bug#9455: RFE: split --balanced

[Bob Proulx]

severity 9455 wishlist
thanks

Hi Dave!

Dave Yost wrote:
> Z% for x in 1 2 3 4 5 6 7
> for> do echo $x ; done | split --lines=3 \
> pipe> && for x in x?? ; do echo "=== $x" ; cat $x ; done

Sure.  Just fyi but GNU seq can produce sequences of numbers very
easily.  I think this is a little more concise example.

  seq 1 7 | split --lines=3
  head *

But do you always mean stdin?  Or mostly is this a file?  Because...

> In some applications, you would like split to more evenly apportion
> the output to the files, like this:
> 
> Z% for x in 1 2 3 4 5 6 7
> for> do echo $x ; done | split --balanced --lines=3 \
> pipe> && for x in x?? ; do echo "=== $x" ; cat $x ; done
> === xaa
> 1
> 2
> 3
> === xab
> 4
> 5
> === xac
> 6
> 7

I think it would be really hard to know if the user wanted the extra
lines in the first file.  It would be easier to gather that last widow
line up into the next to last file.  Or easier to leave it alone in
its own file.

  seq 1 7 > input.txt
  numsplits=3
  num=$(wc -l < input.txt)
  perfile=$(($num / $numsplits))
  split --lines=$perfile < input.txt
  head x*

  seq 1 17 > input.txt
  numsplits=3
  num=$(wc -l < input.txt)
  perfile=$(($num / $numsplits))
  split --lines=$perfile < input.txt
  head x*

And from there you could get creative and make a decision based upon
the number of lines in that last file.

  if [ $(($num % $numsplits)) -lt $(($num / 2)) ]; then ...

If the widow lines are less than half the number of total lines in the
files then that last file could be concatenated into the next to last
file, the implementation of which I will leave as an exercise.  Just
as an example.  Or they could be put in the first file.  But I think
what the user would want in something like this is so varied that
there isn't any one natural result.  So I think this is better left to
the caller to decide.

Bob