Robert Kern wrote:
It is David's desire to
distribute numpy builds with optimized BLASes that does not fit into
eggs, not numpy. Plain numpy eggs are really straightforward.
There may be a misunderstanding: I would not mind distributing something
without optimized BLAS. Not using atlas at
On Thu, Apr 16, 2009 at 01:27, David Cournapeau
da...@ar.media.kyoto-u.ac.jp wrote:
Robert Kern wrote:
It is David's desire to
distribute numpy builds with optimized BLASes that does not fit into
eggs, not numpy. Plain numpy eggs are really straightforward.
There may be a misunderstanding: I
Fadhley Salim wrote:
I've been asked to provide Numpy Scipy as python egg files.
Unfortunately Numpy and Scipy do not make official releases of their
product in .egg form for a Win32 platform - that means if I want eggs
then I have to compile them myself.
I think having a simple .exe
On Thu, Apr 16, 2009 at 12:50:59AM -0500, Robert Kern wrote:
PEAK's setuptools is fragile with complex packages. This is the core
reason numpy is not distributed as an egg, as the implications on what
numpy would have to do to 'fit' in an egg compromise numpy's instal
quality (read: random
Robert Kern wrote:
I have found that people are more willing to accept that they have to
do something different to get a technically chalenging feature (i.e.
use a particular installer to get a safe, optimized BLAS) than to
accept that a relatively straightforward feature is not available at
David Cournapeau wrote:
Robert Kern wrote:
I have found that people are more willing to accept that they have to
do something different to get a technically chalenging feature (i.e.
use a particular installer to get a safe, optimized BLAS) than to
accept that a relatively straightforward
On Thu, Apr 16, 2009 at 01:52, David Cournapeau
da...@ar.media.kyoto-u.ac.jp wrote:
David Cournapeau wrote:
Robert Kern wrote:
I have found that people are more willing to accept that they have to
do something different to get a technically chalenging feature (i.e.
use a particular installer
Robert Kern wrote:
easy_install is documented to be able to find and convert a
bdist_wininst .exe on the fly, so I believe that should be sufficient.
Yes, I've tried locally on a bdist_wininst exe, and easy_install could
install it. I have not tested it from the network.
It might be
On Thu, Apr 16, 2009 at 04:02:01PM +0900, David Cournapeau wrote:
What about just pushing the non optimized bdist_wininst installer on
pypi ?
With a clear note saying that they are non optimised.
G.
___
Numpy-discussion mailing list
On Thu, Apr 16, 2009 at 02:02, David Cournapeau
da...@ar.media.kyoto-u.ac.jp wrote:
Robert Kern wrote:
easy_install is documented to be able to find and convert a
bdist_wininst .exe on the fly, so I believe that should be sufficient.
Yes, I've tried locally on a bdist_wininst exe, and
Hi all,
I have orthonormal set of vectors B = [b_0, b_1,..., b_k-1],
b_i from R^n (k may be less than n), and vector a from R^n
What is most efficient way in numpy to get r from R^n and c_0, ...,
c_k-1 from R:
a = c_0*b_0+...+c_k-1*b_k-1 + r
(r is rest)
Thank you in advance, D.
Hi all,
I have orthonormal set of vectors B = [b_0, b_1,..., b_k-1],
b_i from R^n (k may be less than n), and vector a from R^n
What is most efficient way in numpy to get r from R^n and c_0, ...,
c_k-1 from R:
a = c_0*b_0+...+c_k-1*b_k-1 + r
(r is rest)
Thank you in advance, D.
On 4/16/2009 5:06 AM dmitrey apparently wrote:
I have orthonormal set of vectors B = [b_0, b_1,..., b_k-1],
b_i from R^n (k may be less than n), and vector a from R^n
What is most efficient way in numpy to get r from R^n and c_0, ...,
c_k-1 from R:
a = c_0*b_0+...+c_k-1*b_k-1 + r
(r is
I just installed the latest stable mingw, and made sure that the mingw
bin directory is in my PATH. I used the command you suggested and got
the following output:
http://pastebin.com/m4aea512c
Any suggestions as to what I might be doing wrong?
I'm using standard cpython 2.4.4 from Python.org
No joy yet - this time a completely different error. I just tried r6871
and r6872...
http://pastebin.com/d4d240b36
From: numpy-discussion-boun...@scipy.org
[mailto:numpy-discussion-boun...@scipy.org] On Behalf Of Charles R
Harris
Sent: 16 April 2009 02:28
To:
On Thu, Apr 16, 2009 at 11:03 AM, Fadhley Salim
fadhley.sa...@uk.calyon.com wrote:
No joy yet - this time a completely different error. I just tried r6871 and
r6872...
http://pastebin.com/d4d240b36
do you have g77 included in mingw and on your path?
Josef
I agree with Robert
There should be no reason on earth why you cannot use an Egg to package
Numpy. Setuptools is not fragile, it's very stable but requires a bit of
understanding.
