Shell ScriptingAwk (part1)
Awk Programming Language
standard unix language that is geared for text processing and creating formatted reports
but it is very valuable to seismologists because it uses floating point math, unlike integer only bash, and is designed to work with columnar data
syntax similar to C and bash
one of the most useful unix tools at your command
considers text files as fields (columns) and records (lines)
performs floating & integer arithmetic and string operations
uses loops and conditionals
define your own functions (subroutines)
execute unix commands within the scripts and process the results
versionsawk: original awk
nawk: new awk, dates to 1987
gawk: GNU awk has more powerful string functionality
the CERI unix system has all three. You want to use nawk. I suggest adding this line to your .cshrc file
alias awk ‘nawk’
in OS X, awk is already nawk so no changes are necessary
Command line functionality
you can call awk from the command line two ways:
awk [options] ‘{ commands }’ variables infile(s)awk –f scriptfile variables infile(s)
or you can create an executable awk script%cat << EOF > test.awk
#!/usr/bin/nawk
some set of commands
EOF
%chmod 755 test.awk
%./test.awk
How it treats textawk commands are applied to every record or line
of a file
it is designed to separate the data in each line into a field
essentially, each field becomes a member of an array so that the first field is $1, second field $2 and so on.
$0 refers to the entire record
Field Separatorthe default field separator is one or more white
spaces
$1 $2 $3 $4 $5 $6 $7 $8 $9 $10 $11
1 1918 9 22 9 54 49.29 -1.698 98.298 15.0 ehb
Notice that the fields may be integer, floating point (have a decimal point) or strings. Nawk is generally smart enough to figure out how to use them.
Field Separatorthe field separator may be modified by
resetting the FS built in variable
Look at passwd file%head -n1 /etc/passwd
root:x:0:1:Super-User:/:/sbin/sh
Separator is “:”, so reset it.
%awk –F”:” ‘{ print $1, $3}’ /etc/passwd
root 0
printOne of the most common commands used in awk
scripts is print
awk is not sensitive to white space in the commands
%awk –F”:” ‘{ print $1 $3}’ /etc/passwd
root0
two solutions to this%awk –F”:” ‘{ print $1 “ “ $3}’ /etc/passwd
%awk –F”:” ‘{ print $1, $3}’ /etc/passwd
root 0
any string or numeric text can be explicitly output using “”
Assume a starting file like so:
1 1 1918 9 22 9 54 49.29 -1.698 98.298 15.0 0.0 0.0 ehb FEQ x
%awk '{print "latitude:",$9,"longitude:",$10,"depth:",$11}’ SUMA. loc
latitude: -1.698 longitude: 98.298 depth: 15.0
latitude: 9.599 longitude: 92.802 depth: 30.0
latitude: 4.003 longitude: 94.545 depth: 20.0
• Unlike the shell AWK does not evaluate variables within strings.
• The second line, for example, could not be written:
{print "$8\t$3" }
As it would print ”$8 $3.”
• Inside quotes, the dollar sign is not a special character. Outside, it corresponds to a field.
1 1 1918 9 22 9 54 49.29 -1.698 98.298 15.0 0.0 0.0 ehb FEQ x
you can specify a newline in two ways
%awk '{print "latitude:",$9; print "longitude:",$10}’ SUMA. loc
%awk '{print "latitude:",$9”\n”,”longitude:",$10}’ SUMA. loc
latitude: -1.698
longitude: 98.298
a trickIf a field is composed of both strings and
numbers, you can multiple the field by 1 to remove the string.
%head test.tmp
1.5 2008/09/09 03:32:10 36.440N 89.560W 9.4
1.8 2008/09/08 23:11:39 36.420N 89.510W 7.1
1.7 2008/09/08 19:44:29 36.360N 89.520W 8.2
%awk '{print $4,$4*1}' test.tmp
36.440N 36.44
36.420N 36.42
36.360N 36.36
Selective executionawk recognizes regular expressions and
conditionals, which can be used to selective execute awk procedures on certain records
%awk –F”:” ‘ /root/ { print $1, $3}’ /etc/passwd #reg expr
root 0
or within our example script
#!/usr/bin/nawk -f
/root/ { print $1}
if statements are also very useful%awk –F”:” ‘ {if ($1==“root”) print $1, $3}’ /etc/passwd
root 0
or within our example script
{ if ($1==“root”) { print $1 }
note, this particular if syntax is a bit different from your reading, which suggested
%awk –F”:” ‘ $1==“root” {print $1, $3}’ /etc/passwd
the syntax I use is more explicit and more like C or perl, so I essentially have to remember less syntax
Floating Point Arithmeticawk does floating point math!!!!!
