visit itself,
3457 before it also makes all its children accept the visitor. The children are only
3458 visited if the visit method of their parent returns \var{true}. This pattern
3459 then makes for a prefix traversal of the AST. If postfix traversal is desired,
3460 the visitors also have \method{endVisit} methods for each node type, which is
3461 called after the \method{visit} method for a node. In addition to these visit
3462 methods, there are also the methods \method{preVisit(ASTNode)},
3463 \method{postVisit(ASTNode)} and \method{preVisit2(ASTNode)}. The
3464 \method{preVisit} method is called before the type-specific \method{visit}
3465 method. The \method{postVisit} method is called after the type-specific
3466 \method{endVisit}. The type specific \method{visit} is only called if
3467 \method{preVisit2} returns \var{true}. Overriding the \method{preVisit2} is also
3468 altering the behavior of \method{preVisit}, since the default implementation is
3469 responsible for calling it.
3471 An example of a trivial \type{ASTVisitor} is shown in
3472 \myref{lst:astVisitorExample}.
3475 \begin{minted}{java}
3476 public class CollectNamesVisitor extends ASTVisitor {
3477 Collection<Name> names = new LinkedList<Name>();
3480 public boolean visit(QualifiedName node) {
3486 public boolean visit(SimpleName node) {
3492 \caption{An \type{ASTVisitor} that visits all the names in a subtree and adds
3493 them to a collection, except those names that are children of any
3494 \type{QualifiedName}.}
3495 \label{lst:astVisitorExample}
3498 \section{Property collectors}\label{propertyCollectors}
3499 The prefixes and unfixes are found by property
3500 collectors\typeref{no.uio.ifi.refaktor.extractors.collectors.PropertyCollector}.
3501 A property collector is of the \type{ASTVisitor} type, and thus visits nodes of
3502 type \type{ASTNode} of the abstract syntax tree \see{astVisitor}.
3504 \subsection{The PrefixesCollector}
3505 The \typewithref{no.uio.ifi.refaktor.extractors.collectors}{PrefixesCollector}
3506 finds prefixes that makes up the basis for calculating move targets for the
3507 \refa{Extract and Move Method} refactoring. It visits expression
3508 statements\typeref{org.eclipse.jdt.core.dom.ExpressionStatement} and creates
3509 prefixes from its expressions in the case of method invocations. The prefixes
3510 found are registered with a prefix set, together with all its sub-prefixes.
3512 \subsection{The UnfixesCollector}\label{unfixes}
3513 The \typewithref{no.uio.ifi.refaktor.extractors.collectors}{UnfixesCollector}
3514 finds unfixes within a selection.
3515 \todoin{Give more technical detail?}
3519 \subsection{The ContainsReturnStatementCollector}
3520 \todoin{Remove section?}
3522 \typewithref{no.uio.ifi.refaktor.analyze.collectors}{ContainsReturnStatementCollector}
3523 is a very simple property collector. It only visits the return statements within
3524 a selection, and can report whether it encountered a return statement or not.
3526 \subsection{The LastStatementCollector}
3527 The \typewithref{no.uio.ifi.refaktor.analyze.collectors}{LastStatementCollector}
3528 collects the last statement of a selection. It does so by only visiting the top
3529 level statements of the selection, and compares the textual end offset of each
3530 encountered statement with the end offset of the previous statement found.
3532 \section{Checkers}\label{checkers}
3533 The checkers are a range of classes that checks that text selections comply
3534 with certain criteria. All checkers operates under the assumption that the code
3535 they check is free from compilation errors. If a
3536 \typewithref{no.uio.ifi.refaktor.analyze.analyzers}{Checker} fails, it throws a
3537 \type{CheckerException}. The checkers are managed by the
3538 \type{LegalStatementsChecker}, which does not, in fact, implement the
3539 \type{Checker} interface. It does, however, run all the checkers registered with
3540 it, and reports that all statements are considered legal if no
3541 \type{CheckerException} is thrown. Many of the checkers either extends the
3542 \type{PropertyCollector} or utilizes one or more property collectors to verify
3543 some criteria. The checkers registered with the \type{LegalStatementsChecker}
3544 are described next. They are run in the order presented below.
3546 \subsection{The CallToProtectedOrPackagePrivateMethodChecker}
3547 This checker is used to check that at selection does not contain a call to a
3548 method that is protected or package-private. Such a method either has the access
3549 modifier \code{protected} or it has no access modifier.
3551 The workings of the \type{CallToProtectedOrPackagePrivateMethod\-Checker} is
3552 very simple. It looks for calls to methods that are either protected or
3553 package-private within the selection, and throws an
3554 \type{IllegalExpressionFoundException} if one is found.
3556 \subsection{The DoubleClassInstanceCreationChecker}
3557 The \type{DoubleClassInstanceCreationChecker} checks that there are no double
3558 class instance creations where the inner constructor call takes an argument that
3559 is built up using field references.
3561 The checker visits all nodes of type \type{ClassInstanceCreation} within a
3562 selection. For all of these nodes, if its parent also is a class instance
3563 creation, it accepts a visitor that throws a
3564 \type{IllegalExpressionFoundException} if it encounters a name that is a field
3567 \subsection{The InstantiationOfNonStaticInnerClassChecker}
3568 The \type{InstantiationOfNonStaticInnerClassChecker} checks that selections
3569 do not contain instantiations of non-static inner classes. The
3570 \type{MoveInstanceMethodProcessor} in \name{Eclipse} does not handle such
3571 instantiations gracefully when moving a method. This problem is also related to
3572 bug\ldots \todoin{File Eclipse bug report}
3574 \subsection{The EnclosingInstanceReferenceChecker}
3575 The purpose of this checker is to verify that the names in a text selection are
3576 not referencing any enclosing instances. In theory, the underlying problem could
3577 be solved in some situations, but our dependency on the
3578 \type{MoveInstanceMethodProcessor} prevents this.
3581 \typewithref{no.uio.ifi.refaktor.analyze.analyzers}{EnclosingInstanceReferenceChecker}
3582 is a modified version of the
3583 \typewithref{org.eclipse.jdt.internal.corext.refactoring.structure.MoveInstanceMethod\-Processor}{EnclosingInstanceReferenceFinder}
3584 from the \type{MoveInstanceMethodProcessor}. Wherever the
3585 \type{EnclosingInstanceReferenceFinder} would create a fatal error status, the
3586 checker will throw a \type{CheckerException}.
3588 The checker works by first finding all of the enclosing types of a selection.
3589 Thereafter, it visits all the simple names of the selection to check that they
3590 are not references to variables or methods declared in any of the enclosing
3591 types. In addition, the checker visits \var{this}-expressions to verify that no
3592 such expressions are qualified with any name.
3594 \subsection{The ReturnStatementsChecker}\label{returnStatementsChecker}
3595 The checker for return statements is meant to verify that a text selection is
3596 consistent regarding return statements.
3598 If the selection is free from return statements, then the checker validates. So
3599 this is the first thing the checker investigates.
3601 If the checker proceeds any further, it is because the selection contains one
3602 or more return statements. The next test is therefore to check if the last
3603 statement of the selection ends in either a return or a throw statement. The
3604 responsibility for checking that the last statement of the selection eventually
3605 ends in a return or throw statement, is put on the
3606 \type{LastStatementOfSelectionEndsInReturnOrThrowChecker}. For every node
3607 visited, if the node is a statement, it does a test to see if the statement is a
3608 return, a throw or if it is an implicit return statement. If this is the case,
3609 no further checking is done. This checking is done in the \code{preVisit2}
3610 method \see{astVisitor}. If the node is not of a type that is being handled by
3611 its type-specific visit method, the checker performs a simple test. If the node
3612 being visited is not the last statement of its parent that is also enclosed by
3613 the selection, an \type{IllegalStatementFoundException} is thrown. This ensures
3614 that all statements are taken care of, one way or the other. It also ensures
3615 that the checker is conservative in the way it checks for legality of the
3618 To examine if a statement is an implicit return statement, the checker first
3619 finds the last statement declared in its enclosing method. If this statement is
3620 the same as the one under investigation, it is considered an implicit return
3621 statement. If the statements are not the same, the checker does a search to see
3622 if the statement examined is also the last statement of the method that can be
3623 reached. This includes the last statement of a block statement, a labeled
3624 statement, a synchronized statement or a try statement, that in turn is the last
3625 statement enclosed by one of the statement types listed. This search goes
3626 through all the parents of a statement until a statement is found that is not
3627 one of the mentioned acceptable parent statements. If the search ends in a
3628 method declaration, then the statement is considered to be the last reachable
3629 statement of the method, and thus it is an implicit return statement.
3631 There are two kinds of statements that are handled explicitly: If-statements and
3632 try-statements. Block, labeled and do-statements are handled by fall-through to
3635 If-statements are handled explicitly by overriding their type-specific visit
3636 method. If the then-part does not contain any return or throw statements an
3637 \type{IllegalStatementFoundException} is thrown. If it does contain a return or
3638 throw, its else-part is checked. If the else-part is non-existent, or it does
3639 not contain any return or throw statements an exception is thrown. If no
3640 exception is thrown while visiting the if-statement, its children are visited.
3642 A try-statement is checked very similar to an if-statement. Its body must
3643 contain a return or throw. The same applies to its catch clauses and finally
3644 body. Failure to validate produces an \type{IllegalStatementFoundException}.
3646 If the checker does not complain at any point, the selection is considered valid
3647 with respect to return statements.
3649 \subsection{The AmbiguousReturnValueChecker}
3650 This checker verifies that there are no ambiguous return values in a selection.
3652 First, the checker needs to collect some data. Those data are the binding keys
3653 for all simple names that are assigned to within the selection, including
3654 variable declarations, but excluding fields. The checker also finds out whether
3655 a return statement is found in the selection or not. No further checks of return
3656 statements are needed, since, at this point, the selection is already checked
3657 for illegal return statements \see{returnStatementsChecker}.
