[petsc-dev] [petsc-maint #80119] Re: make install broken on unix?

Lisandro Dalcin dalcinl at gmail.com
Mon Aug 1 21:36:10 CDT 2011


On 30 July 2011 16:18, Jed Brown <jedbrown at mcs.anl.gov> wrote:
> Does that mean we can just add python utilities to that package, or do we
> need to create petsc-util?
>

Currently, if you "pip install petsc", pip will download PETSc's
tarball release (the -lite one), configure, build and install (in
Python's site-package directory). Most of the installed stuff are
header and libraries, and no Python code. Remember this is just a hack
after Barry's loud insistence to get petsc4py easy-installable, even
if the user do not have a previous PETSc install. So they get what
they deserve...

So, I think this package is not appropriate for adding utilities.

I think your only option is to set a lawsuit on me for using the
"petsc" name for my own purpose :-) After we are somehow in software
business, right?. After being asked to pay millions of dollars, I will
eventually give up and start using "libpetsc" for what petsc4py needs
as a dependency...


> On Jul 30, 2011 12:07 PM, "Lisandro Dalcin" <dalcinl at gmail.com> wrote:
>> On 25 July 2011 19:33, Jed Brown <jedbrown at mcs.anl.gov> wrote:
>>> On Mon, Jul 25, 2011 at 10:16, Ethan Coon <ecoon at lanl.gov> wrote:
>>>>
>>>> It's not really a stand-alone script though, so it shouldn't get the
>>>> hashbang or lose the extension.  But it's not a full-fledged package of
>>>> libraries either.  I don't think it should be anywhere in the default
>>>> path
>>>> -- it's the kind of code that you should have to explicitly include
>>>> (either
>>>> via sys.path manipulation or via PYTHONPATH) because too many little
>>>> scripts
>>>> of utility code makes for a nasty, polluted namespace.  There just isn't
>>>> a
>>>> good place (that I know of) in the standard file system hierarchy for
>>>> this
>>>> kind of stuff.
>>>
>>> Making the user/scripts do manual sys.path or PYTHONPATH manipulation
>>> makes
>>> it nearly impossible to make portable scripts that depend on these
>>> functions. Should we start a "proper" Python package containing PETSc
>>> utilities? Then scripts could rely on
>>> from petsc.io import binary_read
>>> or whatever working. Of course it might only contain two functions right
>>> now, but I could add my "petscplot" (makes various types of convergence
>>> and
>>> scaling plots from convergence monitors and -log_summary output), and I'm
>>> sure the community has other useful scripts.
>>
>> Note however that "petsc" package is already registered in PyPI to
>> make "pip install petsc4py" download and install PETSc.
>>
>>
>> --
>> Lisandro Dalcin
>> ---------------
>> CIMEC (INTEC/CONICET-UNL)
>> Predio CONICET-Santa Fe
>> Colectora RN 168 Km 472, Paraje El Pozo
>> 3000 Santa Fe, Argentina
>> Tel: +54-342-4511594 (ext 1011)
>> Tel/Fax: +54-342-4511169
>



-- 
Lisandro Dalcin
---------------
CIMEC (INTEC/CONICET-UNL)
Predio CONICET-Santa Fe
Colectora RN 168 Km 472, Paraje El Pozo
3000 Santa Fe, Argentina
Tel: +54-342-4511594 (ext 1011)
Tel/Fax: +54-342-4511169



More information about the petsc-dev mailing list