The problem I have is not packaging numpy but compiling it on Windows.
If I could compile the C I'd have built the
On Fri, Apr 17, 2009 at 12:10 AM, josef.p...@gmail.com wrote:
On Thu, Apr 16, 2009 at 11:03 AM, Fadhley Salim
fadhley.sa...@uk.calyon.com wrote:
No joy yet - this time a completely different error. I just tried r6871 and
r6872...
http://pastebin.com/d4d240b36
do you have g77 included in
On Fri, Apr 17, 2009 at 12:12 AM, Fadhley Salim
fadhley.sa...@uk.calyon.com wrote:
I agree with Robert
There should be no reason on earth why you cannot use an Egg to package
Numpy.
Actually, there is, although it is not really an egg deficiency. We
use ATLAS as blas/lapack, and ATLAS
As I said before we have a very big set up. The problem is not the
difficulty in making eggs. Eggs are very easy things to make if you can
make the C++ compile first.
Setuptools is a very good and stable project whose purpose is to help
automate very large deployments of Python dependancies.
Eggs are still beneficial:
* People who do not have access to PyPi can still use it (think banks)
* If you can build eggs easily then so can people like me... it becomes
easy to produce optimized eggs for all our Xeon processors.
* Deploying an egg is nothing more than copying a file to the
We need to be able to run both projects concurrently on the same grid.
Setuptools + eggs allows this to happen. If we used .exe installers then
we could only have one single version of any given dependancy at any
time and so we would not be able to run the two projects in paralell.
I think
On Thu, Apr 16, 2009 at 11:50 PM, Fadhley Salim
fadhley.sa...@uk.calyon.com wrote:
I just installed the latest stable mingw, and made sure that the mingw
bin directory is in my PATH. I used the command you suggested and got
the following output:
http://pastebin.com/m4aea512c
Could you make
if you built numpy from source with a site.cfg file pointing to you atlas
libraries, numpy.dot() will use that library natively. no need to import
_dotblas.
Chris
On Tue, Apr 14, 2009 at 6:16 PM, Mathew Yeates myea...@jpl.nasa.gov wrote:
Hi
The line
from _dotblas import dot . is giving
Hi,
Our nightly build system has been detecting tabs in the recent versions
of numpy. The following files appear to have issues:
Checked out revision 6870.
svn checkout ok
PYTHON FILES INDENTED WITH TABS:
./numpy/numpy/distutils/fcompiler/compaq.py
./numpy/numpy/distutils/command/build_ext.py
Thanks, Chris -- I fixed it in trunk. This is the kind of check we
should be running on the buildbot.
Regards
Stéfan
2009/4/16 Christopher Hanley chan...@stsci.edu:
Hi,
Our nightly build system has been detecting tabs in the recent versions
of numpy. The following files appear to have
Does anyone have a binary installer for numpy 1.3.0 and Python 2.6?
I've been able to install from source and all tests passed, but I prefer
official binaries because I have some confidence that there are no
hidden dependencies (important for distributing self-contained apps).
I tried to build
On Fri, Apr 17, 2009 at 4:55 AM, Russell E. Owen ro...@u.washington.edu wrote:
Does anyone have a binary installer for numpy 1.3.0 and Python 2.6?
I've been able to install from source and all tests passed, but I prefer
official binaries because I have some confidence that there are no
hidden
On Fri, Apr 17, 2009 at 12:25 AM, David Cournapeau courn...@gmail.com wrote:
We need to be able to run both projects concurrently on the same grid.
Setuptools + eggs allows this to happen. If we used .exe installers then
we could only have one single version of any given dependancy at any
time
On Thu, Apr 16, 2009 at 7:43 PM, David Cournapeau courn...@gmail.comwrote:
On Fri, Apr 17, 2009 at 12:25 AM, David Cournapeau courn...@gmail.com
wrote:
We need to be able to run both projects concurrently on the same grid.
Setuptools + eggs allows this to happen. If we used .exe installers
On Thu, Apr 16, 2009 at 4:23 PM, Robert Kern robert.k...@gmail.com wrote:
I think it will try to collect all of the possibilities by looking at
PyPI and the download link, then decide on the best one. But it could
be that if there are files on PyPI, it will only consider those. I
don't know.
On Fri, Apr 17, 2009 at 1:38 PM, Robert Kern robert.k...@gmail.com wrote:
On Thu, Apr 16, 2009 at 23:37, David Cournapeau courn...@gmail.com wrote:
On Thu, Apr 16, 2009 at 4:23 PM, Robert Kern robert.k...@gmail.com wrote:
I think it will try to collect all of the possibilities by looking at
On Thu, Apr 16, 2009 at 23:45, David Cournapeau courn...@gmail.com wrote:
Since I am not an egg user myself, I wonder whether it would be useful
to make numpy zip-safe ? Can it be done without numpy relying on
setuptools, or do we have to use setuptools for pkg_resources (maybe
pkg_resources
33 matches
Mail list logo