3659 After the binding keys of the assignees are collected, the checker searches the
3660 part of the enclosing method that is after the selection for references whose
3661 binding keys are among the collected keys. If more than one unique referral is
3662 found, or only one referral is found, but the selection also contains a return
3663 statement, we have a situation with an ambiguous return value, and an exception
3666 %\todoin{Explain why we do not need to consider variables assigned inside
3667 %local/anonymous classes. (The referenced variables need to be final and so
3670 \subsection{The IllegalStatementsChecker}
3671 This checker is designed to check for illegal statements.
3673 Notice that labels in break and continue statements need some special treatment.
3674 Since a label does not have any binding information, we have to search upwards
3675 in the AST to find the \type{LabeledStatement} that corresponds to the label
3676 from the break or continue statement, and check that it is contained in the
3677 selection. If the break or continue statement does not have a label attached to
3678 it, it is checked that its innermost enclosing loop or switch statement (break
3679 statements only) also is contained in the selection.
3681 \todoin{Follow the development in the semantics section\ldots}
3683 \chapter{Technicalities}
3685 \section{Source code organization}
3686 All the parts of this master's project are under version control with
3687 \name{Git}\footnote{\url{http://git-scm.com/}}.
3689 The software written is organized as some \name{Eclipse} plugins. Writing a plugin is
3690 the natural way to utilize the API of \name{Eclipse}. This also makes it possible to
3691 provide a user interface to manually run operations on selections in program
3692 source code or whole projects/packages.
3694 When writing a plugin in \name{Eclipse}, one has access to resources such as the
3695 current workspace, the open editor and the current selection.
3697 The thesis work is contained in the following Eclipse projects:
3700 \item[no.uio.ifi.refaktor] \hfill \\ This is the main Eclipse plugin
3701 project, and contains all of the business logic for the plugin.
3703 \item[no.uio.ifi.refaktor.tests] \hfill \\
3704 This project contains the tests for the main plugin.
3706 \item[no.uio.ifi.refaktor.examples] \hfill \\
3707 Contains example code used in testing. It also contains code for managing
3708 this example code, such as creating an Eclipse project from it before a test
3711 \item[no.uio.ifi.refaktor.benchmark] \hfill \\
3712 This project contains code for running search based versions of the
3713 composite refactoring over selected Eclipse projects.
3715 \item[no.uio.ifi.refaktor.releng] \hfill \\
3716 Contains the rmap, queries and target definitions needed by Buckminster on
3717 the Jenkins continuous integration server.
3721 \subsection{The no.uio.ifi.refaktor project}
3723 \subsubsection{no.uio.ifi.refaktor.analyze}
3724 This package, and its sub-packages, contains code that is used for analyzing
3725 Java source code. The most important sub-packages are presented below.
3728 \item[no.uio.ifi.refaktor.analyze.analyzers] \hfill \\
3729 This package contains source code analyzers. These are usually responsible
3730 for analyzing text selections or running specialized analyzers for different
3731 kinds of entities. Their structures are often hierarchical. This means that
3732 you have an analyzer for text selections, that in turn is utilized by an
3733 analyzer that analyzes all the selections of a method. Then there are
3734 analyzers for analyzing all the methods of a type, all the types of a
3735 compilation unit, all the compilation units of a package, and, at last, all
3736 of the packages in a project.
3738 \item[no.uio.ifi.refaktor.analyze.checkers] \hfill \\
3739 A package containing checkers. The checkers are classes used to validate
3740 that a selection can be further analyzed and chosen as a candidate for a
3741 refactoring. Invalidating properties can be such as usage of inner classes
3742 or the need for multiple return values.
3744 \item[no.uio.ifi.refaktor.analyze.collectors] \hfill \\
3745 This package contains the property collectors. Collectors are used to gather
3746 properties from a text selection. This is mostly properties regarding
3747 referenced names and their occurrences. It is these properties that make up
3748 the basis for finding the best candidates for a refactoring.
3751 \subsubsection{no.uio.ifi.refaktor.change}
3752 This package, and its sub-packages, contains functionality for manipulate source
3756 \item[no.uio.ifi.refaktor.change.changers] \hfill \\
3757 This package contains source code changers. They are used to glue together
3758 the analysis of source code and the actual execution of the changes.
3760 \item[no.uio.ifi.refaktor.change.executors] \hfill \\
3761 The executors that are responsible for making concrete changes are found in
3762 this package. They are mostly used to create and execute one or more Eclipse
3765 \item[no.uio.ifi.refaktor.change.processors] \hfill \\
3766 Contains a refactoring processor for the \MoveMethod refactoring. The code
3767 is stolen and modified to fix a bug. The related bug is described in
3768 \myref{eclipse_bug_429416}.
3772 \subsubsection{no.uio.ifi.refaktor.handlers}
3773 This package contains handlers for the commands defined in the plugin manifest.
3775 \subsubsection{no.uio.ifi.refaktor.prefix}
3776 This package contains the \type{Prefix} type that is the data representation of
3777 the prefixes found by the \type{PrefixesCollector}. It also contains the prefix
3778 set for storing and working with prefixes.
3780 \subsubsection{no.uio.ifi.refaktor.statistics}
3781 The package contains statistics functionality. Its heart is the statistics
3782 aspect that is responsible for gathering statistics during the execution of the
3783 \ExtractAndMoveMethod refactoring.
3786 \item[no.uio.ifi.refaktor.statistics.reports] \hfill \\
3787 This package contains a simple framework for generating reports from the
3788 statistics data generated by the aspect. Currently, the only available
3789 report type is a simple text report.
3794 \subsubsection{no.uio.ifi.refaktor.textselection}
3795 This package contains the two custom text selections that are used extensively
3796 throughout the project. One of them is just a subclass of the other, to support
3797 the use of the memento pattern to optimize the memory usage during benchmarking.
3799 \subsubsection{no.uio.ifi.refaktor.debugging}
3800 The package contains a debug utility class. I addition to this, the package
3801 \code{no.uio.ifi.refaktor.utils.aspects} contains a couple of aspects used for
3804 \subsubsection{no.uio.ifi.refaktor.utils}
3805 Utility package that contains all the functionality that has to do with parsing
3806 of source code. It also has utility classes for looking up handles to methods
3807 and types et cetera.
3810 \item[no.uio.ifi.refaktor.utils.caching] \hfill \\
3811 This package contains the caching manager for compilation units, along with
3812 classes for different caching strategies.
3814 \item[no.uio.ifi.refaktor.utils.nullobjects] \hfill \\
3815 Contains classes for creating different null objects. Most of the classes
3816 are used to represent null objects of different handle types. These null
3817 objects are returned from various utility classes instead of returning a
3818 \var{null} value when other values are not available.
3822 \section{Continuous integration}
3823 The continuous integration server
3824 \name{Jenkins}\footnote{\url{http://jenkins-ci.org/}} has been set up for the
3825 project\footnote{A work mostly done by the supervisor.}. It is used as a way to
3826 run tests and perform code coverage analysis.
3828 To be able to build the \name{Eclipse} plugins and run tests for them with Jenkins, the
3829 component assembly project
3830 \name{Buckminster}\footnote{\url{http://www.eclipse.org/buckminster/}} is used,
3831 through its plugin for Jenkins. Buckminster provides for a way to specify the
3832 resources needed for building a project and where and how to find them.
3833 Buckminster also handles the setup of a target environment to run the tests in.
3834 All this is needed because the code to build depends on an \name{Eclipse}
3835 installation with various plugins.
3837 \subsection{Problems with AspectJ}
3838 The Buckminster build worked fine until introducing AspectJ into the project.
3839 When building projects using AspectJ, there are some additional steps that need
3840 to be performed. First of all, the aspects themselves must be compiled. Then the
3841 aspects need to be woven with the classes they affect. This demands a process
3842 that does multiple passes over the source code.
3844 When using AspectJ with \name{Eclipse}, the specialized compilation and the
3845 weaving can be handled by the \name{AspectJ Development
3846 Tools}\footnote{\url{https://www.eclipse.org/ajdt/}}. This works all fine, but
3847 it complicates things when trying to build a project depending on \name{Eclipse}
3848 plugins outside of \name{Eclipse}. There is supposed to be a way to specify a
3849 compiler adapter for javac, together with the file extensions for the file types
3850 it shall operate. The AspectJ compiler adapter is called
3851 \typewithref{org.aspectj.tools.ant.taskdefs}{Ajc11CompilerAdapter}, and it works
3852 with files that has the extensions \code{*.java} and \code{*.aj}. I tried to
3853 setup this in the build properties file for the project containing the aspects,
3854 but to no avail. The project containing the aspects does not seem to be built at
3855 all, and the projects that depend on it complain that they cannot find certain
3858 I then managed to write an \name{Ant}\footnote{\url{https://ant.apache.org/}}
3859 build file that utilizes the AspectJ compiler adapter, for the
3860 \code{no.uio.ifi.refaktor} plugin. The problem was then that it could no longer
3861 take advantage of the environment set up by Buckminster. The solution to this
3862 particular problem was of a ``hacky'' nature. It involves exporting the plugin
3863 dependencies for the project to an Ant build file, and copy the exported path
3864 into the existing build script. But then the Ant script needs to know where the
3865 local \name{Eclipse} installation is located. This is no problem when building
3866 on a local machine, but to utilize the setup done by Buckminster is a problem
3867 still unsolved. To get the classpath for the build setup correctly, and here
3868 comes the most ``hacky'' part of the solution, the Ant script has a target for
3869 copying the classpath elements into a directory relative to the project
3870 directory and checking it into Git. When no \code{ECLIPSE\_HOME} property is set
3871 while running Ant, the script uses the copied plugins instead of the ones
3872 provided by the \name{Eclipse} installation when building the project. This
3873 obviously creates some problems with maintaining the list of dependencies in the
3874 Ant file, as well as remembering to copy the plugins every time the list of
3875 dependencies changes.
