Introduction to Shell Programming

what is shell programming?

about cygwin
review of basic UNIXTM
pipelines of commands
about shell scripts
some new commands
variables
parameters and shift
command substitution
cut
if-then-else-fi
for-in-do-done
sed
a nal example
gotchas
exercises
references

conducted by
Mike Kupferschmid
Scientific Programming Consultant
VCC 322, x6558, kupfem@rpi.edu
What is Shell Programming?

putting UNIXTM commands in a le

almost always special-purpose code
often one-time code
seldom used where speed is important
often used to manipulate les
 About cygwin

a good but not perfect emulation of unix

included in standard RPI laptop image
to download (long) go to www.cygwin.com
if you want real unix get Linux (site licensed)
it is possible to dual-boot Linux with Windows
ask at the Help Desk
watch for an ACM installfest
starting cygwin
start All Programs Cygwin XTerm
opens unix window with command prompt
$
RCShome is a link to your RCS home directory
to print a le from cygwin
open the le with notepad filename
use the notepad print function
to cut and paste you need a 3-button mouse
cygwin names executables a.exe
 Review of Basic UNIXTM
familiar commands often useful in shell scripts
cat concatenate les
cp copy a le
date print the date and time
grep scan for a string
head show rst lines of a le
tail show last lines of a le
mv move or rename a le
rm -f remove les (silently)
wc count lines, words, characters
wc output format varies between systems
path names of les and directories
schedule relative
/home/37/jones/schedule absolute
wild cards in lenames
* matches zero or more characters
? matches exactly 1 character
redirection
> redirects std-out to a le
>> appends std-out to a le
< redirects std-in from a le
 Pipelines of Commands
send std-out of one command to std-in of another

look e
shows spelling words that begin with e
look e | more
displays the words one page at a time

often use echo to feed a pipeline

echo "count me" | wc
prints 1 2 9
echo * | wc -w
counts les in current directory
 About Shell Scripts

type shell program text in a le using an editor:

#! /bin/sh
# this is a comment
body of program
to continue a line append \
this is the rest of the continued line
exit 0

chmod +x scriptfile make the le executable

not needed in cygwin

scriptfile execute the program

sh -v scriptfile print input lines as read
sh -x scriptfile print commands as executed

shell programs often use temporary les in /tmp

and send unwanted outputs to /dev/null
 Some New Commands
Useful in Shell Programs

basename extract le name from path name

cmp -s compare les (silently)
cut extract selected parts of a line
expr evaluate an expression
mail send email (not in cygwin)
sed -e stream editor
sleep suspend execution for given time
tr translate characters
true, false provide truth values
whoami print current username
head -1 read a line from the keyboard
 Some Example Scripts

In the following examples, the text of the shell

script is shown on top, and the result of executing
it interactively is shown below. The text of each
example script is in a le available for
downloading, so that you can try the scripts
without having to type them in. The name of
each le is given in a comment in each script.
These examples show the scripts being executed
in the directory /home/mike/Classes/shell. When
you run a script in your own directory, some of
the names appearing in the output will be
dierent.
#!/bin/sh
# hi
echo "Hello, world!"
exit 0

unix[1] hi
Hello, world!
unix[2]
#!/bin/sh
# himike
name=Mike
echo "Hello, $name!"
exit 0

unix[3] himike
Hello, Mike!
unix[4]
#!/bin/sh
# rem
rm junk
echo "The return code from rm was $?"
exit 0

unix[5] touch junk

unix[6] rem
The return code from rm was 0
unix[7] rem
rm: junk: No such file or directory
The return code from rm was 2
#!/bin/sh
# quiet
rm junk 2> /dev/null
echo "The return code from rm was $?"
exit 0

unix[8] touch junk

unix[9] quiet
The return code from rm was 0
unix[10] quiet
The return code from rm was 2
#!/bin/sh
# pars
echo "There are $# parameters."
echo "The parameters are $@"
echo "The script name is $0"
echo "The first parameter is $1"
echo "The second parameter is $2"
exit 0

