Rearrange columns using cut

For the cut(1) man page:

Use one, and only one of -b, -c or -f. Each LIST is made up of one range, or many ranges separated by commas. Selected input is written in the same order that it is read, and is written exactly once.

It reaches field 1 first, so that is printed, followed by field 2.

Use awk instead:

awk '{ print $2 " " $1}' file.txt

using just the shell,

while read -r col1 col2
do
  echo $col2 $col1
done <"file"

Using join:

join -t $'\t' -o 1.2,1.1 file.txt file.txt

Notes:

• -t $'\t' In GNU join the more intuitive -t '\t' without the $ fails, (coreutils v8.28 and earlier?); it's probably a bug that a workaround like $ should be necessary. See: unix join separator char.

• join needs two filenames, even though there's just one file being worked on. Using the same name twice tricks join into performing the desired action.

• For systems with low resources join offers a smaller footprint than some of the tools used in other answers:

  wc -c $(realpath `which cut join sed awk perl`) | head -n -1
    43224 /usr/bin/cut
    47320 /usr/bin/join
   109840 /bin/sed
   658072 /usr/bin/gawk
  2093624 /usr/bin/perl
  

Just as an addition to answers that suggest to duplicate the columns and then to do cut. For duplication, paste etc. will work only for files, but not for streams. In that case, use sed instead.

cat file.txt | sed s/'.*'/'&\t&'/ | cut -f2,3

This works on both files and streams, and this is interesting if instead of just reading from a file with cat, you do something interesting before re-arranging the columns.

By comparison, the following does not work:

cat file.txt | paste - - | cut -f2,3

Here, the double stdin placeholder paste does not duplicate stdin, but reads the next line.

Using sed

Use sed with basic regular expression's nested subexpressions to capture and reorder the column content. This approach is best suited when there are a limited number of cuts to reorder columns, as in this case.

The basic idea is to surround interesting portions of the search pattern with \( and \), which can be played back in the replacement pattern with \# where # represents the sequential position of the subexpression in the search pattern.

For example:

$ echo "foo bar" | sed "s/\(foo\) \(bar\)/\2 \1/"

yields:

bar foo

Text outside a subexpression is scanned but not retained for playback in the replacement string.

Although the question did not discuss fixed width columns, we will discuss here as this is a worthy measure of any solution posed. For simplicity let's assume the file is space delimited although the solution can be extended for other delimiters.

Collapsing Spaces

To illustrate the simplest usage, let's assume that multiple spaces can be collapsed into single spaces, and the the second column values are terminated with EOL (and not space padded).

File:

bash-3.2$ cat f
Column1    Column2
str1       1
str2       2
str3       3
bash-3.2$ od -a f
0000000    C   o   l   u   m   n   1  sp  sp  sp  sp   C   o   l   u   m
0000020    n   2  nl   s   t   r   1  sp  sp  sp  sp  sp  sp  sp   1  nl
0000040    s   t   r   2  sp  sp  sp  sp  sp  sp  sp   2  nl   s   t   r
0000060    3  sp  sp  sp  sp  sp  sp  sp   3  nl 
0000072

Transform:

bash-3.2$ sed "s/\([^ ]*\)[ ]*\([^ ]*\)[ ]*/\2 \1/" f
Column2 Column1
1 str1
2 str2
3 str3
bash-3.2$ sed "s/\([^ ]*\)[ ]*\([^ ]*\)[ ]*/\2 \1/" f | od -a
0000000    C   o   l   u   m   n   2  sp   C   o   l   u   m   n   1  nl
0000020    1  sp   s   t   r   1  nl   2  sp   s   t   r   2  nl   3  sp
0000040    s   t   r   3  nl
0000045

Preserving Column Widths

Let's now extend the method to a file with constant width columns, while allowing columns to be of differing widths.