3877 The Ant script described above is run by Jenkins before the Buckminster setup
3878 and build. When setup like this, the Buckminster build succeeds for the projects
3879 not using AspectJ, and the tests are run as normal. This is all good, but it
3880 feels a little scary, since the reason for Buckminster not working with AspectJ
3883 The problems with building with AspectJ on the Jenkins server lasted for a
3884 while, before they were solved. This is reflected in the ``Test Result Trend''
3885 and ``Code Coverage Trend'' reported by Jenkins.
3887 \chapter{Benchmarking}\label{sec:benchmarking}
3888 This part of the master's project is located in the \name{Eclipse} project
3889 \code{no.uio.ifi.refaktor.benchmark}. The purpose of it is to run the equivalent
3890 of the \type{SearchBasedExtractAndMoveMethodChanger}
3891 \see{searchBasedExtractAndMoveMethodChanger} over a larger software project,
3892 both to test its robustness but also its effect on different software metrics.
3894 \section{The benchmark setup}
3895 The benchmark itself is set up as a \name{JUnit} test case. This is a convenient
3896 setup, and utilizes the \name{JUnit Plugin Test Launcher}. This provides us with
3897 a fully functional \name{Eclipse} workbench. Most importantly, this gives us
3898 access to the Java Model of \name{Eclipse} \see{javaModel}.
3900 \subsection{The ProjectImporter}
3901 The Java project that is going to be used as the data for the benchmark, must be
3902 imported into the JUnit workspace. This is done by the
3903 \typewithref{no.uio.ifi.refaktor.benchmark}{ProjectImporter}. The importer
3904 requires the absolute path to the project description file. This file is named
3905 \code{.project} and is located at the root of the project directory.
3907 The project description is loaded to find the name of the project to be
3908 imported. The project that shall be the destination for the import is created in
3909 the workspace, on the base of the name from the description. Then an import
3910 operation is created, based on both the source and destination information. The
3911 import operation is run to perform the import.
3913 I have found no simple API call to accomplish what the importer does, which
3914 tells me that it may not be too many people performing this particular action.
3915 The solution to the problem was found on \name{Stack
3916 Overflow}\footnote{\url{https://stackoverflow.com/questions/12401297}}. It
3917 contains enough dirty details to be considered inconvenient to use, if not
3918 wrapping it in a class like my \type{ProjectImporter}. One would probably have
3919 to delve into the source code for the import wizard to find out how the import
3920 operation works, if no one had already done it.
3922 \section{Statistics}
3923 Statistics for the analysis and changes is captured by the
3924 \typewithref{no.uio.ifi.refaktor.aspects}{StatisticsAspect}. This an
3925 \emph{aspect} written in \name{AspectJ}.
3927 \subsection{AspectJ}
3928 \name{AspectJ}\footnote{\url{http://eclipse.org/aspectj/}} is an extension to
3929 the Java language, and facilitates combining aspect-oriented programming with
3930 the object-oriented programming in Java.
3932 Aspect-oriented programming is a programming paradigm that is meant to isolate
3933 so-called \emph{cross-cutting concerns} into their own modules. These
3934 cross-cutting concerns are functionalities that span over multiple classes, but
3935 may not belong naturally in any of them. It can be functionality that does not
3936 concern the business logic of an application, and thus may be a burden when
3937 entangled with parts of the source code it does not really belong. Examples
3938 include logging, debugging, optimization and security.
3940 Aspects are interacting with other modules by defining advices. The concept of
3941 an \emph{advice} is known from both aspect-oriented and functional
3942 programming\citing{wikiAdvice2014}. It is a function that modifies another
3943 function when the latter is run. An advice in AspectJ is somewhat similar to a
3944 method in Java. It is meant to alter the behavior of other methods, and contains
3945 a body that is executed when it is applied.
3947 An advice can be applied at a defined \emph{pointcut}. A pointcut picks out one
3948 or more \emph{join points}. A join point is a well-defined point in the
3949 execution of a program. It can occur when calling a method defined for a
3950 particular class, when calling all methods with the same name,
3951 accessing/assigning to a particular field of a given class and so on. An advice
3952 can be declared to run both before, after returning from a pointcut, when there
3953 is thrown an exception in the pointcut or after the pointcut either returns or
3954 throws an exception. In addition to picking out join points, a pointcut can
3955 also bind variables from its context, so they can be accessed in the body of an
3956 advice. An example of a pointcut and an advice is found in
3957 \myref{lst:aspectjExample}.
3960 \begin{minted}{aspectj}
3961 pointcut methodAnalyze(
3962 SearchBasedExtractAndMoveMethodAnalyzer analyzer) :
3963 call(* SearchBasedExtractAndMoveMethodAnalyzer.analyze())
3964 && target(analyzer);
3966 after(SearchBasedExtractAndMoveMethodAnalyzer analyzer) :
3967 methodAnalyze(analyzer) {
3968 statistics.methodCount++;
3969 debugPrintMethodAnalysisProgress(analyzer.method);
3972 \caption{An example of a pointcut named \method{methodAnalyze},
3973 and an advice defined to be applied after it has occurred.}
3974 \label{lst:aspectjExample}
3977 \subsection{The Statistics class}
3978 The statistics aspect stores statistical information in an object of type
3979 \type{Statistics}. As of now, the aspect needs to be initialized at the point in
3980 time where it is desired that it starts its data gathering. At any point in time
3981 the statistics aspect can be queried for a snapshot of the current statistics.
3983 The \type{Statistics} class also includes functionality for generating a report
3984 of its gathered statistics. The report can be given either as a string or it can
3985 be written to a file.
3987 \subsection{Advices}
3988 The statistics aspect contains advices for gathering statistical data from
3989 different parts of the benchmarking process. It captures statistics from both
3990 the analysis part and the execution part of the composite \ExtractAndMoveMethod
3993 For the analysis part, there are advices to count the number of text selections
3994 analyzed and the number of methods, types, compilation units and packages
3995 analyzed. There are also advices that counts for how many of the methods there
3996 are found a selection that is a candidate for the refactoring, and for how many
3997 methods there are not.
3999 There exist advices for counting both the successful and unsuccessful executions
4000 of all the refactorings. Both for the \ExtractMethod and \MoveMethod
4001 refactorings in isolation, as well as for the combination of them.
4003 \section{Optimizations}
4004 When looking for possible optimizations for the benchmarking process, I used the
4005 \name{VisualVM}\footnote{\url{http://visualvm.java.net/}} \gloss{profiler} for
4006 the Java Virtual Machine to both profile the application and also to make memory
4009 \subsection{Caching}
4010 When \gloss{profiling} the benchmark process before making any optimizations, it
4011 early became apparent that the parsing of source code was a place to direct
4012 attention towards. This discovery was done when only \emph{analyzing} source
4013 code, before trying to do any \emph{manipulation} of it. Caching of the parsed
4014 ASTs seemed like the best way to save some time, as expected. With only a simple
4015 cache of the most recently used AST, the analysis time was speeded up by a
4016 factor of around 20. This number depends a little upon which type of system the
4019 The caching is managed by a cache manager, that now, by default, utilizes the
4020 not so well known feature of Java called a \emph{soft reference}. Soft
4021 references are best explained in the context of weak references. A \emph{weak
4022 reference} is a reference to an object instance that is only guaranteed to
4023 persist as long as there is a \emph{strong reference} or a soft reference
4024 referring the same object. If no such reference is found, its referred object is
4025 garbage collected. A strong reference is basically the same as a regular Java
4026 reference. A soft reference has the same guarantees as a week reference when it
4027 comes to its relation to strong references, but it is not necessarily garbage
4028 collected if there are no strong references to it. A soft reference \emph{may}
4029 reside in memory as long as the JVM has enough free memory in the heap. A soft
4030 reference will therefore usually perform better than a weak reference when used
4031 for simple caching and similar tasks. The way to use a soft/weak reference is to
4032 as it for its referent. The return value then has to be tested to check that it
4033 is not \var{null}. For the basic usage of soft references, see
4034 \myref{lst:softReferenceExample}. For a more thorough explanation of weak
4035 references in general, see\citing{weakRef2006}.
4038 \begin{minted}{java}
4040 Object strongRef = new Object();
4043 SoftReference<Object> softRef =
4044 new SoftReference<Object>(new Object());
4046 // Using the soft reference
4047 Object obj = softRef.get();
4052 \caption{Showing the basic usage of soft references. Weak references is used the
4053 same way. {\footnotesize (The references are part of the \code{java.lang.ref}
4055 \label{lst:softReferenceExample}
4058 The cache based on soft references has no limit for how many ASTs it caches. It
4059 is generally not advisable to keep references to ASTs for prolonged periods of
4060 time, since they are expensive structures to hold on to. For regular plugin
4061 development, \name{Eclipse} recommends not creating more than one AST at a time to
4062 limit memory consumption. Since the benchmarking has nothing to do with user
4063 experience, and throughput is everything, these advices are intentionally
4064 ignored. This means that during the benchmarking process, the target \name{Eclipse}
4065 application may very well work close to its memory limit for the heap space for
4066 long periods during the benchmark.
4068 \subsection{Candidates stored as mementos}
4069 When performing large scale analysis of source code for finding candidates to
4070 the \ExtractAndMoveMethod refactoring, memory is an issue. One of the inputs to
4071 the refactoring is a variable binding. This variable binding indirectly retains
4072 a whole AST. Since ASTs are large structures, this quickly leads to an
4073 \type{OutOfMemoryError} if trying to analyze a large project without optimizing
4074 how we store the candidates' data. This means that the JVM cannot allocate more
4075 memory for our benchmark, and it exits disgracefully.
