Re: Support bytesize comparison in sort

[Andrew D Jewell]

> I like the idea, but I object to using up yet another single-letter
> option for this (they're not renewable - the ASCII character set has a
> fixed size).  I suggest that we use a long option only.

I agree halfway. As a separate command line argument, using up 
another single character might (possibly) be too aggressive.

However, we do need to allocate a single character for it as part of 
the -k field specification; otherwise, as you said, how can you sort 
some fields this way and some fields that way. I'd recommend h, to 
match the -h of ls and df.


As for the implementation, unless I'm confused, it only deals with 
K,M and G. It should at least support T if not P, E, Z and Y.

I don't know if this is important, but I also don't see anything that 
would normalize numbers, so comparing 1000000 to 1M wouldn't do the 
right thing. To fix that would open another can of worms, in that 
you'd need to know if your M was 1000*1000 or 1024*1024.


adj