unix[11] pars apple orange

There are 2 parameters.
The parameters are apple orange
The script name is ./pars
The first parameter is apple
The second parameter is orange
unix[12]
#!/bin/sh
# shifter
echo $1
shift
echo $1
shift
echo $1
shift
echo $1
exit 0
unix[13] shifter one two three four five
one
two
three
four
unix[14] shifter one two three
one
two
three

unix[15] shifter one two

one
two

shift: shift count must be <= $#

unix[16]
#!/bin/sh
# sorter
rm -f /tmp/sorted
sort $1 > /tmp/sorted
cp /tmp/sorted $1
rm -f /tmp/sorted
exit 0
unix[17] more names
Jeff
Alan
Nancy
Yossl
Scott
Harriet
Chris
unix[18] sorter names
unix[19] more names
Alan
Chris
Harriet
Jeff
Nancy
Scott
Yossl
unix[20]
#!/bin/sh
# hiyou
name=whoami
echo "Hello, $name!"
exit 0

unix[21] hiyou
Hello, kupfem!
unix[22]
#!/bin/sh
# hiyou2
echo "Hello, whoami!"
exit 0

unix[23] hiyou2
Hello, kupfem!
unix[24]
#!/bin/sh
# countem
echo "File \"$1\" contains \
exactly wc $1 | cut -c6-7 lines."
exit 0

unix[25] countem text

File "text" contains exactly 21 lines.
unix[26]
 cut
reads std-in, extracts selected elds, writes std-out

cut -c1,4,7 characters 1, 4, and 7

cut -c1-3,8 characters 1 thru 3, and 8
cut -c-5,10 characters 1 thru 5, and 10
cut -c3- characters 3 thru last

cut -f1,4,7 tab-separated elds 1, 4, and 7

cut -d" " -f1 blank-separated eld 1

echo a.b.c | cut -d"." -f2 yields b

#!/bin/sh
# compile
if [ "$SRCDIR" = "" ]
then
echo "using default source directory"
SRCDIR=$HOME/src
else
echo "using source directory $SRCDIR"
fi
g77 $SRCDIR/$1
exit $?
unix[27] export SRCDIR=pwd
unix[28] compile hello.f
using source directory /home/mike/Classes/shell
unix[29] echo $?
0
unix[30] a.out
hello
unix[31] export SRCDIR=""
unix[32] compile hello.f
using default source directory
g77: /home/mike/src/hello.f:
No such file or directory
unix[33] echo $?
1
unix[34]
#!/bin/sh
# finder
grep $1 text > /dev/null
if [ $? -eq 0 ]
then
echo "found"
fi
exit 0

unix[35] finder ancient

found
unix[36] finder modern
unix[37]
#!/bin/sh
# compares
echo "true yields 0, false yields 1"
x="005"
[ "$x" = "005" ]
echo "Are strings 005 and 005 equal? $?"
[ "$x" = "5" ]
echo "Are strings 005 and 5 equal? $?"
[ $x -eq 005 ]
echo "Are integers 005 and 005 equal? $?"
[ $x -eq 5 ]
echo "Are integers 005 and 5 equal? $?"
exit 0
unix[38] compares
true yields 0, false yields 1
Are strings 005 and 005 equal? 0
Are strings 005 and 5 equal? 1
Are integers 005 and 005 equal? 0
Are integers 005 and 5 equal? 0
unix[39]
#!/bin/sh
# empty
if [ -s $1 ]
then
echo "The file $1 has contents."
exit 0
else
echo "The file $1 is absent or empty."
exit 1
fi

unix[40] empty text

The file text has contents.
unix[41] empty xxxx
The file xxxx is absent or empty.
unix[42] echo $?
1
unix[43]
#!/bin/sh
# adder
sum=0
for x in $@
do
sum=expr $sum + $x
done
echo "The sum is $sum."
exit 0