4077 A possible solution could be to just allow the JVM to allocate even more memory,
4078 but this is not a dependable solution. The allocated memory could easily
4079 supersede the physical memory of a machine, which would make the benchmark go
4082 Thus, the candidates' data must be stored in another format. Therefore, we use
4083 the \gloss{mementoPattern} to store variable binding information. This is done
4084 in a way that makes it possible to retrieve a variable binding at a later point.
4085 The data that is stored to achieve this, is the key to the original variable
4086 binding. In addition to the key, we know which method and text selection the
4087 variable is referenced in, so that we can find it by parsing the source code and
4088 search for it when it is needed.
4090 \section{Handling failures}
4094 \chapter{Case Studies}
4096 In this chapter I am going to present a few case studies. This is done to give
4097 an impression of how the search-based \ExtractAndMoveMethod refactoring
4098 performs when giving it a larger project to take on. I will try to answer where
4099 it lacks, in terms of completeness, as well as showing its effect on refactored
4102 The first and primary case, is refactoring source code from the \name{Eclipse
4103 JDT UI} project. The project is chosen because it is a well-known open-source
4104 project, still in development, with a large code base that is written by many
4105 different people over several years. The code is installed in a large number of
4106 \name{Eclipse} applications worldwide, and many other projects build on the
4107 Eclipse platform. For a long time, it was even the official IDE for Android
4108 development. All this means that Eclipse must be seen as a good representative
4109 for professionally written Java source code. It is also the home for most of the
4110 JDT refactoring code.
4112 For the second case, the \ExtractAndMoveMethod refactoring is fed the
4113 \code{no.uio.ifi.refaktor} project. This is done as a variation of the
4114 ``dogfooding'' methodology, where you use your own tools to do your job, also
4115 referred to as ``eating your own dog
4116 food''\citing{harrisonDogfooding2006}.
4119 For conducting these experiments, three software tools are used. Two of the
4120 tools both use Eclipse as their platform. The first is our own tool, described
4121 in \myref{sec:benchmarking}, written to be able to run the \ExtractAndMoveMethod
4122 refactoring as a batch process. It analyzes and refactors all the methods of a
4123 project in sequence. The second is JUnit, which is used for running the
4124 project's own unit tests on the target code both before and after it is
4125 refactored. The last tool that is used is a code quality management tool, called
4126 \name{SonarQube}. It can be used to perform different tasks for assuring code
4127 quality, but we are only going to take advantage of one of its main features,
4128 namely quality profiles.
4130 A quality profile is used to define a set of coding rules that a project is
4131 supposed to comply with. Failure to following these rules will be recorded as
4132 so-called ``issues'', marked as having one of several degrees of severities,
4133 ranging from ``info'' to ``blocker'', where the latter one is the most severe.
4134 The measurements done for these case studies are therefore not presented as
4135 fine-grained software metrics results, but rather as the number of issues for
4138 In its analysis, \name{SonarQube} discriminates between functions and accessors.
4139 Accessors are methods that are recognized as setters or getters.
4141 In addition to the coding rules defined through quality profiles,
4142 \name{SonarQube} calculates the complexity of source code. The metric that is
4143 used is cyclomatic complexity, developed by Thomas J. McCabe in
4144 1976\citing{mccabeComplexity1976}. In this metric, functions have an initial
4145 complexity of 1, and whenever the control flow of a function splits, the
4146 complexity increases by
4147 one\footnote{\url{http://docs.codehaus.org/display/SONAR/Metric+definitions}}.
4148 Accessors are not counted in the complexity analysis.
4150 Specifications for the computer used during the experiments are shown in
4151 \myref{tab:experimentComputerSpecs}.
4154 \caption{Specifications for experiment computer.}
4155 \label{tab:experimentComputerSpecs}
4157 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{0.35}R{1.65}@{}}
4159 \spancols{2}{Hardware} \\
4161 Model & Lenovo ThinkPad Edge S430 \\
4162 Processor & Intel\textregistered{} Core\texttrademark{}
4163 i5-3210M\linebreak[4] (2.5 GHz/3.1 GHz (turbo),
4164 2 cores, 4 threads, 3 MB Cache) \\
4165 Memory & 8 GB DDR3 1600 MHz \\
4166 Storage & 500 GB HDD (7200 RPM) + 16 GB SSD Cache for Lenovo Hard Disk Drive
4167 Performance Booster \\
4169 \spancols{2}{Operating system} \\
4171 Distribution & Ubuntu 12.10 \\
4172 Kernel & Linux 3.5.0-49-generic (x86\_64) \\
4179 \section{The \name{SonarQube} quality profile}
4180 The quality profile that is used with \name{SonarQube} in these case studies has got
4181 the name \name{IFI Refaktor Case Study} (version 6). The rules defined in the
4182 profile are chosen because they are the available rules found in \name{SonarQube} that
4183 measures complexity and coupling. Now follows a description of the rules in the
4184 quality profile. The values that are set for these rules are listed in
4185 \myref{tab:qualityProfile1}.
4188 \item[Avoid too complex class] is a rule that measures cyclomatic complexity
4189 for every statement in the body of a class, except for setters and getter.
4190 The threshold value set is its default value of 200.
4192 \item[Classes should not be coupled to too many other classes ] is a rule that
4193 measures how many other classes a class depends upon. It does not count the
4194 dependencies of nested classes. It is meant to promote the Single
4195 Responsibility Principle. The metric for the rule resembles the CBO metric
4196 that is described in \myref{sec:CBO}, but is only considering outgoing
4197 dependencies. The max value for the rule is chosen on the basis of an
4198 empirical study by Raed Shatnawi, which concludes that the number 9 is the
4199 most useful threshold for the CBO metric\citing{shatnawiQuantitative2010}.
4200 This study is also performed on Eclipse source code, so this threshold value
4201 should be particularly well suited for the Eclipse JDT UI case in this
4204 \item[Control flow statements \ldots{} should not be nested too deeply] is
4205 a rule that is meant to counter ``Spaghetti code''. It measures the nesting
4206 level of \emph{if}, \emph{for}, \emph{while}, \emph{switch} and \emph{try}
4207 statements. The nesting levels start at 1. The max value set is its default
4210 \item[Methods should not be too complex] is a rule that measures cyclomatic
4211 complexity the same way as the ``Avoid too complex class'' rule. The max
4212 value used is 10, which ``seems like a reasonable, but not magical, upper
4213 limit``\citing{mccabeComplexity1976}.
4215 \item[Methods should not have too many lines] is a rule that simply measures
4216 the number of lines in methods. A threshold value of 20 is used for this
4217 metric. This is based on my own subjective opinions, as the default value of
4218 100 describes method bodies that do not even fit on most screens.
4220 \item[NPath Complexity] is a rule that measures the number of possible
4221 execution paths through a function. The value used is the default value of
4222 200, which seems like a recognized threshold for this metric.
4224 \item[Too many methods] is a rule that measures the number of methods in a
4225 class. The threshold value used is the default value of 10.
4231 \caption{The \name{IFI Refaktor Case Study} quality profile (version 6).}
4232 \label{tab:qualityProfile1}
4234 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{1.5}R{0.5}@{}}
4236 \textbf{Rule} & \textbf{Max value} \\
4238 Avoid too complex class & 200 \\
4239 Classes should not be coupled to too many other classes (Single
4240 Responsibility Principle) & 9 \\
4241 Control flow statements \ldots{} should not be nested too deeply &
4243 Methods should not be too complex & 10 \\
4244 Methods should not have too many lines & 20 \\
4245 NPath Complexity & 200 \\
4246 Too many methods & 10 \\
4253 A precondition for the source code that is going to be the target for a series
4254 of \ExtractAndMoveMethod refactorings, is that it is organized as an Eclipse
4255 project. It is also assumed that the code is free from compilation errors.
4257 \section{The experiment}
4258 For a given project, the first job that is done, is to refactor its source code.
4259 The refactoring batch job produces three things: The refactored project,
4260 statistics gathered during the execution of the series of refactorings, and an
4261 error log describing any errors happening during this execution. See
4262 \myref{sec:benchmarking} for more information about how the refactorings are
4265 After the refactoring process is done, the before- and after-code is analyzed
4266 with \name{SonarQube}. The analysis results are then stored in a database and
4267 displayed through a \name{SonarQube} server with a web interface.
4269 The before- and after-code is also tested with their own unit tests. This is
4270 done to discover any changes in the semantic behavior of the refactored code,
4271 within the limits of these tests.
4273 \section{Case 1: The Eclipse JDT UI project}
4274 This case is the ultimate test for our \ExtractAndMoveMethod refactoring. The
4275 target source code is massive. With its over 300,000 lines of code and over
4276 25,000 methods, it is a formidable task to perform automated changes on it.
4277 There should be plenty of situations where things can go wrong, and, as we shall
4280 I will start by presenting some statistics from the refactoring execution,
4281 before I pick apart the \name{SonarQube} analysis and conclude by commenting on
4282 the results from the unit tests. The configuration for the experiment is
4283 specified in \myref{tab:configurationCase1}.
4286 \caption{Configuration for Case 1.}
4287 \label{tab:configurationCase1}
4289 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{0.67}L{1.33}@{}}
4291 \spancols{2}{Benchmark data} \\
4293 Launch configuration & CaseStudy.launch \\
4294 Project & no.uio.ifi.refaktor.benchmark \\
4295 Repository & gitolite@git.uio.no:ifi-stolz-refaktor \\
4296 Commit & 43c16c04520746edd75f8dc2a1935781d3d9de6c \\
4298 \spancols{2}{Input data} \\
4300 Project & org.eclipse.jdt.ui \\
4301 Repository & git://git.eclipse.org/gitroot/jdt/eclipse.jdt.ui.git \\
4302 Commit & f218388fea6d4ec1da7ce22432726c244888bb6b \\
4303 Branch & R3\_8\_maintenance \\
4304 Tests suites & org.eclipse.jdt.ui.tests.AutomatedSuite,
4305 org.eclipse.jdt.ui.tests.refactoring.all.\-AllAllRefactoringTests \\
4310 \subsection{Statistics}
4311 The statistics gathered during the refactoring execution is presented in
4312 \myref{tab:case1Statistics}.
