Re: Input pattern

[Ali Roustaei]

This is great that it works to some extent in the git version. For my
application I don't need wrapping behavior too and probably there are a
bunch of others who don't need. I'll check the git version, thanks.

On Sat, Mar 26, 2016 at 12:46 PM, Ole Tange  wrote:

> On Wed, Mar 23, 2016 at 6:07 PM, Ali Roustaei 
> wrote:
> > On Tue, Mar 22, 2016 at 1:47 AM, Ole Tange  wrote:
> >> On Mon, Mar 21, 2016 at 4:37 PM, Ali Roustaei 
> >> wrote:
> :
> >> > My script (say myscript) has three input parameters H, Re and Bn. Re
> and
> >> > Bn
> >> > are tied so that for each Re input there is only a specific Bn.
> :
> >> Currently this is not possible using a single command, but I have had
> >> this problem, too, so I would welcome a patch that made it possible.
> :
> >> The idea being that :::+ links the following input source to the
> >> previous --xapply style.
> :
> >> For your current situation this may work:
> >>
> >>   parallel --xapply -I ,, --arg-sep ,,, parallel eval echo myscript {}
> >> ,, ::: A B ,,, C D E ,,, F G H
> :
> > Thanks Ole. So right now it is not a built in feature. And thanks for the
> > workaround solution, it does its job for me right now.
>
> The git version now has :::+ and + They are limited in the way
> that values do not wrap, instead there must be the same amount of
> values. It is the way I use it most of the time, and it was not
> possible for me to get the wrapping to work.
>
> The magic happens in expand_combinations().
>
> /Ole
>

Re: Citation requirement and the GPL

[Martin Møller Skarbiniks Pedersen]

On 27 March 2016 at 23:35, Christopher Jefferson 
wrote:


> I am happy to accept that, but I feel that this requirement violates
> the GPL, the GPL FAQ specifically covers this case:
>
>
Nope.
I don't think you are right saying that the GPL FAQ in this section
http://www.gnu.org/licenses/gpl-faq.en.html#RequireCitation
covers the GNU parallel case.
The difference is that the GPL FAQ uses the words "_require_ citation"

However
GNU parallel says _please. It does not say you _must_ or _shall_. It says
"please" and Ole wrote that you "should". He didn't know write "must".


More precise GNU parallel says as the first line:
"Academic tradition requires you to cite works you base your article on."
I think and hope that you agree on that statement.

And the second line:
"When using programs that use GNU Parallel to process data for publication
please cite:"

_please cite_

Regards
Martin M. S. Pedersen

Re: Citation requirement and the GPL

[Christopher Jefferson]

On 27 March 2016 at 18:26, Ole Tange  wrote:
> On Sun, Mar 27, 2016 at 11:20 AM, Christopher Jefferson
>  wrote:
> :
>> Would I cite 'parallel' if it was an integral part of my research
>> project (for example, a parallelisation project)? Of course! Further,
>> if I reported times taken I would cite parallel, as it would be an
>> integral part of recreating those times.
>>
>> If I used parallel to simply speed up a set of analysis of data? Then
>> I would not cite it, in the same way I wouldn't cite xargs, grep, cut,
>> awk and bash, which might be used in the same scripts.
>
> As the speedup of analysis is a common use case, GNU Parallel
> (--citation) mentions this situation in this way:
>
>   When using programs that use GNU Parallel to process data for
> publication please cite: [...]
>
> Do you feel the wording is unclear? If so: What should it say to
> convey the message that you should cite when using GNU Parallel for
> speeding up of analysis of data?

If that is what you want, then the current wording is fine.

> You do not find similar wording in xargs, grep, cut, awk and bash -
> thus no need to cite those.

I personally disagree with this -- I do not cite software or papers
because their authors want citations. I cite because I believe the
citation helping users understand my paper, reproduce it's results, or
place it in context.

> If you prefer not to cite GNU Parallel at all, then please refrain
> from using GNU Parallel.

I am happy to accept that, but I feel that this requirement violates
the GPL, the GPL FAQ specifically covers this case:

http://www.gnu.org/licenses/gpl-faq.en.html#RequireCitation

Perhaps parallel should be changed to another, less free, license?
Otherwise I could simply take parallel and delete the request for
citation?

Chris

Re: GNU Parallel 20160322 ('Bruxelles') released

[Shlomi Fish]

Hi Ole,

thanks for your reply.

