http://www.spinellis.gr/pubs/jrnl/2005-IEEESW-TotT/html/v29n1.html
This is an HTML rendering of a working paper draft that led to a publication. The publication should always be cited in preference to this draft using the following reference:

This document is also available in PDF format.

The document's metadata is available in BibTeX format.

Find the publication on Google Scholar

This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.

Diomidis Spinellis Publications


© 2012 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

Refactoring on the Cheap

Diomidis Spinellis

The refactorings that a good integrated development environment can perform are impressive. Yet, there are many reasons to master some cheap-and-cheerful alternative approaches. First, there will always be refactorings that your IDE won’t support. Also, although your IDE might offer excellent refactoring support for some programming languages, it could fall short on others. Modern projects increasingly mix and match implementation languages, and switching to a specialized IDE for each language is burdensome and inefficient. Finally, IDE-provided refactorings resemble an intellectual straightjacket. If you only know how to use the ready-made refactorings, you’ll miss out on opportunities for other code improvements.

In this column, I describe how you can harness the sophistication of your editor and the power of command-line tools to perform many simple refactorings on your own. As a bonus, you’ll see that you can write and run most of them in less time than is required to fire up your favorite IDE.

Within Files

The basic tool for performing a refactoring within a file is the editor’s substitution command used in conjunction with regular expressions (see “Dear Editor,” IEEE Software, vol. 22, no. 2, 2005, pp. 14–15). In this section’s examples, I’m using the common Unix substitution syntax s/old/new/ and vi’s regular expression syntax; other editors offer similar functionality. Splitting a simple method call into two can be as easy as typing

s/getX()/getX().getY()/g

However, if the original method has arguments that must now be passed by the second method, you need to get creative and include the bracket in the substitution pattern in order to move the arguments to getY:

s/getX(/getX().getY(/g

Regular expressions aren’t powerful enough to parse a typical programming language, so you’ll often have to resort to tricks like this one to handle brackets and braces. Another useful regular expression substitution technique involves capturing part of the old pattern and reusing it in the new string. For instance, if you want to change calls to the function raiseRating, where its first argument is a variable representing an object, into method calls, you might give a command like

s/raiseRating(\([^,]*\),/\1.raiseRating(/g

Here, the text within the first \(\) pair will get stored in a regular expression variable, which you can then employ in the substitution pattern using \1. The [^,]* idiom inside the brackets matches everything up to the first comma. Again, this isn’t bullet-proof—some legitimate expressions might contain a comma—but it works 95 percent of the time (or does 95 percent of the job).

Capitalizing on the regularity of the text you process can save a lot of pain. The following will transform named HTML headers into list-item hyperlinks:

s,<h2><a name="\([^>]*\)>\(.*\)</a><\/h2>,<li> <a href="\#\1>\2<\/a><\/li>,

I use it regularly to keep the questions and answers in my FAQ documents in sync. It works only because I write the headings on a single line, but the nearest robust alternative (XSLT) would require prohibitively more work.

Across Files

Quite often, you’ll want to apply substitution commands not to a single file but to all files in a directory or throughout your project. The canonical way to do this under Unix (and also Mac OS X and Microsoft Windows with an installation of Cygwin) is to use the stream editor sed. This can take as an argument some editing commands and apply them to the files you specify. Modern versions of sed can perform a substitution in place. Thus, with a command like

sed -i -e 's/Employee/Person/g' proc*.scala

you can change Employee into Person on all Scala files in the current directory whose name starts with proc. You can precede the substitution command s by one or two regular expressions to specify the lines or a range of lines on which the command will apply. If you wanted to change getWidget into getWidgetReference in all C++ lines containing const_iterator, you’d run

sed -i -e '/const_iterator/s/getWidget/getWidgetReference/g' *.cpp

Moreover, if you wanted to change bProxy to buildProxy only within javadoc block comments in all Java files in the current directory, you’d run

