Bug#847524: ITP: distributed -- Dask Distributed computing for Python
On Fri, 2016-12-09 at 10:42 +0300, Konstantin Khomoutov wrote:
> On Thu, 08 Dec 2016 16:20:15 -0800
> Diane Trout <diane@ghic.org> wrote:
>
> > Package: wnpp
> > Severity: wishlist
> > Owner: Diane Trout <diane@ghic.org>
> >
> > * Package name : distributed
>
> The package's name is really unfortunate: the word "distributed" is
> too
> broad, and in our era of everything distributed squatting on it is a
> very bad idea IMO. Please consider naming it dask-distributed
> or python-dask-distributed instead.
I agree its a terribly generic name.
Usually, debianized python packages use the upstream name foo as the
source name and then python-foo as the binary name. especially when foo
is the module you import in python.
But considering distributed is so generic I too was tempted to name the
python-dask-distributed. I asked on #debian-python, but they suggested
use the name you use in the import.
I'll see if I can get some of the more experienced people in the Python
Modules team to weigh in on the name.
Upstream sometimes uses dask.distributed in the documentation, which
might be a better choice.
> If this bit:
>
> > This is the distributed component of dask, and looks really useful.
> >
> > Who wouldn't want some APIs that parallize over multiple machines
> > random python code, numpy and pandas?
> >
>
> ...is a part of the extended package's description, it should be
> reworded: package descriptions should be neutral, dry and up to the
> point. That is, passages like "looks really useful" and "Who
> wouldn't
> want to use this piece of software?" are unfit for the purpose.
Nope those were my personal comments in response to the reportbug wnpp
template prompt "why is this package useful".
This is what's in my proposed control file right now.
Package: python3-distributed
Architecture: all
Depends: ${misc:Depends}, ${python3:Depends}
Recommends: python3-dask
#Suggests: python3-bokeh
Description: Dask Distributed computing for Python 3
Dask.distributed is a lightweight library for distributed computing in
Python. It extends both the concurrent.futures and dask APIs to
moderate sized clusters.
.
This contains the Python 3 version
The next line introduces the -doc package. the bokeh suggests is there
because it adds additional functionality, but its commented out because
getting bokeh in involves dealing with javascript builds
It is somewhat pedantic to be referring to the python 3 version, as I'm
not currently including the python 2 version.
I installed my current package on my works cluster, if you want to look
at it here's the .dsc
http://woldlab.caltech.edu/woldlab-apt/pool/main/d/distributed/distribu
ted_1.14.3-1.dsc
Also thank you for taking the time to review my ITPs. I didn't realize
that people watched the bugs.
Diane
Reply to: