It is commonplace for a developer to find himself working at any given time on multiple different, distinct changes to a particular bit of source code. This isn't necessarily due to poor planning or some form of digital masochism. A software engineer often spots bugs in his peripheral vision while working on some nearby chunk of source code. Or perhaps he's halfway through some large change when he realizes the solution he's working on is best committed as several smaller logical units. Often, these logical units aren't nicely contained in some module, safely separated from other changes. The units might overlap, modifying different files in the same module, or even modifying different lines in the same file.
Developers can employ various work methodologies to keep these logical changes organized. Some use separate working copies of the same repository to hold each individual change in progress. Others might choose to create short-lived feature branches in the repository and use a single working copy that is constantly switched to point to one such branch or another. Still others use diff and patch tools to back up and restore uncommitted changes to and from patch files associated with each change. Each of these methods has its pros and cons, and to a large degree, the details of the changes being made heavily influence the methodology used to distinguish them.
Subversion provides a changelists feature that adds yet another method to the mix. Changelists are basically arbitrary labels (currently at most one per file) applied to working copy files for the express purpose of associating multiple files together. Users of many of Google's software offerings are familiar with this concept already. For example, Gmail doesn't provide the traditional folders-based email organization mechanism. In Gmail, you apply arbitrary labels to emails, and multiple emails can be said to be part of the same group if they happen to share a particular label. Viewing only a group of similarly labeled emails then becomes a simple user interface trick. Many other Web 2.0 sites have similar mechanisms—consider the “tags” used by sites such as YouTube and Flickr, “categories” applied to blog posts, and so on. Folks understand today that organization of data is critical, but that how that data is organized needs to be a flexible concept. The old files-and-folders paradigm is too rigid for some applications.
Subversion's changelist support allows you to create changelists by applying labels to files you want to be associated with that changelist, remove those labels, and limit the scope of the files on which its subcommands operate to only those bearing a particular label. In this section, we'll look in detail at how to do these things.
You can create, modify, and delete changelists using the svn changelist command. More accurately, you use this command to set or unset the changelist association of a particular working copy file. A changelist is effectively created the first time you label a file with that changelist; it is deleted when you remove that label from the last file that had it. Let's examine a usage scenario that demonstrates these concepts.
Harry is fixing some bugs in the calculator application's mathematics logic. His work leads him to change a couple of files:
$ svn status M integer.c M mathops.c $
While testing his bug fix, Harry notices that his changes
bring to light a tangentially related bug in the user
interface logic found in
decides that he'll go ahead and fix that bug, too, as a
separate commit from his math fixes. Now, in a small working
copy with only a handful of files and few logical changes,
Harry can probably keep his two logical change groupings
mentally organized without any problem. But today he's going
to use Subversion's changelists feature as a special favor to
the authors of this book.
Harry first creates a changelist and associates with it the two files he's already changed. He does this by using the svn changelist command to assign the same arbitrary changelist name to those files:
$ svn changelist math-fixes integer.c mathops.c A [math-fixes] integer.c A [math-fixes] mathops.c $ svn status --- Changelist 'math-fixes': M integer.c M mathops.c $
As you can see, the output of svn status reflects this new grouping.
Harry now sets off to fix the secondary UI problem. Since he knows which file he'll be changing, he assigns that path to a changelist, too. Unfortunately, Harry carelessly assigns this third file to the same changelist as the previous two files:
$ svn changelist math-fixes button.c A [math-fixes] button.c $ svn status --- Changelist 'math-fixes': button.c M integer.c M mathops.c $
$ svn changelist --remove button.c D [math-fixes] button.c $ svn changelist ui-fix button.c A [ui-fix] button.c $
$ svn changelist ui-fix button.c D [math-fixes] button.c A [ui-fix] button.c $ svn status --- Changelist 'ui-fix': button.c --- Changelist 'math-fixes': M integer.c M mathops.c $
Harry now has two distinct changelists present in his
working copy, and svn status will group its
output according to these changelist determinations. Notice
that even though Harry hasn't yet modified
button.c, it still shows up in the output
of svn status as interesting because it has
a changelist assignment. Changelists can be added to and
removed from files at any time, regardless of whether they
contain local modifications.
