Talk:Code refactoring

deleted paragraph about refactoring historically being avoided
The paragraph about refactoring being historically avoided in general is wrong. It's long been recognized as the right thing to do but often only been put off for lack of resources and automated tools. 132.160.49.90 (talk) 18:59, 19 September 2013 (UTC)

(function calling from functions)
(heading inserted for readability ... said: Rursus (bork²) 08:01, 9 July 2009 (UTC))

Can anyone help me on the following sentence? "After a certain point, it becomes clear that functions can benefit from using functions themselves."

I am translating this article to Chinese, but I can't catch it. Thanks.--202.99.60.155 02:11, 17 Jan 2005 (UTC)

It means that it is often useful to break a complex function into calls of less-complex functions. For example, a complex function that reads weather sensors, computes the wheather for next week and produces maps could be broken into three smaller functions that are called in sequence.

--Stephan Leclercq 09:04, 17 Jan 2005 (UTC)

In software engineering, refactoring is *strictly* bound to object oriented code. The term comes from 'factorization'. In OO design, 'to factorize' means 'to distribute responsibilities among classes and objects'.

If you want to talk about behaviour-preserving transformations in structural programs, (or in assembly code or declarative code or whatever) there are other terms, such as 'to restructurize' or 'to reengineer' or similar. Leave refactoring to OOP.

-- [Peter]


 * I disagree. Refactoring is not strictly bound to OOP.  Refactoring is about behavior-preserving transformations, cleaning code, etc.  Concepts that are common to all languages.  It is true that some types of refactorings are more or less tied to OO concepts, but refactoring as a whole is not. --Henrik S. Hansen 16:16, 22 October 2005 (UTC)


 * AFAIK the word 'to refactor' comes from 'to factorize', which means 'to distribute responsibilities between objects and classes'. --Pdemb 11:55, 8 January 2006 (UTC)


 * "To factorize" means to express something as a product of factors. You can factorize a number, a polynomial, a matrix. It does not imply a distribution of responsibilities among objects, although often that's what it achieves. For example, if you factorize a positive-definite matrix into the product of two triangular matrices, the responsibility of solving the associated system of linear equations is distributed to the two triangular matrices as forward substitution and backward substitution, respectively. --SergioPi 15.03, 29 July 2006.


 * In software engineering, the term 'factorization' was introduced by Peter Deutsch. I don't have the source paper right now, but I believe that he defined factorization to be the distribution of responsibilities among classes in object-oriented software system. So, refactorization should mean re-distribution of these responsibilities. Pdemb 19:46, 17 September 2006 (UTC)


 * I know this is an old discussion, but just in case anyone's still interested, the earliest use of the term "factorization" (or at least a related term) to refer to splitting responsibilities of code between different modules that I'm aware of was by Edsger Dijkstra in his 1972 Turing Award Lecture; he used the term "well-factored" on a number of occasions to describe software that adhered to good design principles. Quite clearly, he was not talking about object-oriented systems at the time. JulesH (talk) 16:01, 26 March 2009 (UTC)


 * Thank you guys! But CompSci doesn't use language as other sciences or tech spheres. One might have wished that at the very least the terminology would have be coordinated towards maths, engineering and itself, but consider CompSci "vector" vs. Math "vector", CompSci C++ "method" vs. general engineering and science "method" and CompSci "method" for that part, CompSci "factorization" vs. Math "factorization". Many sciences blather around like deaf fools, but CompSci gives me gooseflesh of horror. ... said: Rursus (bork²) 08:08, 9 July 2009 (UTC)

Which type of testing ensures that refactoring does not change the behavior of the code
Unit tests, especially automated ones in the context of test-first or test-driven development, ensure that refactoring does not change the behavior of the code. The Rod 18:53, 30 September 2005 (UTC)

I find that after refactoring is done, unit testing misses a bunch of ingegrated issues - so do a full run test after refactoring. —Preceding unsigned comment added by 212.150.171.20 (talk) 15:33, 30 November 2010 (UTC)
 * Frequent full test runs are very useful and using a continuous integration server is a convenient way of doing that. But unit tests can be run after every single refactoring step. If you follow TDD, then it's not something you do only after an hour of refactoring. If you use JUnit Max then it automatically runs your unit tests all the time, in the same way that Eclipse continuously compiles your code and adds squiggly lines as appropriate. Martijn Meijering (talk) 16:41, 30 November 2010 (UTC)

Tools
Would it be appropriate to add information about tools that support refactoring, in this page? (For example, Eclipse etc) --peterl 01:43, 4 August 2006 (UTC)


 * Code_refactoring now exists but is in need of citations. ~Kvng (talk) 15:23, 11 December 2018 (UTC)

Factorization example
The factorization example given in the text has no relevance whatsoever to the fine art of refactoring. Please refactor this paragraph. Stud. polyt. Kalle Hagen 09:44 4 December 2006

Code Smells
An article on Refactoring with no reference to Code Smells? Ideally it should right up there in the introductory para. No later than the source code section. DSParillo 16:58, 17 January 2007 (UTC)


 * Currently discussed in the first section of the article. ~Kvng (talk) 15:27, 11 December 2018 (UTC)

Spin-off proposal
I'd like this page to be only about code refactoring and the wiki-based idea of "refactoring" articles or talk pages be discussed in another page.

The intro is overly-general and imprecise:


 * Refactoring is the process of modifying a computer program or other material to improve its structure or readability, while explicitly preserving its meaning or behavior.

First off, it will change its behavior in many cases. It just won't change the results of the system tests. Only simply refactorings leave "behavior" essentially unchanged, such as the ExtractMethod facility found in Visual Studio.

The bit about preserving other text's "meaning" is misleading. When refactoring talk pages, we deliberately remove nasty-meaning language like "You're such a jerk about your edits" => "I disagree with your edits." --Uncle Ed 13:05, 9 February 2007 (UTC)

Paragraph 4 critique
The following is the start of the article's fourth paragraph: There are however certain problems to overcome in refactoring. It is a comparably new idea and there is not enough experience over its long term effects, thus the limitations which apply to it are not yet studied enough. Another problem area of refactoring are the databases – in today’s 3 – tiered architectures the business layer is strictly related to the database schema, which makes the refactoring impossible or very difficult. This paragraph is fairly weak, in both rhetoric and content.

"It is a comparably new idea and there is not enough experience over its long term effects" is oddly phrased, and I think the word "comparatively" is intended; "comparably" means similar or equivalent. The gist is inaccurate in my opinion. While the identifiable buzzword "refactoring" may only go back as far as the Fowler book, certainly the practice as informal "code clean-up" with minimal system behavior change has been going on for decades; it's a natural activity for anyone to undertake who's written a large code base and seen its entropy increase to the point that it's difficult to make changes.

The next sentence begins another rudimentary grammatical error - "area", is singular, "databases" is plural. The claim is also extremely over-simplified in its approach to architectural layers and styles. In today's N-tiered applications there is nearly always some layer of abstraction, however crude, between relational database tables and business objects, and a coherent approach to refactoring in either the business tier or the database tier would naturally look to that abstraction layer (be it ORM, JDBC/ODBC, or just a hard-coded translation module) to determine what's possible. One might even refactor to strengthen this translation layer first, and then refactor the target tier after looser coupling between layers has been achieved.

The rest of the paragraph exhibits similar problems. Hyphens are used to run sentences together; this is substandard form, appropriate for e-mail but not for a reference article. The interface difficulty described is commonplace, and the example solution offered just one of many approaches that could be taken. Overall I think the fourth paragraph should go, it's misleading and poorly written.

--Chriscorbell 21:07, 10 September 2007 (UTC)

Software Methodology
I've made a small fortune by combining automated unit testing with code refactoring. From the very start, I doubled my programming speed and reduced my "bug rate" to near zero.

