The editors Emacs and vi are used to make changes in files
interactively. Yet sometimes you want to make particular changes in
many files or repeat the same editing again in the future, or to do
editing from the command line of Unix and pipe the results to yet
other commands. For these purposes, which are not as esoteric as they
may appear at first, sed is indispensable. For example,
suppose your files contain data of the form 1.12D+33, but your
graphics program doesn't like the D for an exponent. Or suppose
you want to convert your paper from nroff to LaTeX. Or perhaps
you want to delete all lines containing the string copyright (c)
by IBM from a group of files.
These operations of
substitution and deletion make sed a must to learn (you'll
have time in your jail cell).
The sed command, like awk, which we describe next, are used as a programming language to create sed scripts. These scripts are useful for complicated or repetitive debugging chores. Quite often, sed is used on the command line with simple options. These options are of some historical interest since they are based on ed, the old Unix line editor now mercifully extinct and forgotten. When used from the command line, only a subset of the full sed options is needed.
In our first example we start with the data file graph.dat containing x-y pairs:
4.05D-09 11.35D+03 File graph.dat 9.73D-09 9.89D+03 10.8D-09 9.63D+03
We must change the D to an e for our graphics program.
Rather than do this by hand with an editor, we use sed,
which is quicker and gives us a tool to edit future files:
% sed -e "s/D/e/g" graph.dat | xplot Edit and pass directly to xplot.
Notice that, if we had not added the
g option for global changes, only the first D per line
would have been changed:
% sed -e "s/D/e/" graph.dat Leave off the g and use standard output. 4.05e-09 11.35D+03 9.73e-09 9.89D+03 10.8e-09 9.63D+03
In the next example we first delete all comments from a Fortran file,
that is, all lines starting with a c or C, and then use
the y option to translate all uppercase letters into lowercase
(the same function as tr provides):
% sed -e "/^[c,C]/d" prog.f >prog2.f Strip comments and put into prog2.f. % sed -e "y/abcdefghijklmnopqrstuvwxyz/ABCDEFGHIJKLMNOPQRSTUVWXYZ/" prog.f > prog2.f