Why doesn't `cat foo.txt | while read` fail if file not found?

Question

I use this loop to iterate through files or, if there are no files, read stdin:

#!/bin/bash
set -e
cat "$@" | while read arr ; do
    echo "Got this line ${arr}"
done

The problem is that if the file doesn't exist, it doesn't error out.

You can see the full example here: https://github.com/StackExchange/blackbox/blob/master/tools/mk_rpm_fpmdir

cat does return error code 1 if any file is not found. However the error doesn't cause the program to stop.

How can I iterate through $@ and fail if a file is not found?

Answer 1

There are several available options. One is to simply ditch cat, and thus have no pipeline (which has the advantageous side effect of avoiding subshell use, and thus allowing changes to shell state -- variables set, etc -- to last beyond the inner loop):

set -e
for f; do # in "$@" is implicit
  while read -r -a arr; do
    printf 'Got line: '
    printf '%q ' "${arr[@]}"
    printf '\n'
  done <"$f"
done

The above also goes out of its way to print the content read in a way that entirely preserves the array read, and prints its contents unambiguously (distinguishing an array containing two elements foo bar baz from one containing three elements foo bar baz).

---

Another is to set pipefail, which will cause a pipeline to be considered failed if any component of a pipeline returns a nonzero exit status, including cat:

set -e
set -o pipefail
cat "$@" | while IFS= read -r; do
  printf 'Got line: %q\n' "$REPLY"
done

This works by overriding the default behavior by which only the right-hand side of a pipeline matters for purposes of considering the overall exit status of the command.

Answer 2

for f in "${@}"; do
    test -f "${f}" || exit # fail if file doesn't exist
    cat "${f}" | awk '{print "Got this line" $0}'
done