But while carefully avoiding "advocacy", I'd like to explain how refactoring works together with automated tests. Or perhaps it is better explained in Extreme programming? --Uncle Ed (talk) 02:03, 23 January 2008 (UTC)


 * I think your expertise would be greatly appreciated for this article. You can have a positive view of the subject of the article, as well as you adhere to Wikipedia's policy of keeping a neutral point of view throughout. I don't believe WP:COI applies here. I personally wouldn't mind a sub-section somewhere in this article about combining refactoring with testing, but perhaps extreme programming is a better article to go for. It's really up to you. -FrankTobia (talk) 03:48, 23 January 2008 (UTC)


 * Thanks. I rarely write software any way but Test first these days, unless I'm learning something completely new and I'm just experimenting with various methods and features. But once I'm comfortable with a new language or API (like PHP and all its file and string functions), I find it better to "write a failing test first" - and then "write just enough code to make the test pass". I guess this is Kent Beck's style.


 * I'll need someone to keep me honest, lest I lapse into effusive praise of test-first development. But I'm always open to being told to "step off" when I confuse advocacy and objectivity. --Uncle Ed (talk) —Preceding comment was added at 18:09, 23 January 2008 (UTC)


 * Haha, fair enough. I'm happy to keep you honest. Of course anything with a source would be amazing: I don't have access to those most excellent refactoring books myself. Let me know on my talk page if there are any related pages I should add to my watch list. -FrankTobia (talk) 01:35, 25 January 2008 (UTC)

Refactoring to improve performance?
Including this seems a little controversial to me. Performance is not a "non-functional" aspect of a system's behaviour. And while some refactorings clearly improve performance, and some techniques to improve performance are clearly refactorings, I think performance optimization is a distinct process. And as Fowler says in Refactoring:


 * Only changes made to make the software easier to understand are refactorings. A good contrast is performance optimization.  Like refactoring, performance optimization does not usually change the behavior of a component (other than its speed); it only alters the internal structure.  However the purpose is different.

I'm going to remove improving performance from the list of reasons to refactor. JulesH (talk) 13:00, 12 March 2009 (UTC)


 * In fact, I'd go further and say that 'external non-functional' requirements are not part of refactoring, and that it is always (rather than usually) performed for one of the reasons in the list, which is in fact a superset of Fowler's definition. JulesH (talk) 13:33, 26 March 2009 (UTC)

Redesign (software)
In regards to the article Redesign (software), I read that this is apparently "Code refactoring" applicable to the entire software solution. Do you think that the redesign article needs to exist, since the difference between the two appears to be scale - and that would be subjective (i.e. how much code can we refactor before its considered redesign)? My opinion is that article should simply redirect here... thoughts? Jwoodger (talk) 00:28, 9 February 2010 (UTC)


 * Refactoring applies small behaviour preserving changes while redesign deeply changes the code and thus behaviour. I do not think this is comparable at all since the scale and impact are totally different. JnRouvignac (talk) 07:52, 24 November 2016 (UTC)

Can't see the forest for the trees
I want to emphasize, as someone who's done this for a living at times, that the refactoring process and a refactoring need to be distinguished. While a single "refactoring" ought not to make any changes in functionality, the purpose of doing a series of refactorings is to support changes in functionality. --Uncle Ed (talk) 21:51, 4 February 2011 (UTC)

Most of this article is naive
I have read this article and it makes me so depressed to observe the naivity of the programmers who have written it and many others who have subsequently commented about it without knowledgeable criticism.

It is heavily OOP oriented and mired in obfuscation and completely unsupported claims. The history section is mostly bunkem - ludicrous comments such as "In the past refactoring was avoided in development processes" displays an ignorance born of the current age. The statement "Although refactoring code has been done informally for years" demonstrates that the writer is seemingly aware that throughout software history refactoring has been performed - but then goes on to cite various papers from the quite "recent" past (1990's onwards). Programming has been taking place since the advent of computers (at least 40 years before then) and so has "refactoring" (it's just that no-one called it that until recently - Doh!). During my 40 years, I have never heard of the term "code smell" used but have seen its consequences.

