On Wed, Jan 2, 2013 at 8:04 PM, Drew Frank <[email protected]> wrote:

> The --joblog option saves a bunch of useful data that I would like to be
> able to access along with the saved results files. A natural way to do that
> would be to add a special value for --joblog (probably ":") that changes its
> behavior to create a separate log file for each result file. Then when
> running parallel with "--results out --joblog :", you end up with files
> like:
>
> out/[arg-value pairs]/stdout
> out/[arg-value pairs]/stderr
> out/[arg-value pairs]/joblog

Looking at the code this seems to be a fairly big change. Big change
often risk breaking other functionality or slowing down the code.

Most of the information can be gotten by wrapping each command (using
time and echo $? etc) and thus get the output in stderr.

I can see why it is beneficial to have the output in a standardized format.

So I am not against extending GNU Parallel with that, but since:

* I do not personally need the change
* The change is significant with a risk of breaking/slowing existing code

then I will not personally make the patch for this. In other words: If
I receive a patch and this patch does not break/slow down existing
code, then I will look favorably at it.


/Ole

Reply via email to