4315 \caption{Statistics after batch refactoring the Eclipse JDT UI project with
4316 the \ExtractAndMoveMethod refactoring.}
4317 \label{tab:case1Statistics}
4319 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{1.5}R{0.5}@{}}
4321 \spancols{2}{Time used} \\
4323 Total time & 98m38s \\
4324 Analysis time & 14m41s (15\%) \\
4325 Change time & 74m20s (75\%) \\
4326 Miscellaneous tasks & 9m37s (10\%) \\
4328 \spancols{2}{Numbers of each type of entity analyzed} \\
4331 Compilation units & 2,097 \\
4334 Text selections & 591,500 \\
4336 \spancols{2}{Numbers for \ExtractAndMoveMethod refactoring candidates} \\
4338 Methods chosen as candidates & 2,552 \\
4339 Methods NOT chosen as candidates & 25,115 \\
4340 Candidate selections (multiple per method) & 36,843 \\
4342 \spancols{2}{\ExtractAndMoveMethod refactorings executed} \\
4344 Fully executed & 2,469 \\
4345 Not fully executed & 83 \\
4346 Total attempts & 2,552 \\
4348 \spancols{2}{Primitive refactorings executed} \\
4349 \spancols{2}{\small \ExtractMethod refactorings} \\
4351 Performed & 2,483 \\
4352 Not performed & 69 \\
4353 Total attempts & 2,552 \\
4355 \spancols{2}{\small \MoveMethod refactorings} \\
4358 Not performed & 14 \\
4359 Total attempts & 2,483 \\
4365 \subsubsection{Execution time}
4366 I consider the total execution time of approximately 1.5 hours, on a regular
4367 laptop computer, as being acceptable. It clearly makes the batch process
4368 unsuitable for doing any on-demand analysis or changes, but it is good enough
4369 for running periodic jobs, like over-night analysis.
4371 As the statistics show, 75\% of the total time goes into making the actual code
4372 changes. The time consumers are here the primitive \ExtractMethod and
4373 \MoveMethod refactorings. Included in the change time is the parsing and
4374 precondition checking done by the refactorings, as well as textual changes done
4375 to files on disk. All this parsing and disk access is time-consuming, and
4376 constitutes a large part of the change time.
4378 In comparison, the pure analysis time, used to find suitable candidates, only
4379 makes up for 15\% of the total time consumed. This includes analyzing almost
4380 600,000 text selections, while the number of attempted executions of the
4381 \ExtractAndMoveMethod refactoring is only about 2,500. So the number of executed
4382 primitive refactorings is approximately 5,000. Assuming the time used on
4383 miscellaneous tasks are used mostly for parsing source code for the analysis, we
4384 can say that the time used for analyzing code is at most 25\% of the total time.
4385 This means that for every primitive refactoring executed, we can analyze around
4386 360 text selections. So, with an average of about 21 text selections per method,
4387 it is reasonable to say that we can analyze over 15 methods in the time it
4388 takes to perform a primitive refactoring.
4390 \subsubsection{Refactoring candidates}
4391 Out of the 27,667 methods that were analyzed, 2,552 methods contained selections
4392 that were considered candidates for the \ExtractAndMoveMethod refactoring. This
4393 is roughly 9\% off the methods in the project. These 9\% of the methods had on
4394 average 14.4 text selections that were considered possible refactoring
4397 \subsubsection{Executed refactorings}
4398 2,469 out of 2,552 attempts on executing the \ExtractAndMoveMethod refactoring
4399 were successful, giving a success rate of 96.7\%. The failure rate of 3.3\%
4400 stems from situations where the analysis finds a candidate selection, but the
4401 change execution fails. This failure could be an exception that was thrown, and
4402 the refactoring aborts. It could also be the precondition checking for one of
4403 the primitive refactorings that gives us an error status, meaning that if the
4404 refactoring proceeds, the code will contain compilation errors afterwards,
4405 forcing the composite refactoring to abort. This means that if the
4406 \ExtractMethod refactoring fails, no attempt is done for the \MoveMethod
4407 refactoring. \todo{Redundant information? Put in benchmark chapter?}
4409 Out of the 2,552 \ExtractMethod refactorings that were attempted executed, 69 of
4410 them failed. This gives a failure rate of 2.7\% for the primitive refactoring.
4411 In comparison, the \MoveMethod refactoring had a failure rate of 0.6 \% of the
4412 2,483 attempts on the refactoring.
4414 The failure rates for the refactorings are not that bad, if we also take into
4415 account that the pre-refactoring analysis is incomplete.\todo{see \ldots}
4417 \subsection{\name{SonarQube} analysis}
4418 Results from the \name{SonarQube} analysis are shown in
4419 \myref{tab:case1ResultsProfile1}.
4422 \caption{Results for analyzing the Eclipse JDT UI project, before and after
4423 the refactoring, with \name{SonarQube} and the \name{IFI Refaktor Case Study}
4424 quality profile. (Bold numbers are better.)}
4425 \label{tab:case1ResultsProfile1}
4427 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{1.5}R{0.25}R{0.25}@{}}
4429 \textnormal{Number of issues for each rule} & Before & After \\
4431 Avoid too complex class & 81 & \textbf{79} \\
4432 Classes should not be coupled to too many other classes (Single
4433 Responsibility Principle) & \textbf{1,098} & 1,199 \\
4434 Control flow statements \ldots{} should not be nested too deeply & 1,375 &
4436 Methods should not be too complex & 1,518 & \textbf{1,452} \\
4437 Methods should not have too many lines & 3,396 & \textbf{3,291} \\
4438 NPath Complexity & 348 & \textbf{329} \\
4439 Too many methods & \textbf{454} & 520 \\
4441 Total number of issues & 8,270 & \textbf{8,155} \\
4444 \spancols{3}{Complexity} \\
4446 Per function & 3.6 & \textbf{3.3} \\
4447 Per class & \textbf{29.5} & 30.4 \\
4448 Per file & \textbf{44.0} & 45.3 \\
4450 Total complexity & \textbf{84,765} & 87,257 \\
4453 \spancols{3}{Numbers of each type of entity analyzed} \\
4455 Files & 1,926 & 1,926 \\
4456 Classes & 2,875 & 2,875 \\
4457 Functions & 23,744 & 26,332 \\
4458 Accessors & 1,296 & 1,019 \\
4459 Statements & 162,768 & 165,145 \\
4460 Lines of code & 320,941 & 329,112 \\
4462 Technical debt (in days) & \textbf{1,003.4} & 1,032.7 \\
4467 \subsubsection{Diversity in the number of entities analyzed}
4468 The analysis performed by \name{SonarQube} is reporting fewer methods than found
4469 by the pre-refactoring analysis. \name{SonarQube} discriminates between
4470 functions (methods) and accessors, so the 1,296 accessors play a part in this
4471 calculation. \name{SonarQube} also has the same definition as our plugin when
4472 it comes to how a class is defined. Therefore it seems like \name{SonarQube}
4473 misses 277 classes that our plugin handles. This can explain why the {SonarQube}
4474 report differs from our numbers by approximately 2,500 methods,
4476 \subsubsection{Complexity}
4477 On all complexity rules that works on the method level, the number of issues
4478 decreases with between 3.1\% and 6.5\% from before to after the refactoring. The
4479 average complexity of a method decreases from 3.6 to 3.3, which is an
4480 improvement of about 8.3\%. So, on the method level, the refactoring must be
4481 said to have a slightly positive impact. This is due to the extraction of a lot
4482 of methods, making the average method size smaller.
4484 The improvement in complexity on the method level is somewhat traded for
4485 complexity on the class level. The complexity per class metric is worsened by
4486 3\% from before to after. The issues for the ``Too many methods'' rule also
4487 increases by 14.5\%. These numbers indicate that the refactoring makes quite a
4488 lot of the classes a little more complex overall. This is the expected outcome,
4489 since the \ExtractAndMoveMethod refactoring introduces almost 2,500 new methods
4492 The only number that can save the refactoring's impact on complexity on the
4493 class level, is the ``Avoid too complex class'' rule. It improves with 2.5\%,
4494 thus indicating that the complexity is moderately better distributed between the
4495 classes after the refactoring than before.
4497 \subsubsection{Coupling}
4498 One of the hopes when starting this project, was to be able to make a
4499 refactoring that could lower the coupling between classes. Better complexity at
4500 the method level is a not very unexpected byproduct of dividing methods into
4501 smaller parts. Lowering the coupling on the other hand, is a far greater task.
4502 This is also reflected in the results for the only coupling rule defined in the
4503 \name{SonarQube} quality profile, namely the ``Classes should not be coupled to
4505 other classes (Single Responsibility Principle)'' rule.
4507 The number of issues for the coupling rule is 1,098 before the refactoring, and
4508 1,199 afterwards. This is an increase in issues of 9.2\%. These numbers can be
4509 interpreted two ways. The first possibility is that our assumptions are wrong,
4510 and that increasing indirection does not decrease coupling between classes. The
4511 other possibility is that our analysis and choices of candidate text selections
4512 are not good enough. I vote for the second possibility. (Voting against the
4513 public opinion may also be a little bold.)
4515 \subsubsection{An example of what makes the number of dependency issues grow}
4516 \Myref{lst:sonarJDTExampleBefore} shows a portion of the class
4517 \typewithref{org.eclipse.jdt.ui.actions}{ShowActionGroup} from the JDT UI
4518 project before it is refactored with the search-based \ExtractAndMoveMethod
4519 refactoring. Before the refactoring, the \type{ShowActionGroup} class has 12
4520 outgoing dependencies (reported by \name{SonarQube}).
