What should standard output of --results be?

[Ole Tange]

Having read https://savannah.gnu.org/bugs/index.php?51261 I am
wondering whether --results is doing the right thing.

Currently this:

  parallel --results res echo ::: foo

saves 'foo' into res/1/foo/stdout, after which GNU Parallel reads back
the file and outputs it to standard output.

It is easy to ignore standard output, if you do not need it:

  parallel --results res echo ::: foo > /dev/null

But it still means the file will be read.

The bug report is about this: they have a performance penalty by this
last reading which they send directly to /dev/null.

I cannot recall that I have ever used the output on standard output
from --results, so I am wondering if it is doing the right thing. If
you have used the output please weigh in and give your opinion.

--results should of course continue to save into the dir/args/stdout
file structure, so the question here is only what should happen to
standard output.

I see at least 3 alternatives:

* Keep as is

  $ parallel --results res echo ::: foo
  foo

* No output to standard output

  $ parallel --results res echo ::: foo
  <<nothing>>

* Output of the stdout filename to standard output (similar to --files)

  $ parallel --results res echo ::: foo
  res/1/foo/stdout

Do you have any thoughts?


/Ole