On Sun, 27 Mar 2016 20:00:57 +0200
Ole Tange  wrote:
> All in all I come to the conclusion that the benefits are small and
> the downsides are just a bit bigger than the benefits. So for the
> foreseeable future there will not be an extra format.
> 

I see. I understand and accept that.

Regards,

Shlomi Fish

> 
> /Ole
> 



-- 
-
Shlomi Fish   http://www.shlomifish.org/
Freecell Solver - http://fc-solve.shlomifish.org/

Chuck Norris helps God help those that help themselves.
— http://www.shlomifish.org/humour/bits/facts/Chuck-Norris/

Please reply to list if it's a mailing list post - http://shlom.in/reply .

Re: Using --dryrun records to the joblog as a successful event, which ignores "real" runs.

[Ole Tange]

On Mon, Sep 21, 2015 at 2:07 PM, Pariksheet Nanda
 wrote:

> I'm using GNU Parallel version 20150810 when both --joblog and --dryrun
> together.  This has the side effect of recording the dryrun as "successful"
> in the joblog.

Fixed in current release.

/Ole

Re: GNU Parallel 20160322 ('Bruxelles') released

[Ole Tange]

On Thu, Mar 24, 2016 at 3:11 PM, Shlomi Fish  wrote:
:
> One note: can the tarball also be supplied as tar.xz? I checked and it reduces
> the size quite a bit.

I have considered that.

I get 1053860 (xz) vs. 1265407 (bz2). So a saving of 200 KB. Less than
a photo these days.

GNU Parallel prides itself on being compatible with very old operating
systems. These systems do not have an xz unpacker so the .xz version
would have to be in addition to the .bz2 version. The xz-file takes
around 80% extra disk space on the ftp-mirrors.

These days most people will be getting GNU Parallel as a package from
their distribution (thanks to you amongst others), and this package
will be compressed using whatever the distribution uses. So these
people will not see any difference.

The packagers (such as yourself) will see the difference, but there
are not very many of you so the possible bandwidth saving may eat
itself in the extra mirroring that has to happen to the mirrorsites.

Then there are people who downloads the source themselves because they
need the newest version. With the exception of people on this list
they most likely will not download every single release, but instead
only download when they need a new feature/bugfix. My belief is that
these are also relatively few, and the possible saving in bandwidth
has to be measured against the extra harddisk space that an xz-file
will take - and keep taking also in the future.

Finally there is the hassle of setting up and keeping another format
working to save 200 KB per release.

All in all I come to the conclusion that the benefits are small and
the downsides are just a bit bigger than the benefits. So for the
foreseeable future there will not be an extra format.


/Ole

Re: Citation requirement and the GPL

[Ole Tange]

On Sun, Mar 27, 2016 at 11:20 AM, Christopher Jefferson
 wrote:
:
> Would I cite 'parallel' if it was an integral part of my research
> project (for example, a parallelisation project)? Of course! Further,
> if I reported times taken I would cite parallel, as it would be an
> integral part of recreating those times.
>
> If I used parallel to simply speed up a set of analysis of data? Then
> I would not cite it, in the same way I wouldn't cite xargs, grep, cut,
> awk and bash, which might be used in the same scripts.

As the speedup of analysis is a common use case, GNU Parallel
(--citation) mentions this situation in this way:

  When using programs that use GNU Parallel to process data for
publication please cite: [...]

Do you feel the wording is unclear? If so: What should it say to
convey the message that you should cite when using GNU Parallel for
speeding up of analysis of data?

You do not find similar wording in xargs, grep, cut, awk and bash -
thus no need to cite those. But GNU R has 'citation()' which says:

  We have invested a lot of time and effort in creating R, please cite
it when using it for data analysis.

It should therefore be cited when used for data analysis - even if you
could have used another tool to do the same. The same goes for Numpy
(citation as per http://www.scipy.org/citing.html) and Python.

As mentioned in --citation you should feel free to use
\nocite{Tange2011a} - in other words: Your article does not have to
mention GNU Parallel, as long as it is in the list of references.

If you prefer not to cite GNU Parallel at all, then please refrain
from using GNU Parallel.

The reason for citations is covered in length in:
https://lists.gnu.org/archive/html/parallel/2013-11/msg6.html


/Ole

Re: Citation requirement and the GPL

[Hans Schou]

2016-03-27 11:20 GMT+02:00 Christopher Jefferson :


> If I used parallel to simply speed up a set of analysis of data? Then
> I would not cite it, in the same way I wouldn't cite xargs


This is actually what 'parallel' does for most people: Replacing xargs to
get more speed

parallel in it self does not do anything useful - you have to add a command
to get some usefulness.

./hans