Shell Programming and Scripting

Hi,
Below is the code I am using. I am trying to list only those numbers which has a + symbol in it

Code:

 cat num | awk -F"+" '{if (/^$/)
{
}
else
{if ( $0 ~ egrep "^[0-9+-]+$" )
  {
    if ( $0 ~ grep "+" )
     {print $0}
  }
}}'

I am getting the following error:

Code:

awk: 0602-521 There is a regular expression error.
        ?*+ not preceded by valid expression.

 The input line number is 3.
 The source line number is 7.

Please help.

Code:

awk 'NF>1' FS=+ infile

I am not looking for an alternate way to do this. I would like to know what went wrong with the syntax I am using. Please help.

This is your version, cleaned up so it works, along with test results (test file shown below).

Code:

cat test | awk -F"+" '
{
   if ( $0 ~ /^$/ )
   {
   }
   else
   {
      if ( $0 ~ /^[0-9+-]+$/ )
      {
         if ( $0 ~ /\+/ )
         {print $0}
      }
   }
}'
+1
0+1
1+

Personally, I'd just go with the following, which yields almost the same results, but which could easily altered to yield the exact same results, as what you had (after it was cleaned up), and it is much simpler to read:

# Looking for numbers with a '+' in front of of trailing them, as you might see in a mathematical expression), which matches the output you get in my cleaned up version of your code:

Code:

awk '/\+[0-9]/ || /[0-9]\+/' test
+1
 0+1
 1+
let x=x+1

You did not state whether your input file contained 'statements' such as the above let statement but, if it does, your version will not find them because you are 'anchoring' your regular expression to the beginning and end of the line. Also, your version will not match numbers which contain trailing spaces, which I did not demonstrate in my example but which you can easily test yourself.

Code:

cat test
+1
0+1
1
a
+a
1+
a+
let x=x+1

What went wrong with what your code? The answer is that the '+' is a metacharacter in regular expressions and it must be escaped with a backslash if you want to look for a literal plus sign. (The line just before the print statement in your code.)

According to the Posix regexp standards, + "Matches the preceding element one or more times."

Code:

awk '/+[0-9]/' test
awk: line 1: regular expression compile failed (missing operand)
+[0-9]

fails because there is no preceding operand to match one or more times, just as there is none in your regexp ( "+" ).

and this:

Code:

awk '/[0-9]+/' test
+1
0+1
1
1+

does not fail but does not yield the desired results because, according to the Posix definition, '1' is matched one or more times and we are not looking for a literal '+'.

Other things that didn't go wrong yet because your script failed before it could reach them:

You are attempting to call grep and egrep from within your awk script but you cannot do that without using the awk system call, nor do you need to since awk understands regular expressions.

While it is syntactically correct, I'm not really sure why you are changing the awk field separator to a '+' sign and only applying your logic to 'non-blank' lines. With a properly structured regular expression, as the example above demonstrates, only the lines you are interested in are printed and none of that is required

Thanks a lot for all your explanations and efforts. As you had rightly pointed out, I need not grep for blank lines having specified the field separator as '+'. Also, I missed to escape the metacharacter. This cleaned up code works fine:

Code:

cat num | awk -F"+" '
{if ( $0 ~ /^[0-9+-]+$/ )
      {
         if ( $0 ~ /\+/ )
         {print $0}
      }
   }'

This is Useless Use of Cat.

Code:

awk '{ commands .. }' num

According to the useless use of cat web page that you refer to:

"Briefly, here's the collected wisdom on using cat:

The purpose of cat is to concatenate (or "catenate") files. If it's only one file, concatenating it with nothing at all is a waste of time, and costs you a process."

By extension of the logic in that argument, ANY use of cat is a useless use of cat since you can concatenate files (the purpose of cat, per the above quote) using only awk (per the below example) without costing yourself a process. In fact, this is true of most Unix commands that you would normally pipe the output of "cat file" to so perhaps they should just remove the cat command from Unix entirely for being redundant and, therefore, unnecessary?

awk '{print}' file1 file2 file3
1 line
2 lines
3 lines

awk '{print}' file1
1 line

awk '{print}' file2
2 lines

awk '{print}' file3
3 lines