I would like to save egrep output to a bash array:
arr=( $(egrep -Rn 'regex') )
If there so happens to be a '*' in the egrep result, it appears like bash is expanding the '*' to be all files in current directory. And the expansion plus results of egrep are then saved into arr.
How do I fix this? I want the '*' in the grep results to be unaltered.
To use the idiom attempted in the question "correctly" might look something like this:
# DON'T DO THIS.
set -f # turn off globbing
IFS=$'\n' # word-split only on newlines
arr=( $(...) ) # populate array
unset IFS # return IFS to defaults (assuming it was there before)
set +f # turn globbing back on
Obviously, there's a lot of room to get this wrong and leave your shell in a state other than the way it started (What if your script had a different initial IFS value? What if this code is sourced from a script that wants globbing to be disabled to work correctly?). Don't do it.
One approach, compatible with bash 3.x, is to use read -a (reading into an array) with IFS (used to separate fields) containing a newline, and -d (used to separate records) set to a NUL:
IFS=$'\n' read -r -d '' -a arr < <(egrep -Rn 'regex' && printf '\0')
The trailing NUL added to the input is present to ensure that read exits successfully; otherwise, this could trigger an abrupt exit if using set -e.
A longer but more explicit approach is to do the iteration yourself:
arr=( )
while IFS= read -r; do
arr+=( "$REPLY" )
done < <(egrep -Rn 'regex')
Another, using bash 4.x features (readarray, AKA mapfile):
readarray -t arr < <(egrep -Rn 'regex')
< < that means anything. < means a thing, and <() means another thing, and they can be used together. :)$(...) is also a subshell. <() is process substitution.
set -f(but this leaves the word-splitting issues/etc. affecting the assignment and isn't recommended as-such).