Re: Question 2 (argfile input syntax)

[Ole Tange]

On Wed, May 17, 2017 at 10:42 PM, paralleluser
<paralleluser@fastmail.net> wrote:

> Question 2:
> There are a multitude of ways to give GNU parallel arguments contained in a 
> file.
>
> 1. cat argfile | parallel
> 2. parallel < argfile
> 3. parallel :::: argfile
> 4. parallel ::: `cat argfile`
> 5. parallel -a argfile
>
> Am I missing any from this list?

The linked versions: :::+ and ::::+, and maybe the using of stdin as -

  cat argfile | parallel echo ::: 1 2 3 :::: -

> I am curious as to which you use when and why.  I think the answer is beyond 
> "personal preference".  For example, I would ...
>
> Use #1 if you plan to do additional pipe work before parallel, for example:
>   cat argfile | grep xyz | parallel

Yup.

> Use #5 if you are going to use "-a" multiple times so you can refer to {1} 
> and {2} etc
>   parallel -a argfile1 -a argfile2  command {1} {2}

Or use #3:

  parallel command {1} {2} :::: argfile1 argfile2

I find that easier to read - especially when mixed with :::

  parallel command {1} {2} {4} {3} :::: argfile1 argfile2 ::: 1 2 3
:::: more files

> I don't see a good reason to use #4

If you use #4 you need to know what you are doing:

  echo 1 2 > argfile
  parallel echo ::: `cat argfile`

So in the exceptional case where you actually want the line split on
space, it is a solution.

But in most cases #3 is the better (and shorter) choice.

> I think #2 and #3 seem about the same.  What advantage is there to use either 
> of them over #1?

#2 has the advantage of not running a cat (but if you are concerned by
that, please read:
http://oletange.blogspot.dk/2013/10/useless-use-of-cat.html )

#3 has the advantage that GNU Parallel sees this as an actual file. So
you can use --pipepart:

  parallel --block -1  --pipepart wc :::: bigfile


/Ole