Multi-line string with extra space (preserved indentation)

Question

I want to write some pre-defined texts to a file with the following:

text="this is line one\n
this is line two\n
this is line three"

echo -e $text > filename

I'm expecting something like this:

this is line one
this is line two
this is line three

But got this:

this is line one
 this is line two
 this is line three

I'm positive that there is no space after each \n, but how does the extra space come out?

Answer 1

Heredoc sounds more convenient for this purpose. It is used to send multiple commands to a command interpreter program like ex or cat

cat << EndOfMessage
This is line 1.
This is line 2.
Line 3.
EndOfMessage

The string after << indicates where to stop.

To send these lines to a file, use:

cat > $FILE <<- EOM
Line 1.
Line 2.
EOM

You could also store these lines to a variable:

read -r -d '' VAR << EOM
This is line 1.
This is line 2.
Line 3.
EOM

This stores the lines to the variable named VAR.

When printing, remember the quotes around the variable otherwise you won't see the newline characters.

echo "$VAR"

Even better, you can use indentation to make it stand out more in your code. This time just add a - after << to stop the tabs from appearing.

read -r -d '' VAR <<- EOM
    This is line 1.
    This is line 2.
    Line 3.
EOM

But then you must use tabs, not spaces, for indentation in your code.

Answer 2

If you're trying to get the string into a variable, another easy way is something like this:

USAGE=$(cat <<-END
    This is line one.
    This is line two.
    This is line three.
END

)

If you indent your string with tabs (i.e., '\t'), the indentation will be stripped out. If you indent with spaces, the indentation will be left in.

NOTES:

• It is significant that the last closing parenthesis is on another line. The END text must appear on a line by itself.
• When using this variable, to preserve non-readable chars (\t, \n) you must quote it. I.e. use echo "$USAGE" (not: echo $USAGE)

Answer 3

echo adds spaces between the arguments passed to it. $text is subject to variable expansion and word splitting, so your echo command is equivalent to:

echo -e "this" "is" "line" "one\n" "this" "is" "line" "two\n"  ...

You can see that a space will be added before "this". You can either remove the newline characters, and quote $text to preserve the newlines:

text="this is line one
this is line two
this is line three"

echo "$text" > filename

Or you could use printf, which is more robust and portable than echo:

printf "%s\n" "this is line one" "this is line two" "this is line three" > filename

In bash, which supports brace expansion, you could even do:

printf "%s\n" "this is line "{one,two,three} > filename

Answer 4

in a bash script the following works:

#!/bin/sh

text="this is line one\nthis is line two\nthis is line three"
echo -e $text > filename

alternatively:

text="this is line one
this is line two
this is line three"
echo "$text" > filename

cat filename gives:

this is line one
this is line two
this is line three

Answer 5

I've found more solutions since I wanted to have every line properly indented:

1. You may use echo:

   echo    "this is line one"   \
       "\n""this is line two"   \
       "\n""this is line three" \
       > filename
   

   It does not work if you put "\n" just before \ on the end of a line.

2. Alternatively, you can use printf for better portability (I happened to have a lot of problems with echo):

   printf '%s\n' \
       "this is line one"   \
       "this is line two"   \
       "this is line three" \
       > filename
   

3. Yet another solution might be:

   text=''
   text="${text}this is line one\n"
   text="${text}this is line two\n"
   text="${text}this is line three\n"
   printf "%b" "$text" > filename
   

   or

   text=''
   text+="this is line one\n"
   text+="this is line two\n"
   text+="this is line three\n"
   printf "%b" "$text" > filename
   

4. Another solution is achieved by mixing printf and sed.

   if something
   then
       printf '%s' '
       this is line one
       this is line two
       this is line three
       ' | sed '1d;$d;s/^    //g'
   fi
   

   It is not easy to refactor code formatted like this as you hardcode the indentation level into the code.

5. It is possible to use a helper function and some variable substitution tricks:

   unset text
   _() { text="${text}${text+
   }${*}"; }
   # That's an empty line which demonstrates the reasoning behind 
   # the usage of "+" instead of ":+" in the variable substitution 
   # above.
   _ ""
   _ "this is line one"
   _ "this is line two"
   _ "this is line three"
   unset -f _
   printf '%s' "$text"
   

Answer 6

The following is my preferred way to assign a multi-line string to a variable (I think it looks nice).

read -r -d '' my_variable << \
_______________________________________________________________________________

String1
String2
String3
...
StringN
_______________________________________________________________________________

The number of underscores is the same (here 80) in both cases.

