Re: [coreutils] added ability in sort to skip n number of lines for each file

[Pádraig Brady]

On 22/11/10 17:28, Pádraig Brady wrote:
> On 18/11/10 16:36, Jim Hester wrote:
>> A common problem when sorting files stems from the file containing 1
>> or more header lines, which should not be sorted.  As of now, the
>> common solution to this problem is to remove the header lines with
>> manually, or to output only the non header lines with tail, awk, or
>> some other program and pipe the results to sort.
> 
> Thanks for the patch!
> 
>> This was likely not
>> deemed a problem when sort was only single threaded, as the printing
>> and pipe was likely still faster than the sort itself.  However with
>> multi-threaded sort this results in the operation bottle necking
>> waiting for more information from the pipe.
> 
> I'm not following the argument above.
> One can always print the header synchronously?
> I.E. the `head` below is guaranteed to run before the `sort`
> 
> printf "z_header\nb\na\n" > file
> (head -n1 file; sort <(tail -n+2 file) <(tail -n+2 file))
> 
> Now the above is awkward and dependent on bash
> (constructs per file), so your idea has some merit I think.

Note the --header option is especially useful for `join`
as it transforms its input, however sort does not and
so might be amenable to a more general solution.
Perhaps something like:

(head --no-header -n1 file.* | head -n1; tail --no-header -n+2 file.* | sort)

I.E. add the --no-header option to suppress the ==> file name <== annotations
which would allow using `head` and `tail` in general for this.

thanks,
Pádraig.