Re: sort feature requests

[Matthew Woehlke]

Balaji S. Srinivasan wrote:
> The coreutils, particularly the text utilities, are an amazing piece of 
> work and crucial for our group's research efforts at Stanford.
> Thank you for taking the time to make this code robust and high quality.
>
> Along these lines, I have a suggestion for a minor improvement. Would it 
> be possible to add the following features to the next version of GNU sort?
>
> 1) Allow the first line of the input to be retained without sorting 
> (e.g. to keep the header of a tab-delimited text file at the top of the 
> file)

This should already be possible with a command such as:
$ ( sed 1q ; sort [args] ) < [file]

...however it seems my version of sed (4.1.2-RH4) does not honor POSIX 
semantics and consumes too much of the input ('head' from coreutils-6.9 
seems to have the same problem?) I did get 'read 4 ; echo $a' to work 
however.

(Ok gnulib folks, I don't have any dates handy, is the fseek stuff newer 
than coreutils-6.9?)

> 2) Allow the specification of tabs as separators with the -t option 
> (this is useful when tab separated fields contain embedded spaces). 
> Perhaps this is currently possible, but '\t' does not seem to work.

If your shell is reasonable, it will let you enter a literal tab 
character. [ctrl]-[v], [tab] seems to work for me (in bash).

> There are various ways to achieve this functionality by using shell 
> scripts or complex pipelines, but it would be great if flags
> enabling these features could be added to the next version.

I don't see how either of the above amounts to 'complex pipelines'. The 
general philosophy seems to be to not add options when a simple, modular 
solution works, therefore I suspect that without greater justification, 
neither of these features is likely to be added.

--
Matthew
"...anyway, that's my 0.02 toasters"
(with apologies to Niel)