unix[44] adder 1 2 3 4 5
The sum is 15.
unix[45]
#!/bin/sh
# fixfor
for fyle in *.for
do
new=echo $fyle | sed -e"s/\.for$/\.f/"
mv $fyle $new
done
exit 0

unix[46] ls *.for
a.for b.for pgm.for xyz.w.for
unix[47] fixfor
unix[48] ls *.f
a.f b.f pgm.f xyz.w.f
#!/bin/sh
# suffix
for fyle in *.$1
do
new=echo $fyle | sed -e"s/\.$1$/\.$2/"
mv $fyle $new
done
exit 0

unix[49] ls *.f
a.f b.f pgm.f xyz.w.f
unix[50] suffix f for
unix[51] ls *.for
a.for b.for pgm.for xyz.w.for
unix[52]
 sed
reads std-in, edits line(s), writes std-out

sed -e"s/old/new/" replace rst old by new

sed -e"s/old/new/g" replace each old by new

old can be a regular expression

^ matches the beginning of the line
$ matches the end of the line
. matches any single character
.* matches zero or more characters
[tT] matches t or T
escape these and other special characters with \

unix[53] echo banana | sed -e"s/a$/\.x/"

banan.x
unix[54] more fruit
xapple
xpear
xplum
xcherry
unix[55] sed -e"s/^x/ /" < fruit
apple
pear
plum
cherry
 A Final Example
#! /bin/sh
# list names of all files containing given words

if [ $# -eq 0 ]
then
echo "findtext word1 word2 word3 ..."
echo "lists names of files containing all given words"
exit 1
fi

for fyle in *
do
bad=0
for word in $*
do
grep $word $fyle > /dev/null 2> /dev/null
if [ $? -ne 0 ]
then
bad=1
break
fi
done
if [ $bad -eq 0 ]
then
echo $fyle
fi
done
exit 0
 Gotchas

Never use test as the name of a variable or a

shell script le.

When using = as an assignment operator, do

not put blanks around it.

When using = as a comparison operator, you

must put blanks around it.

When using if [ ] put spaces around the

brackets (except after ] when it is the last
character on the line).
 Exercises
some hints are given in the files
exer1, exer2, exer2.aix, and exer3.

1. Write a script that counts files. (a) First make it count the
files in the current directory. (b) Now modify your script to
accept a parameter that is the name of a directory, and count
the files in that directory. Try this version on the current
directory (.) and on the /afs/rpi.edu/campus/doc directory.
(c) Further modify your script so that if it is invoked without
a parameter it prints out an explanation of how to use it.
2. If the ls command is given the name of a single extant file
it merely prints that filename back out. (a) Write a script myls
that behaves like ls except that when a single filename
parameter is supplied it produces the output that ls -l would
give for the file. (b) Revise your script so that when a single
filename parameter is given the output produced is the
filename followed by the date and time of its most recent
change and then the size of the file in bytes.
3. A script isyes is required that sets its exit code to 0 if its
parameter is some variation of y or yes, and to 1 otherwise.
(a) Assume the only acceptable parameter values meaning
yes are y, yes, Y, and YES, and solve the problem using only
shell programming features we have discussed. (b) Simplify
and generalize your script by using tr a-z A-Z, which reads
from std-in, translates to upper case, and writes to std-out.
4. Write a script that adds up the sizes reported by ls for the
files in the current directory. The script should print out only
the total number of bytes used.
 References

UNIXTM Shell Programming, Revised Edition, by

Stephen G. Kochan and Patrick H. Wood,
Hayden Books, 1990, ISBN 0-672-48448-X.

The UNIXTM Programming Environment, by Brian

W. Kernighan and Rob Pike, Prentice Hall, 1984,
ISBN 0-13-937681-X.

sed & awk, by Dale Dougherty, OReilly &

Associates, 1991, ISBN 0-937175-59-5.

Mastering Regular Expressions, by Jerey E. F.

Friedl, OReilly & Associates, 1997, ISBN
1-56592-257-3.