#!/bin/sh usage="Usage: die [-x exitstatus] [-k killsig]" while : do case $# in 0) break;; *) case "$1" in -x) case $# in 1) echo "die: -x requires exit status argument $usage" >&2; exit 1;; esac stat="$2"; ; ;; -k) case $# in 1) echo "die: -k requires signal number argument $usage" >&2; exit 1;; esac killsig="$2"; ; ;; # -v) verbose=1; ;; # example of a flag that takes # no arguments -*) echo "Illegal flag $1 $usage" >&2; exit 1;; *) break;; # filename arguments esac esac done case "$stat" in '') case "$killsig" in '') echo "die: You must have one of -x or -k $usage" >&2; exit 1;; *) kill -"$killsig" $$;; esac ;; *) case "$killsig" in '') exit $stat;; *) echo "die: You must have one of -x or -k $usage" >&2; exit 1;; esac ;; esac exit # example of handling filename arguments for fn do : # commands that process $fn done
$ cd ../public/bin $ mkdir ../bin.x $ for i in * > do > < $i sed 's,/home/bsy/bin,/home/solaris/ieng9/cs80w/public/bin,g' > ../bin.x/$i > chmod +x ../bin.x/$i > done $interactively. After checking things out, I removed the original bin directory and renamed bin.x to bin. (BTW, what's the > prompt?)
The sed command used here is the Stream Editor. Its basic, implicit loop is to read in a line of input, apply whatever editing commands are applicable, and then print out the possibly modified input. In the above example, the editing command is to substitute /home/solaris/ieng9/cs80w/public/bin for occurrences of /home/bsy/bin in the input line. The g at the end specifies that this substitution is to be done globally over all occurrences in that line; without the g, only the first occurrence of /home/bsy/bin would be replaced.
sed's default action of printing can be suppressed with the -n flag, whereupon output is performed only when explicitly told to via the editing command p. You should skim through the sed manual page -- you can ignore the complexities of the hold space versus the pattern space for now, but you should know the s, p, d, a, i, c, and q commands.
I gave the following examples, based on the fact that an empty line always separates the email headers and the email body (RFC-822):
Print out email headers:
$ sed -n '1,/^$/p' < email.fileand automatically adding a salutation:
$ sed '/^$/a\ Hello' < email.fileNote: this example doesn't work quite right. Why? (Hint: what might the body look like / contain?)
BEGIN {
cmds; /* A */
}
/pattern/ {
cmds; /* B */
}
NR == 4 {
cmds; /* C */
}
END {
cmds; /* D */
}
The A cmds controlled by the BEGIN pseudo pattern is
run before any input are processed; as you may have guessed, the D
commands are run after all input are processed. The B and C commands
are run when their guards are satisfied: for B, that means that the
regular expression pattern matches the input line, and for C,
that means the boolean expression is true, i.e., the number of input
records so far equals 4 (at the 4th line).
The control flow constructs of awk are very similar to those of C: there are if / else statements, while statements, and for loops just like in C. The data types are different though: objects are not really typed, and the awk program will dynamically convert a string into a number and vice versa:
bash$ awk 'END { print "5" + 2; }' < /dev/null
7
bash$
One area in which awk and C are very different is the kinds of aggregate types available. awk has no struct or union, but has instead an associative array. Instead of the indices being only numbers, they may be arbitrary strings. For example:
#!/software/common/gnu/bin/bash
#
# take input a file containing two fields per line: login name and
# number of points
#
awk '
{
grade[$1] = $2
}
END {
for (student in grade) {
print student, grade[student]
sum += grade[student];
nstud++;
}
print "Mean grade " sum / nstud
}' < ${1:-"/home/solaris/ieng9/cs80w/cs80w/grade.db"}
uses an associative array to store the login/points tuple.
bsy@cse.ucsd.edu, last updated