To store code, use functions, not variables:
parm() {
< file awk '
$1 > 0 {$1 = "abc"; print f $0}
{f = $0 ORS}'
}
parm
If you had to use variables:
parm='
< file awk '\''
$1 > 0 {$1 = "abc"; print f $0}
{f = $0 ORS}'\''
'
eval "$parm"
Is probably the easiest where we only use single quotes, and to insert single quotes, we get out of the quotes and quote the single quote with \'.
See also:
parm=$(cat << 'EOF'
< file awk '
$1 > 0 {$1 = "abc"; print f $0}
{f = $0 ORS}'
EOF
)
eval "$parm"
When the delimiter is quoted in << 'delimiter', then there's no $param, $(( arith )), `cmdsubst`, $(cmdsubst) nor \ line continuation happening inside here documents which makes it easy to write any multiline text containing anything.
I've replaced cat file | (makes little sense to concatenate a single file) with < file, moved $1 = "abc" assignment out of the condition where it was confusing and added a separator (here a newline which also improves legibility) between the two awk statements as required by POSIX.
Using functions also means that it can get parametrised as in:
parm() {
< "$1" awk '
$1 > 0 {$1 = "abc"; print f $0}
{f = $0 ORS}'
}
parm file
Where the first argument is made the stdin of awk or:
parm() {
cat -- "$@" | awk '
$1 > 0 {$1 = "abc"; print f $0}
{f = $0 ORS}'
}
echo -1 3 | parm file1 - file2
Where all the arguments are taken as file paths (or stdin if -¹ or no argument is supplied) and concatenated and the result fed as awk's stdin via a pipe or:
parm() {
awk -- '
$1 > 0 {$1 = new; print f $0}
{f = $0 ORS}' new=abc "$@"
}
echo -1 3 | parm file1 - new=xyz file2
Where all the arguments are passed to awk which can take them as file paths, stdin for -¹ (or if there's no file path argument) or awk variable assignments² ³.
¹ use ./- for the file literally called - in the current working directory
² use ./foo=bar.txt for a file literally called foo=bar.txt in the current working directory.
³ Beware that in GNU awk aka gawk, new=@/text/ is interpreted as new=text with text taken as a regexp. new='\@/text/' works around it but issues a warning, a better approach at working around that misfeature is to call awk as POSIXLY_CORRECT=1 awk ... which disables that misfeature for gawk and should be harmless for other awk implementations.
