Here-document with cat

Question

I am new to Linux, shell scripting, and I'm trying to learn the fundamentals. I am currently on the redirection chapter, where here-documents are being explained. I could not really understand how the code below works in conjunction with the cat command. If some one can please tell me?

cat << EOF
line1 
line2
line3
EOF

I very well understand how << EOF works. The keyword denotes the end of input. My question is; cat takes filename(s) as input stream; so that it can print its content on output stream. How does above work internally? We are not specifying filename anywhere. Does the shell create internally some temporary file and give it to cat? I hope my question is clear.

Answer 1

When cat is not given a filename on the command line (or if the filename is just -), then it switches to reading from its standard input.

That means that with

cat <<END
something something
END

cat will notice that it was not given a filename and will proceed to read from the here-document, which is arriving on its standard input stream.

You can get cat to read from both its standard input and from a file with

cat - filename <<END
something something
END

This will cause the contents of the here-document to be concatenated with the contents of filename. If the order of the arguments was filename - then that would be the order that the data would be concatenated too.

Note that - is not special to the shell in any way, and is interpreted as "standard input" specifically by the cat utility (and a few others). If you have an actual file called - that you need to run cat on, use cat ./-, i.e. specify the filename with a path.

---

For all purposes, you can think of feeding a here-document directly into a utility as a shorthand for creating a temporary file and then invoking the utility with that attached to the standard input stream:

printf 'some contents' >tmpfile
utility <tmpfile
rm -f tmpfile

Here-documents might not actually be implemented this way (it may be a FIFO (named pipe)), but that's not an entirely incorrect way of thinking about it.

Reading the bash sources, it seems as if that particular shell implements here-documents by creating a temporary file using the mkstemp() library function if it's available, otherwise it tries to create a random filename to write to. The contents of the document is then written to that file and the data is provided on the standard input of whatever it is that's reading it, just as if it came from a file, redirected with <. See redir.c and lib/sh/tmpfile.c in the bash source distribution.