Re: Improper behavior in sort utility

[Philip Rowlands]

On Wed, 30 Nov 2005, Mark Frost wrote:

> It appears that sort does not examine the proper fields specified by
> '-k' unless the '-n' option is given.

It would be helpful to know which sort locale is in use here; I'll 
assume it's en_US. Sort keys can be tricky, but I don't think any of the 
examples are bugs...

> localhost$ sort -s -k 3 sort_challenge6
> 1 0 0 1 3
> 0 1 0 2 1
> 0 0 0 3 5
> 1 1 0 4 7
> 1 1 1 1 2
> 1 1 1 2 6
> 0 0 1 2 9
> 0 1 1 3 4
> 1 0 1 4 8
> > > > > > > > > > This sort appears to sort by all fields following field 3

Yes, this is the intended behaviour. Use "-k 3,3" to sort only on field 
3.

> localhost$ sort -s -k 3 -k 5 sort_challenge6
> 1 0 0 1 3
> 0 1 0 2 1
> 0 0 0 3 5
> 1 1 0 4 7
> 1 1 1 1 2
> 1 1 1 2 6
> 0 0 1 2 9
> 0 1 1 3 4
> 1 0 1 4 8
> > > > > > > > > > The second key argument is ignored and results are identical
> to 'sort -s -k 3 sort_challenge6' above

This is equivalent to "-k 3,5 -k 5,5", therefore the keys overlap; the 
second argument is not ignored.

> localhost$ sort -n -s -k 3 -k 5 sort_challenge6
> 0 1 0 2 1
> 1 0 0 1 3
> 0 0 0 3 5
> 1 1 0 4 7
> 1 1 1 1 2
> 0 1 1 3 4
> 1 1 1 2 6
> 1 0 1 4 8
> 0 0 1 2 9
> > > > > > > > > > Adding the '-n' corrects this as well.

I think (don't quote me on this) that using -n is equivalent to adding 
an "n" suffix to each sort key, i.e. "-k 3,5n -k 5,5n". What's happening 
here is that the whitespace terminates the third field number, allowing 
the fifth field tie-break to kick in.


Cheers,
Phil