UNIX for Dummies Questions & Answers

Can sort sort by pair of lines?
infile:

Code:

ID 15
GNJSMKSNS
ID 25
GNJSMKSNS
ID 1
GNJSMKSNS

outfile:

Code:

ID 1
GNJSMKSNS
ID 15
GNJSMKSNS
ID 25
GNJSMKSNS

sort -grk 2 obviously does not work

No. The sort utility sorts lines; not groups of lines. You can however, join pairs of lines, sort the result, and then split the sorted output into pairs of lines again. Assuming that there aren't any tab characters in your input file, a simple way to do it would be:

Code:

awk '{printf("%s%s",$0,(NR%2)?"\t":"\n")}' infile|sort -k2,2n|tr '\t' '\n'

which with the following infile contents (modified so we can be sure that the correct even numbered lines in the output follow the same lines they followed in the input file):

Code:

ID 15
GNJSMKSNS 1
ID 25
GNJSMKSNS 2
ID 1
GNJSMKSNS 3

produces the output:

Code:

ID 1
GNJSMKSNS 3
ID 15
GNJSMKSNS 1
ID 25
GNJSMKSNS 2

Of course, if all of the even numbered lines in your input files are identical (as in your sample); you could delete the even numbered lines, sort the remaining lines, and reinsert the deleted lines.

This is what I was looking for

Code:

 awk '{printf("%s%s",$0,(NR%2)?"\t":"\n")}' input|sort -rk 3|tr '\t' '\n'

Could you please explain me the following parts of the code:
"%s%s"
?
:

awk: Use awk
': with the script:
printf(: print
"%s%s": two strings
,$0: where the 1st string is the current input line (without the trailing <newline> character)
,(NR%2)?"\t":"\n"): and the 2nd string is a <tab> character if the current input line is an odd numbered line and the 2nd string is a <newline> character if the current input line is an even numbered line
': ending the awk script
input: naming the file to be processed by awk
|sort -rk 3: sorting the output from awk using a sort key that is a decreasing alphanumeric sort starting with the 3rd field on the line, and for lines that match on the 3rd field uses a decreasing alphanumeric sort on any remaining fields after the 3rd field, and for lines that match on the 3rd field to the end of the line perform an decreasing alphanumeric sort on field 1, and for lines that still match, perform an decreasing alphanumeric sort on field 2
|tr '\t' '\n': and, finally, converting the <tab> characters in the joined lines sorted by sort back into <newline> characters (thereby splitting the lines apart again).

Note: I originally said that fields 1 and 2 would be sorted in increasing order (which would have been true with sort -k 3r), but, with sort -rk 3, all fields will be sorted in decreasing order.

Hi.

Some versions of paste can do the stitching job:

Code:

$ paste - - < input
ID 15	GNJSMKSNS 1
ID 25	GNJSMKSNS 2
ID 1	GNJSMKSNS 3

Then the sort and severing of the super-lines. See man paste for details.

This was on a system:

Code:

OS, ker|rel, machine: Linux, 2.6.26-2-amd64, x86_64
Distribution        : Debian 5.0.8 (lenny, workstation) 
paste (GNU coreutils) 6.10

Best wishes ... cheers, drl