4522 \begin{listing}[htb]
4523 \begin{minted}[linenos,samepage]{java}
4524 public class ShowActionGroup extends ActionGroup {
4526 private void initialize(IWorkbenchSite site,
4527 boolean isJavaEditor) {
4529 ISelectionProvider provider= fSite.getSelectionProvider();
4530 ISelection selection= provider.getSelection();
4531 fShowInPackagesViewAction.update(selection);
4532 if (!isJavaEditor) {
4533 provider.addSelectionChangedListener(
4534 fShowInPackagesViewAction);
4539 \caption{Portion of the \type{ShowActionGroup} class before refactoring.}
4540 \label{lst:sonarJDTExampleBefore}
4543 During the benchmark process, the search-based \ExtractAndMoveMethod refactoring
4544 extracts the lines 6 to 12 of the code in \myref{lst:sonarJDTExampleBefore}, and
4545 moves the new method to the move target, which is the field
4546 \var{fShowInPackagesViewAction} with type
4547 \typewithref{org.eclipse.jdt.ui.actions}{ShowInPackageViewAction}. The result is
4548 shown in \myref{lst:sonarJDTExampleAfter}.
4550 \begin{listing}[htb]
4551 \begin{minted}[linenos,samepage]{java}
4552 public class ShowActionGroup extends ActionGroup {
4554 private void initialize(IWorkbenchSite site,
4555 boolean isJavaEditor) {
4557 fShowInPackagesViewAction.generated_8019497110545412081(
4558 this, isJavaEditor);
4563 \begin{minted}[linenos,samepage]{java}
4564 public class ShowInPackageViewAction
4565 extends SelectionDispatchAction {
4567 public void generated_8019497110545412081(
4568 ShowActionGroup showactiongroup, boolean isJavaEditor) {
4569 ISelectionProvider provider=
4570 showactiongroup.fSite.getSelectionProvider();
4571 ISelection selection= provider.getSelection();
4573 if (!isJavaEditor) {
4574 provider.addSelectionChangedListener(this);
4579 \caption{Portions of the classes \type{ShowActionGroup} and
4580 \type{ShowInPackageViewAction} after refactoring.}
4581 \label{lst:sonarJDTExampleAfter}
4584 After the refactoring, the \type{ShowActionGroup} has only 11 outgoing
4585 dependencies. It no longer depends on the
4586 \typewithref{org.eclipse.jface.viewers}{ISelection} type. So our refactoring
4587 managed to get rid of one dependency, which is exactly what we wanted. The only
4588 problem is, that now the \type{ShowInPackageViewAction} class has got two new
4589 dependencies, in the \type{ISelectionProvider} and the \type{ISelection} types.
4590 The bottom line is that we eliminated one dependency, but introduced two more,
4591 ending up with a program that has more dependencies now than when we started.
4593 What can happen in many situations where the \ExtractAndMoveMethod refactoring
4594 is performed, is that the \MoveMethod refactoring ``drags'' with it references
4595 to classes that are unknown to the method destination. If the refactoring
4596 happens to be so lucky that it removes a dependency from one class, it might as
4597 well introduce a couple of new dependencies to another class, as shown in the
4598 previous example. In those situations where a destination class does not know
4599 about the originating class of a moved method, the \MoveMethod refactoring most
4600 certainly will introduce a dependency. This is because there is a
4601 bug\footnote{\url{https://bugs.eclipse.org/bugs/show\_bug.cgi?id=228635}} in the
4602 refactoring, making it pass an instance of the originating class as a reference
4603 to the moved method, regardless of whether the reference is used in the method
4606 There is also the possibility that the heuristics used to find candidate text
4607 selections are not good enough, they most certainly are not. I wish I had more
4608 time to fine-tune them, and to complete the analysis part of the project, but
4609 this is simply not the case. This becomes even clearer when analyzing the unit
4610 test results for the after-code.
4612 \subsubsection{Totals}
4613 On the bright side, the total number of issues is lower after the refactoring
4614 than it was before. Before the refactoring, the total number of issues is
4615 8,270, and after it is 8,155. An improvement of only 1.4\%.
4617 Then \name{SonarQube} tells me that the total complexity has increased by
4618 2.9\%, and that the (more questionable) ``technical debt'' has increased from
4619 1,003.4 to 1,032.7 days, also a deterioration of 2.9\%. Although these numbers
4620 are similar, no correlation has been found between them.
4622 \subsection{Unit tests}
4623 The tests that have been run for the \name{Eclipse JDT UI} project, are the
4624 tests in the test suites specified as the main test suites on the JDT UI wiki
4625 page on how to contribute to the
4626 project\footnote{\url{https://wiki.eclipse.org/JDT\_UI/How\_to\_Contribute\#Unit\_Testing}}.
4628 \subsubsection{Before the refactoring}
4629 Running the tests for the before-code of Eclipse JDT UI yielded 4 errors and 3
4630 failures for the \type{AutomatedSuite} test suite (2,007 test cases), and 2
4632 3 failures for the \type{AllAllRefactoringTests} test suite (3,816 test cases).
4634 \subsubsection{After the refactoring}
4635 The test results for the after-code of the Eclipse JDT UI project is another
4636 story. The reason for this is that during the setup for the unit tests, Eclipse
4637 now reports that the project contains 322 fatal errors, and a lot of errors that
4638 probably follows from these. This is another blow for this master's project.
4640 It has now been shown that the \ExtractAndMoveMethod refactoring, in its current
4641 state, produces code that does not compile. Had these errors originated from
4642 only one bug, it would not have been much of a problem, but this is not the
4643 case. By only looking at some random compilation problems in the refactored code,
4644 I came up with at least four different bugs \todo{write bug reports?} that
4645 caused those problems. I then stopped looking for more, since some of the bugs
4646 would take more time to fix than I could justify using on them at this point.
4648 The only thing that can be said in my defense, is that all the compilation
4649 errors could have been avoided if the types of situations that cause them were
4650 properly handled by the primitive refactorings, which again are supported by the
4651 Eclipse JDT UI project. All of the four randomly found bugs that I mentioned
4652 before, are also weaknesses of the \MoveMethod refactoring. If the primitive
4653 refactorings had detected the up-coming errors
4654 in their precondition checking phase, the refactorings would have been aborted,
4655 since this is how the \ExtractAndMoveMethod refactoring handles such situations.
4657 Of course, taking all possible situations into account is an immense task. This
4658 is one of the reasons for the failure. A complete analysis is too big of a task
4659 for this master's project to handle. Looking at it now, this comes as no
4660 surprise, since the task is obviously also too big for the creators of the
4661 primitive \MoveMethod refactoring. This shows that the underlying primitive
4662 refactorings are not complete enough to be fully relied upon for avoiding
4665 Considering all these problems, it is difficult to know how to interpret the
4666 unit test results from after refactoring the Eclipse JDT UI. The
4667 \type{AutomatedSuite} reported 565 errors and 5 failures. Three of the failures
4668 were the same as reported before the refactoring took place, so two of them are
4669 new. For these two cases it is not immediately apparent what makes them behave
4670 differently. The program is so complex that to analyze it to find this out, we
4671 might need more powerful methods than just manually analyzing its source code.
4672 This is somewhat characteristic for imperative programming: The programs are
4673 often hard to analyze and understand.
4675 For the \type{AllAllRefactoringTests} test suite, the three failures are gone,
4676 but the two errors have grown to 2,257 errors. I will not try to analyze those
4679 What I can say, is that it is likely that the \ExtractAndMoveMethod refactoring
4680 has introduced some unintended behavioral changes. Let us say that the
4681 refactoring introduces at least two behavior-altering changes for every 2,500
4682 executions. More than that is difficult to say about the behavior-preserving
4683 properties of the \ExtractAndMoveMethod refactoring, at this point.
4685 \subsection{Conclusions}
4686 After automatically analyzing and executing the \ExtractAndMoveMethod
4687 refactoring for all the methods in the Eclipse JDT UI project, the results do
4688 not look that promising. For this case, the refactoring seems almost unusable as
4689 it is now. The error rate and measurements done tells us this.
4691 The refactoring makes the code a little less complex at the method level. But
4692 this is merely a side effect of extracting methods, and holds little scientific
4693 value. When it comes to the overall complexity, it is increased, although it is
4694 slightly better spread among the classes.
4696 The analysis done before the \ExtractAndMoveMethod refactoring, is currently not
4697 complete enough to make the refactoring useful. It introduces too many errors in
4698 the code, and the code may change its behavior. It also remains to prove that
4699 large scale refactoring with it can decrease coupling between classes. A better
4700 analysis may prove this, but in its present state, the opposite is the fact. The
4701 coupling measurements done by \name{SonarQube} show this.
4703 On the bright side, the performance of the refactoring process is not that bad.
4704 It shows that it is possible to make a tool the way we do, if we can make the
4705 tool do anything useful. As long as the analysis phase is not going to involve
4706 anything that uses too much disk access, a lot of analysis can be done in a
4707 reasonable amount of time.
4709 The time used on performing the actual changes excludes a trial and error
4710 approach with the tools used in this master's project. In a trial and error
4711 approach, you could for instance be using the primitive refactorings used in
4712 this project to refactor code, and only then make decisions based on the effect,
4713 possibly shown by traditional software metrics. The problem with the approach
4714 taken in this project, compared to a trial and error approach, is that using
4715 heuristics beforehand is much more complicated. But on the other hand, a trial
4716 and error approach would still need to face the challenges of producing code
4717 that does compile without errors. If using refactorings that could produce
4718 in-memory changes, a trial and error approach could be made more efficient.
4720 \section{Case 2: The \type{no.uio.ifi.refaktor} project}
4721 In this case we will see a form of the ``dogfooding'' methodology used, when
4722 refactoring our own \type{no.uio.ifi.refaktor} project with the
4723 \ExtractAndMoveMethod refactoring.
