On Thu, Jul 30, 2015 at 11:08 PM, Schweiss, Chip <c...@innovates.com> wrote:
> I have a process that I want run on every folder on a file system. The
> operation has to happen from the deepest child folder first then back to the
> parent. Some folders will have many folders, which is where parallel fits
> well. However, all child folders must be processed before their parent.
>
> Using a recursive script works well so that the operations on all the
> children always happen before the parent no matter how deep the folder tree
> goes. But when combined with 'parallel' in some cases I end up with too
> many processes running at once, choking the system.
>
> Is there a way so that multiple 'parallel' processes become aware of each
> other and never run more than a specified number of processes at once, yet
> still satisfies the dependency order? Or maybe there's a better way to
> approach this problem?
If you use 'find . -type d' you will get it in the exactly wrong
order. So how about simply reversing that:
find . -type d | tac | parallel myprog
/Ole
