Re: [coreutils] cut feature request

[Laurent Lyaudet]

On 07/09/2010 19:52, Eric Blake wrote:
> On 09/07/2010 10:29 AM, Laurent Lyaudet wrote:
> > Hi,
> > it would be useful that cut handles negative numbers parameters for
> > columns and fields as "from the end" parameters.
> > One could select the last field with -f -1 for example. (Currently, -f
> > -1 do the same that -f 1.)
>
> Thanks for the report.  I also wonder if allowing negative end fields 
> for sort -k key specifiers also makes sense.
Thanks for your answer.
> > I didn't find a previous request that is similar in coreutils archive.
> > I think using awk or something similar is overkill for solving that
> > problem.
>
> On the contrary, we tend to take the conservative position that if 
> there is a POSIX-mandated tool that already does what you need, then 
> adding an alternative shortcut to some other tool to do the same thing 
> is very difficult to justify, on the grounds that you cannot rely on 
> our extension, but can rely on awk's behavior, when porting to 
> multiple systems.  In other words, the bar for 'convenience' features 
> is rather high before considering additions like this.
I didn't knew for the conservative position. I thought that, of course, 
it had to be POSIX compliant but there was no
restraint on additional features regarding compatibility with other 
POSIX-compliant systems.
> It may help us decide by showing a sample input file and a command 
> line using your proposed extension, along with any awk alternative to 
> do the same, to convince us that the verbosity required in awk is 
> worth such a shortcut (or conversely, that the awk script is already 
> so simple as to not be worth changing cut, as has already been our 
> argument for the numerous times that people suggest teaching cut how 
> to reorder fields).
I'm sorry I don't have a decisive example. All I'm using is a sequence 
of piped commands on csv and text files like
... | awk -F'/' '{print $FN}' | ...

I would prefer :
... | cut -d'/' -f -1 | ...

If I had to handle flat file and cut the last 5 characters then the 
example would be more convincing
awk '{print substr($0,length($0)-5)}'
instead of
cut -c -5

I see two other arguments for these feature : cut is compiled command, 
awk program is interpreted;
if reading a script two years after writing it, the "what was I doing 
here? effect" (or the
"how was I doing it? effect" if there is a minimal amount of commentary 
in the code) will
be decreased.
> > Let me know what you think about it.
>
> If we do it, I'd like to see it done uniformly across all the 
> coreutils that take partial field specifications.  And although I 
> probably won't be the one writing such a patch, it does sound like it 
> still might be justifiable.
Thanks for considering my request.