How to sort data? I have a set of data that I wish to sort by using a Unix script. The data is similar to the set given below. Columns represent x-, y- and z-coordinates and are supposed to be sorted considering y-coordinates. It is important that each block is sorted individually and that the separating blank row is kept. Data set should be sorted: From 100.000 23.000 150.000 99.000 83.000 369.000 110.000 15.000 123.000 25.000 23.000 23.000 9.000 63.000 81.000 15.000 38.000 23.000 to 110.000 15.000 123.000 100.000 23.000 150.000 99.000 83.000 369.000 25.000 23.000 23.000 15.000 38.000 23.000 9.000 63.000 81.000 Data is retrieved from data.dat and is to be re-entered into data.dat. How can I do this? regards bjorb

Perhaps I cheated, but I split bigfile after line 3 on the first blank: csplit -f ss bigfile /^/3 This created two files, ss00 & ss01 I then sorted on the second field: sort -k 2,2 ss00 > bigfile sort -k 2,2 ss01 >> bigfile

Thanks for a quick reply Nails! It would have worked but maybe I should have mentioned that there are more blocks following the two first. So it would be quite elaborate to use the method you presented. I hope there is a simpler way to do this, if there is any.

This solution prepends a 4-digit group number to each line to control the first sort level.  That data is then piped into the sort command, then the sorted data is piped into cut to delete the first 5 characters of each line. I added group number as fixed-width 4 digits, not for the benefit of the sort, but because it makes it a bit easier to chop it off afterward. Your second fields all have the same decimal alignment, but in case they do not, you need to specify a numerical sort for that sort key, as I have. This code outputs to data.new. Your script would need to rename it to data.dat. awk '{ if (NF==0)    group++ printf "%4.4d %s\n",group,$0 }' data.dat | sort -k1,1 -k3,3n | cut -c-6- > data.new 0000 100.000 23.000 150.000 0000 99.000 83.000 369.000 0000 110.000 15.000 123.000 0001 0001 25.000 23.000 23.000 0001 9.000 63.000 81.000 0001 15.000 38.000 23.000 cat data.new 110.000 15.000 123.000 100.000 23.000 150.000 99.000 83.000 369.000 25.000 23.000 23.000 15.000 38.000 23.000 9.000 63.000 81.000

Sorry it took so long to get back to you. You can try this method: csplit -ks -f xss bigfile '/^$/' {9} rm -f bigfile ls -1 xss*|while read y do echo $y sort -k 2,2 "$y" >> bigfile done # end script I've assumed the file has no more than 9 blocks with {9}. All you have to do is make sure that number is greater than the number of blocks you'll ever have. You'll get an 'out of range' error on the csplit command, but the -k flag saves the created files.