Harry now fixes the user interface problem in
$ svn status --- Changelist 'ui-fix': M button.c --- Changelist 'math-fixes': M integer.c M mathops.c $
The visual grouping that Harry sees in the output of
svn status as shown in our previous section
is nice, but not entirely useful. The
status command is but one of many
operations that he might wish to perform on his working copy.
Fortunately, many of Subversion's other operations understand
how to operate on changelists via the use of the
When provided with a
Subversion commands will limit the scope of their operation to
only those files to which a particular changelist name is
assigned. If Harry now wants to see the actual changes he's
made to the files in his
changelist, he could explicitly list only
the files that make up that changelist on the svn
diff command line.
$ svn diff integer.c mathops.c Index: integer.c =================================================================== --- integer.c (revision 1157) +++ integer.c (working copy) … Index: mathops.c =================================================================== --- mathops.c (revision 1157) +++ mathops.c (working copy) … $
That works okay for a few files, but what if Harry's change touched 20 or 30 files? That would be an annoyingly long list of explicitly named files. Now that he's using changelists, though, Harry can avoid explicitly listing the set of files in his changelist from now on, and instead provide just the changelist name:
$ svn diff --changelist math-fixes Index: integer.c =================================================================== --- integer.c (revision 1157) +++ integer.c (working copy) … Index: mathops.c =================================================================== --- mathops.c (revision 1157) +++ mathops.c (working copy) … $
And when it's time to commit, Harry can again use the
--changelist option to limit the scope of the
commit to files in a certain changelist. He might commit his
user interface fix by doing the following:
$ svn commit -m "Fix a UI bug found while working on math logic." \ --changelist ui-fix Sending button.c Transmitting file data . Committed revision 1158. $
In fact, the svn commit command
provides a second changelists-related option:
--keep-changelists. Normally, changelist
assignments are removed from files after they are committed.
--keep-changelists is provided,
Subversion will leave the changelist assignment on the
committed (and now unmodified) files. In any case, committing
files assigned to one changelist leaves other changelists
$ svn status --- Changelist 'math-fixes': M integer.c M mathops.c $
--changelist option acts only as a
filter for Subversion command targets, and will not add
targets to an operation. For example, on a commit operation
svn commit /path/to/dir, the
target is the directory
and its children (to infinite depth). If you then add a
changelist specifier to that command, only those files in
/path/to/dir that are
assigned that changelist name will be considered as targets
of the commit—the commit will not include files
located elsewhere (such as in
/path/to/another-dir), regardless of
their changelist assignment, even if they are part of the
same working copy as the operation's target(s).
Even the svn changelist command accepts
--changelist option. This allows you to
quickly and easily rename or remove a changelist:
$ svn changelist math-bugs --changelist math-fixes --depth infinity . D [math-fixes] integer.c A [math-bugs] integer.c D [math-fixes] mathops.c A [math-bugs] mathops.c $ svn changelist --remove --changelist math-bugs --depth infinity . D [math-bugs] integer.c D [math-bugs] mathops.c $
Finally, you can specify multiple instances of the
--changelist option on a single command
line. Doing so limits the operation you are performing to
files found in any of the specified changesets.
Subversion's changelist feature is a handy tool for grouping working copy files, but it does have a few limitations. Changelists are artifacts of a particular working copy, which means that changelist assignments cannot be propagated to the repository or otherwise shared with other users. Changelists can be assigned only to files—Subversion doesn't currently support the use of changelists with directories. Finally, you can have at most one changelist assignment on a given working copy file. Here is where the blog post category and photo service tag analogies break down—if you find yourself needing to assign a file to multiple changelists, you're out of luck.