Using "while read..." in a linux script

Question

Could somebody please explain how the following code works?

echo '1 2 3 4 5 6' | while read a b c
do
  echo $c $b $a
done

Specifically, I'd like to know why the output of this loop is 3 4 5 6 2 1, instead of 3 2 1 and 6 5 4 on two separate lines? I can't seem to wrap my mind around it...

score 47 · Accepted Answer · edited Apr 07 '15 at 19:44

47

read reads a whole line from standard input, splits the line into fields and assigns this fields to the given variables. If there are more pieces than variables, the remaining pieces are assigned to the last variable.

In your case $a is assigned 1, $b is assigned 2 and $c the remaining 3 4 5 6.

edited Apr 07 '15 at 19:44

guntbert

13,485

answered Apr 02 '15 at 21:27

Florian Diesch

89,821

Thanks Florian! Now it makes sense... For some reason I thought spaces would delimit each variable reading, but apparently not. I appreciate your help!! – linuxgringo Apr 04 '15 at 02:17

score 25 · Answer 2 · edited Apr 13 '17 at 12:37

Rewriting the loop this way reveals what is happening:

echo '1 2 3 4 5 6' | while read a b c
  do
    echo '(iteration beginning)' a="$a" b="$b" c="$c" '(iteration ending)'
  done

This gives, as its output:

(iteration beginning) a=1 b=2 c=3 4 5 6 (iteration ending)

Notice first that only a single echo command is run. If it were run more than once, you would, among other things, see the (iteration beginning) and (iteration ending) substrings printed more than once.

This is to say that having a while loop here is not really accomplishing anything. The read builtin reads whitespace-separated text¹ into each specified variable. Extra input gets appended to the end of the last variable specified.² Thus variables a and b take on the values 1 and 2 respectively, while c takes on the value 3 4 5 6.

When the loop condition (while read a b c) is evaluated the second time, there's no more input available from the pipe (we only piped it a single line of text), so the read command evaluates to false instead of true and the loop stops (before ever executing the body a second time).

_{1: To be technical and specific, the read builtin, when passed variable names as arguments, reads input, splitting it into separate "words" when it encounters IFS whitespace (see also this question and this article).}

_{2: read's behavior of jamming any extra fields of input into the last variable specified is nonintuitive to many scripters, at first. It becomes easier to understand when you consider that, as Florian Diesch's answer says, read will always (try to) read a whole line--and that read is intended to be usable both with and without a loop.}

Eliah, thanks for taking the time to explain all the details. I suspected that while didn't serve its normal purpose in this example, but then the read command threw me off... Somehow, I interpreted it as "while read a b c is not false, do echo ...". Thank you for explaining how it really worked. I came across this code yesterday and knew it would bug me until I figured it out... lol — linuxgringo, Apr 04 '15 at 02:22
@linuxgringo Actually, the loop body is executed each time read a b c evaluates to true, and the loop condition (read a b c) does run more than once. Bit it only evaluates to true the 1st time. The 2nd time, there's no more input to be read from the pipe, so end of file is encountered, causing read to return false. (See the last section of the output of help read, on "Exit Status," for details, noting that, in shell scripting, zero means true and nonzero means false.) If you piped more than one line of input to while read ..., the loop body would get executed multiple times. — Eliah Kagan, Apr 04 '15 at 03:05

score 0 · Answer 3 · answered Sep 24 '25 at 10:26

As said, read first reads a full line from standard input, i.e, a block of (UTF-8) text finished by a newline, and then it assigns what it read to whatever variables it finds on the input.

Note that at least the bash variant of read does use spaces as delimiters (by default), but the line must be read first (and that requires a newline).

echo '1 2 3 4 5 6' obviously only provides one line of input, so the while loop is just run once. The explanation of the 'strange' behaviour of how it then assigns variables has been already mentioned (i.e., the last variable gets 'all the rest of the match').

But, of course, we can cheat.

The first cheat is simply to place '1 2 3'and '4 5 6' in separate lines:

echo '1 2 3
4 5 6' | while read a b c; do
  echo $c $b $a;
done

(to type the example above directly on the shell, you will need to use Shift + Enter in most cases)

Or, alternatively, use two echo commands, one for each line, and eval the result as a single stream to feed to the while loop:

echo $(echo '1 2 3'; echo '4 5 6') | while read a b c; do
  echo $c $b $a;
done

Too ugly for your tastes (beyond requiring launching a subprocess)? Well, echo has a nice cute flag, -e, which allows introducing backslashed control characters, such as \n, like this:

echo -e '1 2 3\n4 5 6' | while read a b c; do
  echo $c $b $a;
done

Alternatively, you can replace echo with printf. printf, as its name implies, prints structured formatting, inspired by the C family of programming languages.

printf '1 2 3\n4 5 6\n' | while read a b c; do
  echo $c $b $a;
done

Notice that printf, like it's C-style equivalents, doesn't print a terminating newline on its own — you've got to explicitly add it at the end.

Finally, my favourite way: using a here-doc, like this:

cat | while read a b c; do echo $c $b $a; done << EOF
1 2 3
4 5 6
EOF

For best results, it's easier to keep everything in a single line when typing it out on the shell.

There are more variants to the theme, of course, but these should suffice — for now, at least!

Have fun

Using "while read..." in a linux script

3 Answers3