sed -;i -e '/\/\*\*$/,/\*\/$/s/bProxy/buildProxy/g' *.java</p>

The Perl and Ruby scripting languages also offer in-place substitution functionality through command-line invocation options, and their expression evaluators allow you to perform more sophisticated processing. The following command changes the first argument of setFlags from decimal to hexadecimal in all the C and C++ files in the current directory:

perl -pi.bak -e 's/setFlags\((\d+)/sprintf("setFlags(0x%x", $1)/ge' *.c *.cpp

Large projects typically reside in multiple directories. The Unix find and xargs commands are the building blocks for applying commands to large hierarchies. Find will print a list of files that match the criteria you specify. You then pass those through a pipeline to xargs, which will invoke the command you specify in batches of as many files as the operating system allows. The typical invocation for a global substitution through all the project’s files is something like

find project_directory -type f -print0 |
xargs -0 sed -i -e 's/old/new/g'

(The -print0 and -0 options allow you to process file names with embedded spaces.) You can restrict the files onto which you apply the substitution by passing a -name argument to find; here is how you would change intr_handle_t into pci_intr_handle_t only in files with names starting with pci_:

find /usr/src/sys -type f -name pci_*.c |
xargs sed -i -e 's/intr_handle_t/pci_intr_handle_t/g'

On File Paths

Some refactorings involve changing file names or moving files around. You can easily accomplish this by using find to list the corresponding files and sed to craft the text of a command that will accomplish the action you want. You then pipe the generated commands into the shell (sh), which will execute them as if you typed them interactively. Depending on your setup, you’ll want the commands you create to either manipulate the files directly or to call up your version control system to perform the corresponding action. The following command will add an fs prefix to all file names residing in directories ending in fs:

find . -type f |
sed -n -e 's/\(.*\)fs\/\(.*\)/mv "\1fs\/\2" "\1fs\/fs_\2"/p' |
sh

It acts by generating an mv (rename) command for all matching path names. (As invoked, sed will read lines from its standard input and only print the lines for which the substitution succeeds.) If you wanted to issue commands to the Subversion version control system, you’d specify svn rename instead of mv. With similar commands, you can move files around the project’s directory hierarchy or remove files that are no longer needed.

Profiting

There are many habits that can increase your effectiveness in the tasks I’ve outlined here. First, as befits someone who works on the cheap, you must be stingy and lazy. With the ability to easily undo changes either within your editor or through your version control system, don’t sweat coming up with the perfect regular expression that will succeed in every imaginable case. Try out simple commands that could conceivably work and see if they’re good enough. If not, the compiler or a visual inspection should catch the errors, and you can try a slightly more sophisticated version of the command. In particular, the regular expression repetition operators * and + aren’t matching as greedily as you might fear. If you specify a well-chosen string after them, the regular expression engine will backtrack and have the repetition operator match only the part needed for the whole expression to succeed. Many commands in this column utilize this property.

Furthermore, be ready to tolerate a few missed or extraneous changes. If you can easily locate them, correcting them by hand is often faster than crafting the exact command that will work on all cases. Remember, your goal is to be productive; you’re writing throwaway code that no one else will ever see again, so you might as well enjoy employing a few shortcuts. Be opportunistic in the commands you craft, taking advantage of style, coding, and API conventions to achieve your results. Thus, if you can narrow down on the correct method to change because it happens to always appear as an argument to another method, don’t be shy about specifying that in your command.

Regular expressions are nifty, but they aren’t sufficiently powerful for all the tasks you’ll face. Also, the commands can quickly become overwhelmingly complex. You can solve both problems by running the substitutions step by step. Dodge a sequence that confuses the regular expression matcher by converting it into something innocuous, say @v@, running the substitution, and then converting it back to its original form. Break complex substitutions into small steps, verifying the results in each step.

Finally, you can simplify your life if you write code in a way that can aid the refactorings I’ve described here. Be consistent in naming your identifiers, formatting your code, and the way you split it across lines and files. There are many good reasons for writing good code: being able to refactor it on the cheap is just the icing on the cake.