L4. Shell scripting
Learn / L4
Goal: write three scripts, each with one idea: count tests
its arguments, each loops and matches names, mygrep drives
ed from a here-document and cleans up after itself.
What you need
L1: a script is a text file that starts with
#!/bin/sh. Nothing from the assembly lessons.
maux's sh is the Bourne shell of Research Unix Version 7 with a few
later conveniences, and it has no shell functions: V7 had none either. That fixes
what these scripts can use.
Writing them also turned up shell behaviour that is wrong today. These are defects, reported to the release and not yet fixed; the scripts below keep clear of them, and so should yours:
- A pipeline of more than three stages is refused with
sh: syntax error. That one is a documented limit, not a defect (the manual says "at most three stages"). - A
$( )inside anif,fororwhilebody runs more than once, even in a branch that is not taken. A multi-linefortyped at the prompt does the same. Its side effects happen every time, and a variable the body sets is still empty the first time. - Positional parameters (
$1,$#) are not visible inside$( ). That is how this shell is built today; POSIX says they should be visible. - A three-stage pipeline of bare command names, such as
ls | tr a-z A-Z | sort, fails inside a script file withsh: command not foundprinted three times. It works at the prompt, and in a script given absolute command paths, assh < file, or read with. file. - When the last stage of a pipeline cannot be started,
$?is 0.
The files
count: arguments and tests
$# is the number of arguments, $1 the first,
[ -f ... ] tests for an ordinary file, and exit 1 ends the
script with a failure status.
examples/scripts/count — on the disk as /usr/src/examples/SCRIPTS/count#!/bin/sh
# count FILE -- say how many lines FILE has
if [ $# -ne 1 ]
then
echo usage: count file
exit 1
elif [ -f "$1" ]
then
wc -l "$1"
else
echo count: $1: no such file
exit 1
fi
each: loops and case
for f in * loops over the file names the shell expands;
case matches each against patterns; while read line takes
lines from standard input, here a file given after done <. The
$( ) on the last line runs a command and puts its output in the line.
examples/scripts/each — on the disk as /usr/src/examples/SCRIPTS/each#!/bin/sh
# each DIR -- describe every file in DIR
cd $1
for f in *
do
case "$f" in
*.s) kind=source ;;
*.o) kind=object ;;
*) kind=other ;;
esac
echo $f is $kind
done
ls > /tmp/each.$$
while read line
do
echo listed: $line
done < /tmp/each.$$
rm /tmp/each.$$
echo $(ls | wc -l) files
mygrep: a here-document, trap and set -e
Everything between <<EOF and EOF is the standard input
of ed. The name grep is the ed command
g/re/p: global, regular expression, print.
examples/scripts/mygrep — on the disk as /usr/src/examples/SCRIPTS/mygrep#!/bin/sh
# mygrep PATTERN FILE -- grep, written in ed
set -e
tmp=/tmp/mygrep.$$
trap 'rm -f $tmp; exit 1' 2
ed > $tmp <<EOF
r $2
g/$1/p
q
EOF
cat $tmp
rm -f $tmp
The session
Each script is typed with cat > name and ended with
^D (control-D, end of file). cmp then compares what you
typed with the copy on the disk; if it prints a line, you mistyped.
$ mkdir /tmp/l4 $ cd /tmp/l4 $ cat > count [ type the file shown above, a line at a time, then press ^D ] $ cmp count /usr/src/examples/SCRIPTS/count $ cat > each [ type the file shown above, a line at a time, then press ^D ] $ cmp each /usr/src/examples/SCRIPTS/each $ cat > mygrep [ type the file shown above, a line at a time, then press ^D ] $ cmp mygrep /usr/src/examples/SCRIPTS/mygrep $ ./count each 19 each $ ./count usage: count file $ echo $? 1 $ ./count nosuch count: nosuch: no such file $ echo $? 1 $ ./each /usr/src/examples/HELLO hello.s is source Makefile is other listed: hello.s listed: Makefile 2 files $ ./mygrep regular /usr/man/GREP for lines matching the regular expression. Normally, regular expression. The matcher is the same basic regular-expression engine $ grep regular /usr/man/GREP for lines matching the regular expression. Normally, regular expression. The matcher is the same basic regular-expression engine $ ls /tmp | grep mygrep
What just happened
./count with no argument took the first branch and printed the usage
line; with a file that does not exist it took the last. The shell's
$? holds the status of the last command, and the session prints it
with echo $?: 0 means success, anything else failure. That
is the entire contract between a program and a script.
each takes a directory as its operand and changes to it
(cd $1) before it looks at anything; the change belongs to the script's
own process, so your shell is still where you left it. It writes the listing to a
file under /tmp named with $$, the shell's own process
number, so two copies never share one, and reads it back with
while read line ... done < file. The word 2 files on
its last line came from $(ls | wc -l).
mygrep printed the same lines, in the same order, as the real
grep. set -e makes the script stop at the first command that
fails; trap 'rm -f $tmp; exit 1' 2 says "if signal 2 arrives (^C), delete
the temporary file and stop". The last line of the session shows that nothing named
mygrep was left in /tmp.
Try next
Write backup: a script that copies the files you changed to a second
volume, then archives them with nulib2. Give it a usage message and
check its status the way count does.