sed: replace multi-line block with standard input

Question

Given is input_file:

1
2
START
foo
END
3

Goal is to replace content inside including START ... END block with multi-line content from stdin:

1
2
hello 
world
3

What I tried:

sed '/^START$/,/^END$/d;r /dev/stdin' input_file <<EOF 
hello
world
EOF

unfortunately results in

1
hello
world
2
3

I guess r /dev/stdin is invoked sequentially after /^START$/,/^END$/d; and just appends after first line.

Second try:

sed '/^START$/,/^END$/{d;r /dev/stdin
}' input_file <<EOF     
hello
world
EOF

prints

1
2
3

Why do above commands - especially last one - do print the wrong result? And how might I adjust these?

Answer 1

In your first try, the address range is valid for the d only. The r /dev/stdin (append...) is done for the first line; thereafter, it runs into end-of-file. In your second try, the script doesn't encounter the r command. man sed:

  d      Delete pattern space.  Start next cycle.

So everything after the d command is lost (in the relevant address range); it just doesn't r /dev/stdin.

Try this to achieve yout target:

sed -e '/^START$/,/^END$/ { r /dev/stdin' -e';d};' file3 <<EOF 
hello
world
EOF

1
2
hello
world
3

read is the first thing when address range is met, then delete the address range.