4725 In this case I will try to point out some differences from case 1, and how they
4726 impact the execution of the benchmark. The refaktor project is 39 times smaller
4727 than the Eclipse JDT UI project, measured in lines of code. This will make
4728 things a bit more transparent. It will therefore be interesting to see if this
4729 case can shed light on any aspect of our project that were lost in the larger
4732 The configuration for the experiment is specified in
4733 \myref{tab:configurationCase2}.
4736 \caption{Configuration for Case 2.}
4737 \label{tab:configurationCase2}
4739 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{0.67}L{1.33}@{}}
4741 \spancols{2}{Benchmark data} \\
4743 Launch configuration & CaseStudyDogfooding.launch \\
4744 Project & no.uio.ifi.refaktor.benchmark \\
4745 Repository & gitolite@git.uio.no:ifi-stolz-refaktor \\
4746 Commit & 43c16c04520746edd75f8dc2a1935781d3d9de6c \\
4748 \spancols{2}{Input data} \\
4750 Project & no.uio.ifi.refaktor \\
4751 Repository & gitolite@git.uio.no:ifi-stolz-refaktor \\
4752 Commit & 43c16c04520746edd75f8dc2a1935781d3d9de6c \\
4754 Test configuration & no.uio.ifi.refaktor.tests/ExtractTest.launch \\
4759 \subsection{Statistics}
4760 The statistics gathered during the refactoring execution is presented in
4761 \myref{tab:case2Statistics}.
4764 \caption{Statistics after batch refactoring the \type{no.uio.ifi.refaktor}
4765 project with the \ExtractAndMoveMethod refactoring.}
4766 \label{tab:case2Statistics}
4768 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{1.5}R{0.5}@{}}
4770 \spancols{2}{Time used} \\
4772 Total time & 1m15s \\
4773 Analysis time & 0m18s (24\%) \\
4774 Change time & 0m47s (63\%) \\
4775 Miscellaneous tasks & 0m10s (14\%) \\
4777 \spancols{2}{Numbers of each type of entity analyzed} \\
4780 Compilation units & 154 \\
4783 Text selections & 8,609 \\
4785 \spancols{2}{Numbers for \ExtractAndMoveMethod refactoring candidates} \\
4787 Methods chosen as candidates & 58 \\
4788 Methods NOT chosen as candidates & 1,012 \\
4789 Candidate selections (multiple per method) & 227 \\
4791 \spancols{2}{\ExtractAndMoveMethod refactorings executed} \\
4793 Fully executed & 53 \\
4794 Not fully executed & 5 \\
4795 Total attempts & 58 \\
4797 \spancols{2}{Primitive refactorings executed} \\
4798 \spancols{2}{\small \ExtractMethod refactorings} \\
4801 Not performed & 2 \\
4802 Total attempts & 58 \\
4804 \spancols{2}{\small \MoveMethod refactorings} \\
4807 Not performed & 3 \\
4808 Total attempts & 56 \\
4814 \subsubsection{Differences}
4815 There are some differences between the two projects that make them a little
4816 difficult to compare by performance.
4818 \paragraph{Different complexity.}
4819 Although the JDT UI project is 39 times greater than the refaktor project in
4820 terms of lines of code, it is only about 26 times its size measured in numbers
4821 of methods. This means that the methods in the refaktor project are smaller in
4822 average than in the JDT project. This is also reflected in the \name{SonarQube}
4823 report, where the complexity per method for the JDT project is 3.6, while the
4824 refaktor project has a complexity per method of 2.1.
4826 \paragraph{Number of selections per method.}
4827 The analysis for the JDT project processed 21 text selections per method in
4828 average. This number for the refaktor project is only 8 selections per method
4829 analyzed. This is a direct consequence of smaller methods.
4831 \paragraph{Different candidates to methods ratio.}
4832 The differences in how the projects are factored are also reflected in the
4833 ratios for how many methods that are chosen as candidates compared to the total
4834 number of methods analyzed. For the JDT project, 9\% of the methods were
4835 considered to be candidates, while for the refaktor project, only 5\% of the
4836 methods were chosen.
4838 \paragraph{The average number of possible candidate selection.}
4839 For the methods that are chosen as candidates, the average number of possible
4840 candidate selections for these methods differ quite much. For the JDT project,
4841 the number of possible candidate selections for these methods was 14.44
4842 selections per method, while the candidate methods in the refaktor project had
4843 only 3.91 candidate selections to choose from, in average.
4845 \subsubsection{Execution time}
4846 The differences in complexity, and the different candidate methods to total
4847 number of methods ratios, is shown in the distributions of the execution times.
4848 For the JDT project, 75\% of the total time was used on the actual changes,
4849 while for the refaktor project, this number was only 63\%.
4851 For the JDT project, the benchmark used on average 0.21 seconds per method in
4852 the project, while for the refaktor project it used only 0.07 seconds per
4853 method. So the process used 3 times as much time per method for the JDT project
4854 than for the refaktor project.
4856 While the JDT project is 39 times larger than the refaktor project measured in
4857 lines of code, the benchmark used about 79 times as long time on it than for the
4858 refaktor project. Relatively, this is about twice as long.
4860 Since the details of these execution times are not that relevant to this
4861 master's project, only their magnitude, I will leave them here.
4863 \subsubsection{Executed refactorings}
4864 For the composite \ExtractAndMoveMethod refactoring performed in case 2, 53
4865 successful attempts out of 58 gives a success rate of 91.4\%. This is 5.3
4866 percentage points worse than for case 1.
4868 \subsection{\name{SonarQube} analysis}
4869 Results from the \name{SonarQube} analysis are shown in
4870 \myref{tab:case2ResultsProfile1}.
4872 Not much is to be said about these results. The trends in complexity and
4873 coupling are the same. We end up a little worse after the refactoring process
4877 \caption{Results for analyzing the \var{no.uio.ifi.refaktor} project, before
4878 and after the refactoring, with \name{SonarQube} and the \name{IFI Refaktor
4879 Case Study} quality profile. (Bold numbers are better.)}
4880 \label{tab:case2ResultsProfile1}
4882 \begin{tabularx}{\textwidth}{@{}>{\bfseries}L{1.5}R{0.25}R{0.25}@{}}
4884 \textnormal{Number of issues for each rule} & Before & After \\
4886 Avoid too complex class & 1 & 1 \\
4887 Classes should not be coupled to too many other classes (Single
4888 Responsibility Principle) & \textbf{29} & 34 \\
4889 Control flow statements \ldots{} should not be nested too deeply & 24 &
4891 Methods should not be too complex & 17 & \textbf{15} \\
4892 Methods should not have too many lines & 41 & \textbf{40} \\
4893 NPath Complexity & 3 & 3 \\
4894 Too many methods & \textbf{13} & 15 \\
4896 Total number of issues & \textbf{128} & 129 \\
4899 \spancols{3}{Complexity} \\
4901 Per function & 2.1 & 2.1 \\
4902 Per class & \textbf{12.5} & 12.9 \\
4903 Per file & \textbf{13.8} & 14.2 \\
4905 Total complexity & \textbf{2,089} & 2,148 \\
4908 \spancols{3}{Numbers of each type of entity analyzed} \\
4910 Files & 151 & 151 \\
4911 Classes & 167 & 167 \\
4912 Functions & 987 & 1,045 \\
4913 Accessors & 35 & 30 \\
4914 Statements & 3,355 & 3,416 \\
4915 Lines of code & 8,238 & 8,460 \\
4917 Technical debt (in days) & \textbf{19.0} & 20.7 \\
4922 \subsection{Unit tests}
4923 The tests used for this case are the same that has been developed throughout the
4926 The code that was refactored for this case suffered from some of the problems
4927 discovered in case 1. This means that the after-code for case 2 also contained
4928 compilation errors, but they were not as many. The code contained only 6 errors
4929 that made the code not compile.
4931 All of the errors made, originated from the same bug. It is a bug that happens
4932 in situation where a class instance creation is moved from between packages, and
4933 the class for the instance is package-private. The \MoveMethod refactoring does
4934 not detect that there will be a visibility problem, and neither does it promote
4935 the package-private class to be public.
4937 Since the errors were easy to fix manually, I corrected them and ran the unit
4938 tests as planned. Before the refactoring, all tests passed. All tests also
4939 passed after the refactoring, with the six error corrections. Since the
4940 corrections done are not of a kind that could make the behavior of the program
4941 change, it is likely that the refactorings done to the
4942 \type{no.uio.ifi.refaktor} project did not change its behavior. This is also
4943 supported by the informal experiment presented next.
4945 \subsection{An informal experiment}
4946 To complete the task of ``eating my own dog food'', I conducted an informal
4947 experiment where I used the refactored version of the \type{no.uio.ifi.refaktor}
4948 project, with the corrections, to again refaktor ``itself''.
4950 The experiment produced code containing the same six errors as after the
4951 previous experiment. I also compared the after-code from the two experiments
4952 with a diff-tool. The only differences found were different method names. This
4953 is expected, since the method names are randomly generated by the
4954 \ExtractAndMoveMethod refactoring.
4956 The outcome of this simple experiment makes me more confident that the
4957 \ExtractAndMoveMethod refactoring made only behavior-preserving changes to the
4958 \type{no.uio.ifi.refaktor} project, apart from the compilation errors.
4960 \subsection{Conclusions}
4961 The differences in complexity between the Eclipse JDT UI project and the
4962 \type{no.uio.ifi.refaktor} project, clearly influenced the differences in their
4963 execution times. This is mostly because fewer of the methods were chosen to be
4964 refactored for the refaktor project than for the JDT project. This makes it
4965 difficult to know if there are any severe performance penalties associated with
4966 refactoring on a large project compared to a small one.
4968 The trends in the \name{SonarQube} analysis are the same for this case as for
4969 the previous one. This gives more confidence in the these results.
