Comparing Three Different Versions with diff3

For each set of differences, diff3 displays a row of equal signs (====) followed by 1, 2, or 3, indicating which file is different; if no number is specified, then all three files differ. Then, using ed-like notation (), the differences are described for each file.

$ diff3 test1 test2 test3 ====3 1:1c 2:1c apples 3:0a ==== 1:3c walnuts 2:3c grapes 3:2,3c walnuts chestnuts

With the output of diff3, it is easy to keep track of which file is which; however, the prescription given is a little harder to decipher. To bring these files into agreement, the first range of text (after ====3) shows that you would have to add apples at the beginning of the third file (:0a). The second range tells you to change line 3 of the second file to line 3 of the first file; and change lines 2 and 3 of the third file, effectively dropping the last line.

The diff3 command also has a -e option for creating an editing script for ed. It doesn't work quite the way you might think. Basically, it creates a script for building the first file from the second and third files.

$ diff3 -e test1 test2 test3 3c walnuts chestnuts . 1d . w q

If you reverse the second and third files, a different script is produced:

$ diff3 -e test1 test3 test2 3c grapes . w q

As you might guess, this is basically the same output as doing a diff on the first and third files. (The only difference in the output is the result of a rather errant inconsistency between diff and diff3. The System V version of diff3 produces an ed script that ends with the commands that save the edited version of the file. The Berkeley diff3, and both versions of diff, require that you supply the w and q. We show the System V version of diff3 here.)

- DD from UNIX Text Processing, Hayden Books, 1987, Chapter 11