it stores all variables as strings, but when math operators are applied, it converts the strings to floating point numbers if the string consists of numeric characters
the reading calls this stringy variables
Arithmetic OperatorsAll basic arithmetic is left to right associative
+ : addition- : subtraction* : multiplication / : division% : remainder or modulus ^ : exponent other standard C programming operators
Assignment Operators= : set variable equal to value on right+= : set variable equal to itself plus the value on
right-= : set variable equal to itself minus the value on
right*= : set variable equal to itself times the value on
right/= : set variable equal to itself divided by value on
right%= : set variable equal to the remainder of itself
divided by the value on the right^= : set variable equal to the itself to the
exponent following the equal sign
Unary OperationsA unary expression contains one operand and one
operator
++ : increment the operand by 1
if ++ occurs after, $x++, the original value of the operand is used in the expression and then incremented
if ++ occurs before, ++$x, the incremented value of the operand is used in the expression
-- : decrement the operand by 1+ : unary plus maintains the value of the operand, x=+x - : unary minus negates the value of the operand, -
1*x=-x! : logical negation evaluates if the operand is true
(returns 1) or false (returns 0)
Relational OperatorsReturns 1 if true and 0 if false
!!! opposite of bash test command
All relational operators are left to right associative
< : test for less than<= : test for less than or equal to> : test for greater than>= : test for greater than or equal to== : test for equal to!= : test for not equal
Boolean (Logical) Operators
Boolean operators return 1 for true and 0 for false
&& : logical AND; tests that both expressions are true
left to right associative
|| : logical OR ; tests that one or both of the expressions are true
left to right associative
! : logical negation; tests that expression is true
Unlike bash, the comparison and relational operators don’t have different syntax for strings and numbers.
ie: == in awk rather than == or eq using test
Comparison Operators~ : pattern match
!~ : pattern does not match
&& : logical AND
|| : logical OR
== : equals (numeric or string)
!= : does not equal (numeric or string)
Built-In VariablesFS: Field Separator
NR: record number is another useful built-in awk variable
it takes on the current line number, starting from 1
%awk –F”:” ‘ {if (NR==1) print $1, $3}’ /etc/passwd
root 0
this is useful when headers are present in a file
RS : record separator specifies when the current record ends and the next begins
default is “\n” or newlineuseful option is “”, or a blank line
OFS : output field separatordefault is “ “ or a whitespace
ORS : output record separatordefault is a “\n” or newline
NF : number of fields in the current recordthink of this as awk looking ahead to the next RS to
count the number of fields in advance
FILENAME : stores the current filename
OFMT : output format for numbers example OFMT=“%.6f” would make all numbers
output as floating points
Accessing shell variables in nawk
3 methods to access shell variables inside a nawk script ...
1. Assign the shell variables to awk variables after the body of the script, but before you specify the input file
VAR1=3VAR2=“Hi”
awk '{print v1, v2}' v1=$VAR1 v2=$VAR2 input_file
3 HiNote that I am sneaking in the concept of awk variables here (v1,v2)
There are a couple of constraints with this method
- Shell variables assigned using this method are not available in the BEGIN section
- If variables are assigned after a filename, they will not be available when processing that filename
awk '{print v1, v2}' v1=$VAR1 file1 v2=$VAR2 file2
In this case, v2 is not available to awk when processing file1.
Also note: awk variables are referred to by just their name (no $ in front)
awk '{print v1, v2, NF, NR}' v1=$VAR1 file1 v2=$VAR2 file2
2. Use the -v switch to assign the shell variables to awk variables.
This works with nawk, but not with all flavours of awk.
nawk -v v1=$VAR1 -v v2=$VAR2 '{print v1, v2}' input_file
3. Protect the shell variables from awk by enclosing them with "'" (i.e. double quote
- single quote - double quote).
awk '{print "'"$VAR1"'", "'"$VAR2"'"}' input_file