4971 By refactoring our own code and using it again to refactor our code, we showed
4972 that it is possible to write an automated composite refactoring that works for
4973 many cases. That it probably did not alter the behavior of a smaller project
4974 shows us nothing more than that though, and might just be a coincidence.
4977 \todoin{Write? Or wrap up in final conclusions?}
4980 \chapter{Conclusions and Future Work}
4981 This chapter will conclude this master's thesis. It will try to give justified
4982 answers to the research questions posed \see{sec:researchQuestions} and present
4983 some future work that could be done to take this project to the next level.
4985 \section{Conclusions}
4986 One of the motivations for this thesis was to create a fully automated composite
4987 refactoring that could be used to make program source code better in terms of
4988 coupling between classes. Earlier, in \mysimpleref{sec:CBO}, it was shown that a
4989 composition of the \ExtractMethod and the \MoveMethod refactorings reduces the
4990 coupling between two classes in an ideal situation. The Eclipse IDE implements
4991 both these refactorings, as well as providing a framework for analyzing source
4992 code, so it was considered a suitable tool to build upon for our project.
4994 The search-based \ExtractAndMoveMethod refactoring was created by utilizing the
4995 analysis and refactoring support of Eclipse, and a small framework was built
4996 for executing large scale refactoring with it. The refactoring was set up to
4997 analyze and execute changes on the Eclipse JDT UI project. Statistics was
4998 gathered during this process and the resulting code was analyzed through
4999 SonarQube. The project's own unit tests were also performed to find out if our
5000 refactoring introduces any behavior-altering changes in the code it refactor.
5002 \paragraph{Answering the main research question}
5003 The first and greatest challenge was to find out if the \ExtractAndMoveMethod
5004 refactoring could be automated, in all tasks ranging from analysis to executing
5005 changes. It is now confirmed that this can be done, since it has been
5006 implemented as a part of the work done for this project. It has also been shown
5007 that the refactoring can be used to refactor large code bases, through the case
5008 study done on the Eclipse JDT UI project.
5010 If we ask if using the existing Eclipse refactorings for this task is
5011 \emph{easy}, this is another question. The refactorings provided by the JDT UI
5012 project are clearly not meant to be combined in any way. The preconditions for
5013 one refactoring are not always easily retrievable after the execution of
5014 another. Also, the refactorings are all assuming that the code they shall
5015 refactor is textualized. This means that the source code must be parsed between
5016 the executions of each refactoring. Another problem with this dependency on
5017 textual changes is that you cannot make a composition of two refactorings appear
5018 as one change if their changes overlap. This will make the undo-history of the
5019 refactoring show two changes instead of one, and is not nice for usability it
5020 the refactoring would be used as an on-demand refactoring in an IDE.
5022 Apart from the problems with implementing the actual refactoring, the analysis
5023 framework is quite nicely solved in Eclipse. The AST generated when parsing
5024 source code supports using visitors to traverse it, and this works without
5027 \paragraph{Is the refactoring efficient enough?}
5028 By being able to process over 300,000 pure lines of code in about 1.5 hours on a
5029 mid-level laptop computer, the search-based \ExtractAndMoveMethod refactoring
5030 must be said to perform reasonably well.
5032 \paragraph{What about breaking the source code?}
5033 The case studies showed that our safety measures that rely on the precondition
5034 checking of the existing primitive refactorings are not good enough in practice.
5035 If we were going to assure that code we change compiles, we would need to
5036 consider all possible situations where the refactoring could fail and search for
5037 them in our analysis. It is an open question if this is even feasible. Our
5038 analysis is incomplete, and so is the analysis for the \ExtractMethod and the
5039 \MoveMethod refactorings.
5041 Our refactoring does not take any precautions to preserve behavior. A few
5042 running \todo{check out test report} and failing unit test for the JDT UI
5043 project after the refactoring indicate that our refactoring probably causes some
5044 changes to the way a program behaves.
5046 \paragraph{Is the quality of the code improved?}
5047 For coupling, there is no evidence that the refactoring improves the quality of
5048 source code. Shall we believe the SonarQube analysis from the case studies, our
5049 refactoring makes classes more coupled after the refactoring than before, in the
5050 general case. This is probably because our analysis and heuristics for finding
5051 the best candidates for the refactoring are not adequate.
5053 \paragraph{Is the refactoring useful?}
5054 In its present state, the refactoring cannot be said to be very useful. It
5055 generates too many compilation errors for it to fall into that category. On the
5056 other hand, if the problems with the search-based \ExtractAndMoveMethod
5057 refactoring were to be solved it could be useful in some situations.
5059 If the refactoring was perfected, it could of course be used as a regular
5060 on-demand automated refactoring on a per method base (or per class, package or
5063 As it is now, the refactoring is not very well suited to be set to perform
5064 unattended refactoring. But if we could find a way to filter out the changes
5065 that create compilation errors, we could use the refactoring to look for
5066 improvement points in a software project. This process could for instance be
5067 scheduled to run at regular times, possibly after a nightly build or the like.
5068 Then the results could be made available, and an administrator could be set to
5069 review them and choose whether or not they should be performed.
5071 \section{Future work}
5072 \todoin{Find out if a complete analysis is feasible}
5073 \todoin{Complete the analysis}
5074 \todoin{Make refactorings safer (behavior)}
5075 \todoin{Improve heuristics/introduce metrics}
5081 \chapter{Eclipse Bugs Submitted}
5082 \newcommand{\submittedBugReport}[1]{The submitted bug report can be found on
5085 \section{Eclipse bug 420726: Code is broken when moving a method that is
5086 assigning to the parameter that is also the move
5087 destination}\label{eclipse_bug_420726}
5089 was found when analyzing what kinds of names that were to be considered as
5090 \emph{unfixes} \see{unfixes}.
5093 The bug emerges when trying to move a method from one class to another, and when
5094 the target for the move (must be a variable, local or field) is both a parameter
5095 variable and also is assigned to within the method body. \name{Eclipse} allows this to
5096 happen, although it is the sure path to a compilation error. This is because we
5097 would then have an assignment to a \var{this} expression, which is not allowed
5099 \submittedBugReport{https://bugs.eclipse.org/bugs/show\_bug.cgi?id=420726}
5101 \paragraph{The solution}
5102 The solution to this problem is to add all simple names that are assigned to in
5103 a method body to the set of unfixes.
5105 \section{Eclipse bug 429416: IAE when moving method from anonymous
5106 class}\label{eclipse_bug_429416}
5108 this bug during a batch change on the \type{org.eclipse.jdt.ui} project.
5111 This bug surfaces when trying to use the \refa{Move Method} refactoring to move a
5112 method from an anonymous class to another class. This happens both for my
5113 simulation as well as in \name{Eclipse}, through the user interface. It only occurs
5114 when \name{Eclipse} analyzes the program and finds it necessary to pass an
5115 instance of the originating class as a parameter to the moved method. I.e. it
5116 wants to pass a \var{this} expression. The execution ends in an
5117 \typewithref{java.lang}{IllegalArgumentException} in
5118 \typewithref{org.eclipse.jdt.core.dom}{SimpleName} and its
5119 \method{setIdentifier(String)} method. The simple name is attempted created in
5121 \methodwithref{org.eclipse.jdt.internal.corext.refactoring.structure.\\MoveInstanceMethodProcessor}{createInlinedMethodInvocation}
5122 so the \type{MoveInstanceMethodProcessor} was early a clear suspect.
5124 The \method{createInlinedMethodInvocation} is the method that creates a method
5125 invocation where the previous invocation to the method that was moved was
5126 located. From its code it can be read that when a \var{this} expression is going
5127 to be passed in to the invocation, it shall be qualified with the name of the
5128 original method's declaring class, if the declaring class is either an anonymous
5129 class or a member class. The problem with this, is that an anonymous class does
5130 not have a name, hence the term \emph{anonymous} class! Therefore, when its
5131 name, an empty string, is passed into
5132 \methodwithref{org.eclipse.jdt.core.dom.AST}{newSimpleName} it all ends in an
5133 \type{IllegalArgumentException}.
5134 \submittedBugReport{https://bugs.eclipse.org/bugs/show\_bug.cgi?id=429416}
5136 \paragraph{How I solved the problem}
5137 Since the \type{MoveInstanceMethodProcessor} is instantiated in the
5138 \typewithref{no.uio.ifi.refaktor.change.executors}{MoveMethod\-RefactoringExecutor},
5139 and only need to be a
5140 \typewithref{org.eclipse.ltk.core.refactoring.participants}{MoveProcessor}, I
5141 was able to copy the code for the original move processor and modify it so that
5142 it works better for me. It is now called
5143 \typewithref{no.uio.ifi.refaktor.change.processors}{ModifiedMoveInstanceMethodProcessor}.
5144 The only modification done (in addition to some imports and suppression of
5145 warnings), is in the \method{createInlinedMethodInvocation}. When the declaring
5146 class of the method to move is anonymous, the \var{this} expression in the
5147 parameter list is not qualified with the declaring class' (empty) name.
5149 \section{Eclipse bug 429954: Extracting statement with reference to local type
5150 breaks code}\label{eclipse_bug_429954}
5151 The bug was discovered when doing some changes to the way unfixes is computed.
5154 The problem is that \name{Eclipse} is allowing selections that references variables of
5155 local types to be extracted. When this happens the code is broken, since the
5156 extracted method must take a parameter of a local type that is not in the
5157 methods scope. The problem is illustrated in
5158 \myref{lst:extractMethod_LocalClass}, but there in another setting.
5159 \submittedBugReport{https://bugs.eclipse.org/bugs/show\_bug.cgi?id=429954}
5161 \paragraph{Actions taken}
5162 There are no actions directly springing out of this bug, since the Extract
5163 Method refactoring cannot be meant to be this way. This is handled on the
5164 analysis stage of our \refa{Extract and Move Method} refactoring. So names representing
5165 variables of local types is considered unfixes \see{unfixes}.
5166 \todoin{write more when fixing this in legal statements checker}