Merging and More with diff3

Learn how to use diff3 to compare and merge files.

As you’ve seen over the course of the last two “Power Tools” columns, one reason to use the command-line GNU diffutils is that, compared to windowed difference programs, diffutils can give you much more control over how variations are found and displayed. Another compelling argument is that some diffutils can create new files from merged versions of other files — either interactively, or by using the output of one utility as input to another. You can also use diff --from-file to compare one file to two or more other files. patch can also merge files automatically.

The diff3 utility compares three files, and can also merge the changes from two other files into a common ancestor file. Let’s dig in and see how it works.

What’s different about diff3?

Most diffutils compare two files, but diff3 compares three. It’s often used to compare two differing files that are both based on a common ancestor. For instance, if two programmers copy and then modify the same program file, diff3 could show both sets of changes.

diff3 can also merge changes from two files into a third file — typically in a case like the one we just mentioned. The options -m and --merge do this.

The -e and --ed options create an ed script that lets you merge changes into a separate file. ed scripts were common before patch came on the scene, and remain effective. However, because ed scripts generally use absolute line numbers, such scripts can cause corruption or fail completely when used to edit files that are somewhat different than the file the script was based upon. patch is far more forgiving.

Showing differences with diff3

The diff3 output format is different than the formats of diff. Let’s start with the short description (edited slightly) from the Unix Seventh Edition man page, then look at a few examples.

diff3 [ -ex3 ] file1 file2 file3

diff3 compares three versions of a file, and publishes disagreeing ranges of text flagged with these codes:

====              all three files differ

====1           file1 is different

====2               file2 is different

====3               file3 is different

The type of change needed to convert a given range of one file  to another is indicated in one of two ways:

f : n1 a
Text is to be appended after line number n1 in file f, where f is 1, 2, or 3.

f : n1 , n2 c
Text is to be changed in the range line n1 to line n2.  If there's just one line, only n1 is given.

Now for examples. Let’s look at three sets of three files. In each set, one line in one of the three files is different.

In the first example, shown in Table One, the first line of the first file (file 1a) is different:

Table One: The first file of three differs

File 1a File 1b File 1c
line 1 file a line 1 line 1
line 2 line 2 line 2

If you pass those three filename arguments to diff3 (using the shell wildcard 1? to expand the three names sorted alphanumerically), the output is:

$ diff3 1?
   line 1 file a
   line 1

The first output line, starting with equal signs and a 1, says, “Here’s a difference in the first file.” Next, is the difference. In the first file (1:), line 1 contains (1c) the text “line 1 file a”. In both the second and third files (2: and 3:), line 1 contains (1c) the text “line 1”. Because the text is the same in both the second and third files, the codes for both of those files (2:1c and 3:1c) are shown together, and the line is shown only once afterward.

For the next example (Table Two), the first line of the second file, file 2b, is different.

Table Two: The second file differs from the other two

File 2a File 2b File 2c
line 1 line 1 file b line 1
line 2 line 2 line 2

And here’s the result:

$ diff3 2?
   line 1
   line 1 file b

Again, the 2 after the hunk delimiter ====2 means, “This change is in the second file.” You can see that both the first file (1:1c) and third file (3:1c) have a first line that reads “line 1”. In the second file, though (2:1c), line 1 reads “line 1 file b”.

You’re probably seeing the pattern now. But, to be complete, Table Three shows a change in the third file 3c.

Table Three: An example where the third file differs

File 3a File 3b File 3c
line 1 line 1 line 1 file c
line 2 line 2 line 2
$ diff3 3?
   line 1
   line 1 file c

The other type of change that diff3 indicates is an appended

  • line.Table Four shows three example files; the second file has a

    new line at the end:

    Table Four: A third line in second file

    File 4a File 4b File 4c
    line 1 line 1 line 1
    line 2 line 2 line 2
    line 3 file b

    Here’s the diff3 output:

    $ diff3 4?
       line 3 file b

    It shows that in file1 (1:2a) and file 3 (3:2a), the hunk contains no lines. (In other words, the new text isn’t inserted into these two files.) But in file 2, the new line 3 is changed (2:3c) to read “line 3 file b”.