File:

bash-3.2$ cat f2
Column1    Column2
str1       1
str2       2
str3       3
bash-3.2$ od -a f2
0000000    C   o   l   u   m   n   1  sp  sp  sp  sp   C   o   l   u   m
0000020    n   2  nl   s   t   r   1  sp  sp  sp  sp  sp  sp  sp   1  sp
0000040   sp  sp  sp  sp  sp  nl   s   t   r   2  sp  sp  sp  sp  sp  sp
0000060   sp   2  sp  sp  sp  sp  sp  sp  nl   s   t   r   3  sp  sp  sp
0000100   sp  sp  sp  sp   3  sp  sp  sp  sp  sp  sp  nl
0000114

Transform:

bash-3.2$ sed "s/\([^ ]*\)\([ ]*\) \([^ ]*\)\([ ]*\)/\3\4 \1\2/" f2
Column2 Column1
1       str1      
2       str2      
3       str3      
bash-3.2$ sed "s/\([^ ]*\)\([ ]*\) \([^ ]*\)\([ ]*\)/\3\4 \1\2/" f2 | od -a
0000000    C   o   l   u   m   n   2  sp   C   o   l   u   m   n   1  sp
0000020   sp  sp  nl   1  sp  sp  sp  sp  sp  sp  sp   s   t   r   1  sp
0000040   sp  sp  sp  sp  sp  nl   2  sp  sp  sp  sp  sp  sp  sp   s   t
0000060    r   2  sp  sp  sp  sp  sp  sp  nl   3  sp  sp  sp  sp  sp  sp
0000100   sp   s   t   r   3  sp  sp  sp  sp  sp  sp  nl 
0000114

Lastly although the question's example does not have strings of unequal length, this sed expression support this case.

File:

bash-3.2$ cat f3
Column1    Column2
str1       1      
string2    2      
str3       3      

Transform:

bash-3.2$ sed "s/\([^ ]*\)\([ ]*\) \([^ ]*\)\([ ]*\)/\3\4 \1\2/" f3
Column2 Column1   
1       str1      
2       string2   
3       str3    
bash-3.2$ sed "s/\([^ ]*\)\([ ]*\) \([^ ]*\)\([ ]*\)/\3\4 \1\2/" f3 | od -a
0000000    C   o   l   u   m   n   2  sp   C   o   l   u   m   n   1  sp
0000020   sp  sp  nl   1  sp  sp  sp  sp  sp  sp  sp   s   t   r   1  sp
0000040   sp  sp  sp  sp  sp  nl   2  sp  sp  sp  sp  sp  sp  sp   s   t
0000060    r   i   n   g   2  sp  sp  sp  nl   3  sp  sp  sp  sp  sp  sp
0000100   sp   s   t   r   3  sp  sp  sp  sp  sp  sp  nl 
0000114

Comparison to other methods of column reordering under shell

• Surprisingly for a file manipulation tool, awk is not well-suited for cutting from a field to end of record. In sed this can be accomplished using regular expressions, e.g. \(xxx.*$\) where xxx is the expression to match the column.

• Using paste and cut subshells gets tricky when implementing inside shell scripts. Code that works from the commandline fails to parse when brought inside a shell script. At least this was my experience (which drove me to this approach).

Expanding on the answer from @Met, also using Perl:
If the input and output are TAB-delimited:

perl -F'\t' -lane 'print join "\t", @F[1, 0]' in_file

If the input and output are whitespace-delimited:

perl -lane 'print join " ", @F[1, 0]' in_file

Here,
-e tells Perl to look for the code inline, rather than in a separate script file,
-n reads the input 1 line at a time,
-l removes the input record separator (\n on *NIX) after reading the line (similar to chomp), and add output record separator (\n on *NIX) to each print,
-a splits the input line on whitespace into array @F,
-F'\t' in combination with -a splits the input line on TABs, instead of whitespace into array @F.

@F[1, 0] is the array made up of the 2nd and 1st elements of array @F, in this order. Remember that arrays in Perl are zero-indexed, while fields in cut are 1-indexed. So fields in @F[0, 1] are the same fields as the ones in cut -f1,2.

Note that such notation enables more flexible manipulation of input than in some other answers posted above (which are fine for a simple task). For example:

# reverses the order of fields:
perl -F'\t' -lane 'print join "\t", reverse @F' in_file

# prints last and first fields only:
perl -F'\t' -lane 'print join "\t", @F[-1, 0]' in_file