Re: sort -nu: bug or feature?

[Chris F.A. Johnson]

On Wed, 8 Sep 2004, Andreas Schwab wrote:

> Andrew Noymer <address@hidden> writes:
>
> > sort -nu seems to compare only the first key, which is not what a lot of
> > people expect.  I don't know if this is a bug, though.
>
> sort -n takes the selected fields (the whole line if no -k is used) and
> tries to parse a number out of each starting from the beginning of the
> field, ignoring leading whitespace and discarding trailing garbage.  The
> resulting numbers are the sort keys, which also decide which lines are
> equivalent for -u.
>
> > But I think most folks expect that "sort -nu" == "sort -n | uniq"
>
> The expectation is wrong because sort -n and uniq use different sort keys..
> Especially uniq is always using a string comparison.  Compare this:
>
> $ printf "%s\n" 1 01 | sort -nu
> 1
> $ printf "%s\n" 1 01 | sort -n | uniq
> 01
> 1

    "sort -n" uses a string comparison after aligning the numbers on
    the decimal point, but -u apparently uses a (non-POSIX) numeric
    comparison to determine equivalency:

$ printf "%s\n" 1 01 08 8 | sort -nu
1
08


    Also, there's a bug in the documentation; at the end of the man
    page:

-------------
       The full documentation for sort is maintained as a Texinfo
       manual.  If the info and coreutils sort programs are properly
       installed at your site, the command

              info coreutils sort

       should give you access to the complete manual.
------------------

    This brings up the page on sorting the output of ls; to get the
    info page for sort, use "info sort".


--
        Chris F.A. Johnson                      http://cfaj.freeshell.org
        =================================================================
                Everything in moderation -- including moderation