In the overview section the statement "...If at any point a test fails, you undo your last small change and try again in a different way" displays how poor the training of programmers is today. What you should actually do is find out WHY something fails, not just try a different way. Sadly, however, I have discovered that this is precisely how many of this current generation of programmers "get by".

I have been programming for more than 40 years and "refactored" my own (and other people's) code from the very beginning. In fact, I constantly refactor in my head "on-the-fly" - whilst creating the code in the first place. This is how code should be built IMHO.

During coding, if you have "seen" the code before (for example three or more consecutive lines that are almost identical to ones you have written before), it is already time to consider refactoring by putting the code in a sub-routine (procedure/function) or making it into a reusable macro. Not only will this often make it quicker to write the rest of the program, but it means that the re-organized section of code is more transportable to your next program that might need it. It also means that if the code is subsequently used in another program, it has been partially tested by earlier programs. COPY and PASTE do not acheive the same result since very often the code is subtly changed after copying and probably results in more errors rather than less (the same statements are spread everywhere but have multiple versions - a definite increase in entropy).

I could go on but suffice it to say this article is very misleading. There is really no substitute for good programming and part of that is understanding the code that is being written in its context (so that when something doesn't work, you find out why, not try something new). — Preceding unsigned comment added by 86.142.127.235 (talk) 08:14, 19 August 2011 (UTC)

I concur. It's typical of many articles that spew opinion as if fact; some of which shut out or flame people with differing opinions who would make corrections. Rather than relying on factual evidence, there's a tendency to accept content generally on consensus of those most dedicated to pushing their views. I've done my part here to incrementally address the problem, but there's a long way to go and it involves a systemic problem, and there's no guarantee that positive changes won't be reverted, and many like me who do real research & development don't have the time to deal with the problem appropriately. 132.160.49.90 (talk) 19:11, 19 September 2013 (UTC)

why is the refactoring browser "so-called" ?
Which tool is older than the Refactoring Browser for VW Smalltalk from http://www.refactory.com/tools/refactoring-browser ?

A tool in VA Smalltalk ? But that team wrote eclipse ... was there one in IBM workbench ? Was there one in the Borland C++ IDE for OS/2 ?

Which Object Pascal had such a tool ?

But it is "so-called" ... because an EDITOR is not a REFACTORING tool any more than the bogey under an aircraft is an airfoil. The bogies add drag, not lift. And eclipse ?

G. Robert Shiplett 21:51, 23 July 2013 (UTC)

Reengineering is NOT refactoring
Right now Reengineering (software) points to this article. That is an error. Refactoring is not re-engineering. Re-engineering is analyzing code in legacy environments to capture designs, business rules, and/or to automate transformation of the code into a more maintainable form. Refactoring is just a normal part of the evolution of an object-oriented systems. They are related but different concepts. --MadScientistX11 (talk) 21:44, 4 January 2014 (UTC)
 * Sounds like the same thing to me, one is refactoring done during normal development, the other is long overdue refactoring that should have been done during development but wasn't. Refactoring simply means making tiny behaviour preserving transformations that make code easier to read and maintain. Martijn Meijering (talk) 22:09, 4 January 2014 (UTC)
 * No they aren't the same thing at all. They involve similar types of activities and tools but they are fundamentally different. Analyzing some old assembler or COBOL code to understand the control flow or the business logic is re-engineering. Taking a Java system and reworking it to make it more re-usable, maintainable, etc. is refactoring. In refactoring you are using the same paradigm and language for the before and after. In re-engineering you are usually working in a different paradigm (code that wasn't designed with OO) and certainly in a different programming language. --MadScientistX11 (talk) 04:21, 5 January 2014 (UTC)
 * OK, that's a substantial difference, but it seems to me refactoring could still be a large part of it. I once worked on a numerical code that had originally been written in Fortran, had been run through f2c after a couple of years and then been maintained for a couple of years. I thoroughly refactored it into more OO C++. Running something through f2c is still a behaviour preserving transformation, though the immediate end result isn't more readable, so it's a bit of a stretch to call it refactoring. Martijn Meijering (talk) 10:50, 5 January 2014 (UTC)
 * I agree they are related. Conceptually I would say Refactoring is a kind of re-engineering. We have to be careful to distinguish words from their English use and from specific meanings word have to IT people. So if we are talking just English word meaning than I agree Refactoring is a kind of re-engineering. But for at least some IT people, if you look for example in old reports on Gartner or Forrester before Y2K the phrase "software re-engineering" has a very specific meaning, it means recapturing design from legacy systems which in virtually all cases weren't designed with an Object-Oriented model. I don't think software re-engineering is in vogue now the way it was before Y2K. Anyway, it's not a big deal, I was just trying to link an article to "software re-engineering" where I was talking about recovering designs from COBOL and was surprised that the link ended up here. MadScientistX11 (talk) 14:16, 5 January 2014 (UTC)