    If you’d like more examples, see the online manual page. (It’s worth reading, anyway.)

    Merging with diff3

    One advantage of plain diff over diff3 is that diff’s output can be used as input to patch — to merge the same changes into other files, at other times. diff3 output, on the other hand, is for you to look at.

    When you’re comparing two different files to a third, as diff3 does, changes can conflict. For instance, line 2 may be different in all three files. In those cases, you have to decide which change to keep. If there are no conflicts, though, diff3 can merge two files automatically, and writes the result to its standard output. (To merge more than three files, merge the first three, then re-merge the next into the result from the first three, and so on.)

    When merging files, the order of the three command-line arguments is important. The diffutils manual gives an example: Both you and someone else have modified an ancestor file named older. Your version is named mine and the other person’s is named yours. You want to merge into mine the changes that would turn older into yours. In other words, the other person’s changes should be found, then those changes are merged with your changes. The command-line arguments should be in this order:

    $ diff3 mine older yours

    Let’s see examples of both a clean merge and a merge with a conflict, reusing the files from Table Three.

    The yours file (the third argument, file 3c) has been modified, but the mine file hasn’t. It’s easy to merge those changes because only one of the two “newer” files has changed:

    $ diff3 -m 3?
    line 1 file c
    line 2

    In the same way, in the situation shown in Table One, only mine has been edited, so merging is easy:

    $ diff3 -m 1?
    line 1 file a
    line 2

    The problem comes when both mine and yours are different than the original version. When older is different from both newer files, diff3 flags that, by default. The conflict is shown between a pair of <<<<<<< and >>>>>>> markers; other lines without conflicts are output as-is. Here’s the conflict in this case:

    $ diff3 -m 2?
    <<<<<<< 2b
    line 1 file b
    line 1
    >>>>>>> 2c
    line 2

    That “conflict” may not be a problem for you, though. Here, both you and the other person made the same change to the older file (changing “line 1 file b” to “line 1”). To handle cases like these, diff3 gives you other options to control which output is shown:

  • -x or --overlap-only outputs only the overlapping changes (conflicts where all three files are different)

  • -3 or --easy-only outputs only non-overlapping changes

  • -A or --show-all outputs all changes.

    See the documentation for more information and examples.

    Merging with patch

    Patch is such a popular and easy-to-use tool that it doesn’t need much introduction here. Briefly, though, it’s an automated way to apply diff output to other files. You create a diff listing that shows the changes between the current version of one or more files and the new version (s). You then send that diff listing to others. In turn, they feed the listing to patch, and it makes those same changes in the remote files. (If there’s more than one file, the directory structure should be similar at the source and destination ends.)

    Patch does merging automatically and fairly intelligently. Here are some good tips from the diffutils manual:

    Running the bash command export SIMPLE_BACKUP_SUFFIX=".old" sets that suffix for all GNU programs that make backups. To set the suffix only for patch, you can make an alias that defines the environment variable per-command. (This bash feature works for any Linux utility, not just patch.) For example, if the patch binary is at /usr/bin/patch, try.

    $ alias patch='SIMPLE_BACKUP_SUFFIX=".old" /usr/bin/patch'
    $ patch myfile < patchfile
    patching file myfile
    $ ls myfile*
    myfile myfile.old

    To make a patch file, don’t use diff3. Try diff -aNru instead. See “Avoiding Common Mistakes” in the diffutils manual.

    And Even More...

    There’s a lot more to discover about diffutils in these three columns. The detailed and very helpful documentation for diffutils can be found online.

  • Jerry Peek is a freelance writer and instructor who has used Unix and Linux for more than 25 years. He's happy to hear from readers; see

    [Read previous article] [Read next article]
    [Read Jerry’s other Linux Magazine articles]