Answer 7

Hard to read:

This looks way too bash-like :) (hard to read) for my general taste:

cat << EndOfMessage
This is line 1.
This is line 2.
Line 3.
EndOfMessage

Better, easier-to-read:

Let's get something a little more Pythonic (this is still bash):

text="this is line one
      this is line two
      this is line three\n"
dedent text
printf "$text"             # print to screen
printf "$text" > file.txt  # print to a file

Ah...that's better. :) It reminds me of Python's textwrap.dedent() function which I use here.

Here is what the magic dedent function looks like:

dedent() {
    local -n reference="$1"
    reference="$(echo "$reference" | sed 's/^[[:space:]]*//')"
}

Sample output to screen:

this is line one
this is line two
this is line three

WithOUT calling dedent text first`, the output would look like this:

this is line one
      this is line two
      this is line three

The variable text is passed to dedent by reference, so that what is modified inside the function affects the variable outside the function.

For more details and explanation and references, see my other answer here: Equivalent of python's textwrap dedent in bash

Problem with your original attempt

OP's quote (with my emphasis added):

I'm positive that there is no space after each \n, but how does the extra space come out?

Your original attempt was this:

text="this is line one\n
this is line two\n
this is line three"
echo -e $text

...but your output had an extra space before the 2nd and 3rd line. Why?

By deduction and experimentation, my conclusion is that echo converts the actual newline at the end of the line (the one you got when you actually pressed Enter) into a space. That space therefore shows up before the line just after each \n in the text.

So, the solution is to escape the real newline at the end of each line by putting a backslash \ at the end of any line within the quotes of your string, like this:

text="this is line one\n\
this is line two\n\
this is line three"

echo -e "$text"

Do NOT put a space before those trailing backslashes (like this: text="this is line one\n \) or that space will go right back into your output and cause you the same problem with the extra space!

OR, just use my technique with the dedent function above, which also has the added functionality of being able to indent with your code to look really pretty and nice and readable.

Answer 8

I came hear looking for this answer but also wanted to pipe it to another command. The given answer is correct but if anyone wants to pipe it, you need to pipe it before the multi-line string like this

echo | tee /tmp/pipetest << EndOfMessage
This is line 1.
This is line 2.
Line 3.
EndOfMessage

This will allow you to have a multi line string but also put it in the stdin of a subsequent command.

Answer 9

As per @alphado's answer, and subsequent comments, this is what worked for me and gives 0 exit status.

read -r -d '\0' VAR <<- EOM
    This is line 1.
    This is line 2.
    Line 3.
\0
EOM

• <<- to ignore leading tabs
• set \0 as explicit "end marker"
• this also substitutes variables between the EOMs, so use \$ to print verbatim.

(Just putting this as a separate answer because it got hard to decipher this between all the comments, and did not want to change too much in original answer)

Answer 10

it will work if you put it as below:

AA='first line
\nsecond line 
\nthird line'
echo $AA
output:
first line
second line
third line

Answer 11

There are many ways to do it. For me, piping the indented string into sed works nicely.

printf_strip_indent() {
   printf "%s" "$1" | sed "s/^\s*//g" 
}

printf_strip_indent "this is line one
this is line two
this is line three" > "file.txt"

This answer was based on Mateusz Piotrowski's answer but refined a bit.

Answer 12

Just to mention a simple one-line concatenation as it can be useful sometimes.

# for bash

v=" guga "$'\n'"   puga "

# Just for an example.
v2="bar "$'\n'"   foo "$'\n'"$v"

# Let's simplify the previous version of $v2.
n=$'\n'
v3="bar ${n}   foo ${n}$v"

echo "$v3" 

You'll get something like this

bar 
   foo 
 guga 
   puga 

All leading and ending white spaces will be preserved right for

echo "$v3" > filename

Answer 13

Or keeping text indented with whitespaces:

#!/bin/sh

sed 's/^[[:blank:]]*//' >filename <<EOF
    this is line one
    this is line two
    this is line three
EOF

Same but using a varible:

#!/bin/sh

text="$(sed 's/^[[:blank:]]*//' << whatever
    this is line one
    this is line two
    this is line three
)"

echo "$text" > filename

;-)

Answer 14

In bash and dash (the ones I tested, should work for yours), you can use some Python-style triple-quote strings:

echo """Hello
World!"""

This outputs:

Hello
World!