Add a prefix string to beginning of each line

问题:

I have a file as below:

line1
line2
line3

And I want to get:

prefixline1
prefixline2
prefixline3

I could write a Ruby script, but it is better if I do not need to.

prefix will contain /. It is a path, /opt/workdir/ for example.

回答1:

# If you want to edit the file in-place
sed -i -e 's/^/prefix/' file

# If you want to create a new file
sed -e 's/^/prefix/' file > file.new

If prefix contains /, you can use any other character not in prefix, or escape the /, so the sed command becomes

's#^#/opt/workdir#'
# or
's/^/\/opt\/workdir/'

回答2:

awk '$0="prefix"$0' file > new_file

回答3:

You can use Vim in Ex mode:

ex -sc '%s/^/prefix/|x' file

1. % select all lines

2. s replace

3. x save and close

回答4:

If your prefix is a bit complicated, just put it in a variable:

prefix=path/to/file/

Then, you pass that variable and let awk deal with it:

awk -v prefix="$prefix" '{print prefix $0}' input_file.txt

回答5:

If you have Perl:

perl -pe 's/^/PREFIX/' input.file

回答6:

Using the shell:

#!/bin/bash
prefix="something"
file="file"
while read -r line
do
 echo "${prefix}$line"
done <$file > newfile
mv newfile $file

回答7:

While I don't think pierr had this concern, I needed a solution that would not delay output from the live "tail" of a file, since I wanted to monitor several alert logs simultaneously, prefixing each line with the name of its respective log.

Unfortunately, sed, cut, etc. introduced too much buffering and kept me from seeing the most current lines. Steven Penny's suggestion to use the -s option of nl was intriguing, and testing proved that it did not introduce the unwanted buffering that concerned me.

There were a couple of problems with using nl, though, related to the desire to strip out the unwanted line numbers (even if you don't care about the aesthetics of it, there may be cases where using the extra columns would be undesirable). First, using "cut" to strip out the numbers re-introduces the buffering problem, so it wrecks the solution. Second, using "-w1" doesn't help, since this does NOT restrict the line number to a single column - it just gets wider as more digits are needed.

It isn't pretty if you want to capture this elsewhere, but since that's exactly what I didn't need to do (everything was being written to log files already, I just wanted to watch several at once in real time), the best way to lose the line numbers and have only my prefix was to start the -s string with a carriage return (CR or ^M or Ctrl-M). So for example:

#!/bin/ksh

# Monitor the widget, framas, and dweezil
# log files until the operator hits <enter>
# to end monitoring.

PGRP=$$

for LOGFILE in widget framas dweezil
do
(
    tail -f $LOGFILE 2>&1 |
    nl -s"^M${LOGFILE}>  "
) &
sleep 1
done

read KILLEM

kill -- -${PGRP}

回答8:

Using ed:

ed infile <<'EOE'
,s/^/prefix/
wq
EOE

This substitutes, for each line (,), the beginning of the line (^) with prefix. wq saves and exits.

If the replacement string contains a slash, we can use a different delimiter for s instead:

ed infile <<'EOE'
,s#^#/opt/workdir/#
wq
EOE

I've quoted the here-doc delimiter EOE ("end of ed") to prevent parameter expansion. In this example, it would work unquoted as well, but it's good practice to prevent surprises if you ever have a $ in your ed script.

回答9:

Here's a wrapped up example using the sed approach from this answer:

$ cat /path/to/some/file | prefix_lines "WOW: "

WOW: some text
WOW: another line
WOW: more text

prefix_lines

function show_help()
{
  IT=$(CAT <<EOF
    Usage: PREFIX {FILE}

    e.g.

    cat /path/to/file | prefix_lines "WOW: "

      WOW: some text
      WOW: another line
      WOW: more text
  )
  echo "$IT"
  exit
}

# Require a prefix
if [ -z "$1" ]
then
  show_help
fi

# Check if input is from stdin or a file
FILE=$2
if [ -z "$2" ]
then
  # If no stdin exists
  if [ -t 0 ]; then
    show_help
  fi
  FILE=/dev/stdin
fi

# Now prefix the output
PREFIX=$1
sed -e "s/^/$PREFIX/" $FILE