Linux sort with 'nested' fields

Question

I have a big file with lines of the following format:

field1:field2:field3:field4:field5

I want to sort on field4. However, field4 itself is 'subdivided' like this:

field41_field42_field43_field44

The command to sort field4 is

sort -t_ -k1,1d -k2,2g -k3,3g -k4,4g

So what I have to do is split each line by : and then split the fourth field by _ and perform the specified sorting on it. Is there an easy way to do this?

A small subset of the data can be found here.

Answer 1

I don't like this myself - you could perform a Schwartzian transform by temporarily copying the fourth field to the beginning of the record, sort by it, and then remove it

awk -F':' -vOFS=':' '{print $4,$0}' file | 
sort -k1,1d -k2,2n -k3,3n -k4,4n -t"_" | 
cut -f2- -d":"