running into error with multiline sed statement

Question

I have a very long sed statement that is 6 lines long strung together with -e. I am doing multiple replacements and writing to the same file.

It throws me an error for any sed statement over one line. How can I rectify this?

Some have suggested using variables to reduce the text length, but even still I think that solution won't work for me. Is there another simpler way to make this work?

Thanks.

edit: I just wasn't using my line breaks properly. Working script:

sed -e 's/horse.*/300/g' \
-e 's/goat.*/320/g' \
-e 's/dog.*/310/g' \ 
-e 's/wolf.*/330/g' < "$file1" > "$file2"

Thanks again.

Answer 1

See if this works:

sed 's/horse.*/300/g;s/goat.*/320/g;s/dog.*/310/g;s/wolf.*/330/g' filename.txt

Your script never ends because you are not feeding your SED command with a filename.

Answer 2

To summarize the various options for passing a script (program) [made up of multiple commands] to sed:

Let's assume you want to perform the following 2 commands:

• s/a/A/
• s/b/B/

on input 'ab', so as to get 'AB'.

---

Use a single-line string and separate the commands with ;:

echo 'ab' | sed 's/a/A/; s/b/B/'

Caveat: Some sed implementations - notably, FreeBSD/macOS sed require actual line breaks as part of certain commands, such as for defining labels and branching to them; typically, using separate -e options (see below) is an alternative.

^{For a comprehensive summary of the differences between GNU and FreeBSD/macOS sed, see this answer.}

---

Use a multi-line string and use actual line breaks to separate the commands:

echo 'ab' | sed -e '
              s/a/A/
              s/b/B/
              '

Note: You can also use a here-document with the -f option - see below.

---

Use multiple -e options to pass the commands individually.

echo 'ab' | sed -e 's/a/A/' -e 's/b/B/'

If your option arguments are long and you want to spread the -e options across several lines, terminate all but the last line of the overall command with \ so as to signal to the shell that the command continues on the next line - note that the \ must be the very last character on the line (not even spaces or tabs are allowed after it):

echo 'ab' | sed -e 's/a/A/' \
                -e 's/b/B/'

Note that this is a general requirement of the shell and has nothing to do with sed itself.

---

Write your sed script to a file or use a here-document, and pass it to sed with the -f option:

# Create script file.
cat > myScript <<'EOF'
  s/a/A/
  s/b/B/
EOF

# Pass to `sed` with `-f`:
echo 'ab' | sed -f myScript

# Alternative, with a here-document via stdin, which requires
# the input to come from an actual file:
# Note: GNU sed also accepts `-f -`, but macOS sed does not.
echo 'ab' > test.txt
sed -f /dev/stdin test.txt <<'EOF'
  s/a/A/
  s/b/B/
EOF

---

Note:

• All POSIX-compatible sed implementations should support all these methods.

• You can even mix the -f and -e approaches, where the -e supplied command are appended to the overall script; e.g.: echo 'abc' | sed -f myScript -e 's/c/C/' # -> 'ABC'

• If you use the variant with the here-document as the script (<<'EOF'...), use -f /dev/stdin to be portable (only GNU sed supports -f -).

Answer 3

sed -e 's/horse.*/300/g' \  #replacing horse with horse300
-e 's/goat.*/320/g' \       #replacing goat with goat320   
-e 's/dog.*/310/g' \        #replacing dog with dog310
-e 's/wolf.*/330/g' < "$file1" > "$file2" #replacing wolf with wolf330 ; reading from file1, writing to file2

Above is the working code used for making multiple replacements using the same sed statement. Thanks again for all of the help.