Re: [Nmh-workers] A permute command for nmh 1.7 ?

[Ralph Corderoy]

Hi Jererad,

> Putting the specified messages into the specified order is trivial,
> but where do they go in the overall order of the folder?

As Norm's clarified said, things are just being moved within the same
message numbers.

> If they are to be inserted into one another's slots,
> I see no sane way to define how that would work.

Let's get some sample arguments to this new script.

    $ set `seq 42 | shuf | sed 13q`
    $ echo $*
    25 4 3 31 29 41 9 13 2 26 23 20 10
    $

Norm wants the current 25 to sit in the smallest of those messages instead.

    $ l=`printf '%s\n' $*`
    $ echo "$l" | sort -n | fmt
    2 3 4 9 10 13 20 23 25 26 29 31 41
    $

So 25's content trample's 2's.  Obviously, we want 2's somewhere else
beforehand.

One can imagine backing up all the files and moving the backups to their
destination, but that creates a lot of duplication along the way.  An
alternative is to topologically sort the mapping of source → destination
and use that to do the copies in an order, breaking the cycles with a
backup copy when it's really necessary.  (There's tsort(1) but it
dislikes cycles.)

Here's where each has to end up.

    $ paste <(echo "$l") <(echo "$l" | sort -n) | awk '$1 != $2'
    25      2
    4       3
    3       4
    31      9
    29      10
    41      13
    9       20
    13      23
    2       25
    23      29
    20      31
    10      41
    $

So,

    mv  2   t
    mv  25  2
    mv  t   25

    mv  3   t
    mv  4   3
    mv  t   4

    mv  9   t
    mv  31  9
    mv  20  31
    mv  t   20

    mv  13  t
    mv  41  13
    mv  10  41
    mv  29  10
    mv  23  29
    mv  t   23

Cheers, Ralph.