merge with Rewrite (programming)
I am against this merge. Too often people confuse refactoring and rewrite. They are not the same thing at all.

Refactoring applies small behaviour-preserving changes, while rewrite means throwing away the existing code and restarting from scratch. Refactoring should retain bugs like for like, while a rewrite removes all the previously known bugs while it adds new unknown ones (hopefully less than before and/or less severe). The process of refactoring can be stopped at any moment and the system remains functional at all times. On its hand the rewrite must be completed before it can replace the previous code. If it is not completed then the new code is dead code or unreachable code and is immediately subject to software rot. JnRouvignac (talk) 08:03, 24 November 2016 (UTC)
 * I agree. Although, I think your definition of refactoring is just a bit too rigorous. In the real world, at least in my experience, people can sometimes make minor changes to the functionality while doing refactoring. So for example, if I find that I have some class, hexagon that is created all by itself and then I realize it should inherit from the shape class, I would consider that refactoring even though the behavior may change (e.g., by making hexagon a subclass of shape it may inherit new behaviors that weren't originally needed but now are). But thats just a quibble, in general I definitely agree rewriting or re-enginnering code is not the same as refactoring (I said similar things in an earlier thread). --MadScientistX11 (talk) 16:27, 24 November 2016 (UTC)

I am against merging this document. The concept and term of refactoring come up in computer science education and literature quite often. Some schools even have entire courses dedicated to refactoring and improving code. I myself found this Wikipedia page when searching for a concise, elegant definition of refactoring. This is a very different concept than simply rewriting a program. When rewriting code, it is expected that the behavior will change. When refactoring code, it is expected that it will not. — Preceding unsigned comment added by 174.16.14.244 (talk) 02:59, 12 March 2017 (UTC)

I oppose this merging too, and I agree with the previous comments by MadScientistX11 and "unsigned". Refactoring and rewriting are completely different things. To abuse a car analogy, it's the difference between fixing your car and buying a new one. Dserodio (talk) 22:25, 30 May 2017 (UTC)

Motivation section: debt vs payment
The metaphor of technical debt is introduced, and then refactoring is cited as a payment of technical debt.

Is such a metaphor really helping the article? Doesn't this just say that fixing poorly-organized or poorly-planned code is better than not fixing it? (It doesn't take a genius to figure that out.) :) I'm not sure why that part is kept. TooManyFingers (talk) 21:08, 8 July 2020 (UTC)
 * Fixing poorly-organized or poorly-planned code comes with a cost and with the risk of introducing new errors. In a given business environment leaving such code as it is may be entirely reasonable at a given time. – Tea2min (talk) 07:52, 9 July 2020 (UTC)

Messy change in History section
In the History section, I just removed the redundant "Griswold's Ph.D. thesis, ..." sentence. It was a very messy edit, since the contents of the refs were embedded in it, and had to be shuffled to the next sentence. The diff (changes) is a mess, but the end results are just the removal of the pesky sentence. BMJ-pdx (talk) 08:39, 12 May 2024 (UTC)