Re: --linebuffer implemented

[Svenne Krap]

Hi all.

Sorry for the long delay. Ole poked me semi-recently so here is the
reason for it now...

We use parallel to start some jobs simultaneously. The jobs each write
to stdout (full lines) of status rather infrequently (i.e. around 5
lines / minute / process), but we save the combined output (verbatim) as
documentation and hence cannot accept inter-line interleaving from
multiple processes.

The processes run for quite a while (days sometimes), so status
(including estimation on completion time) while running is highly
desirably (we run it through tee later).

That is the setup for the feature request to Ole.

Svenne

p.s. I know this is probably the use case for a full fledged queue, but
sadly we don't have ressources to implement that in the near future...

On 30-07-2013 17:20, Ole Tange wrote:
> --linebuffer is now implemented in git clone
> git://git.savannah.gnu.org/parallel.git
>
> It is sort of in between --ungroup and --group:
>
> --ungroup does not guard against having half a line coming from job1
> and half a line coming from job2, but prints out immediately.
>
> --group only prints out when the job is finished - thus jobs will not be 
> mixed.
>
> --linebuffer guards against having half a line coming from job1 and
> half a line coming from job2, but prints out when it has a full line.
> It is slower than --group and much slower than --ungroup.
>
> I will let Svenne Krap (who asked for the option) to come up with a
> good example of when this is needed.
>
>
> /Ole
>