Bug#389118: [Deb-scipy-devel] Bug#389118: python-numpy: will not import unless python-numpy-ext is installed
Robert Kern writes: > Thank you. As an upstream numpy developer, I can verify that the contents of > python-numpy-ext cannot be separated from python-numpy. I assume the separation of the base package and the extension packages comes from the Numeric and numarray packages in debian. The extension did make sense to avoid the installation of the blas and lapack libraries; basic functionality can be used without having these libraries be installed. In which ways did numpy change, so that this is not possible anymore? Matthias -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#389118: [Deb-scipy-devel] Bug#389118: python-numpy: will not import unless python-numpy-ext is installed
Matthias Klose wrote: > Robert Kern writes: >> Thank you. As an upstream numpy developer, I can verify that the contents of >> python-numpy-ext cannot be separated from python-numpy. > > I assume the separation of the base package and the extension packages > comes from the Numeric and numarray packages in debian. The extension > did make sense to avoid the installation of the blas and lapack > libraries; basic functionality can be used without having these > libraries be installed. In which ways did numpy change, so that this > is not possible anymore? numpy is a unitary package, now. You can't remove subpackages and still have a functional numpy. Look at the __init__.py: import linalg import fft import random The core has the same optional dependency on BLAS as the linalg subpackage for the numpy.dot() function, and the Debian package python-numpy already has the dependency on LAPACK and BLAS Debian packages, so folding python-numpy-ext into it will not add any more dependencies than already exist. Also, for f2py to be functional, the two files in python-numpy-dev also need to be present. It would save everyone some grief if those files were in python-numpy itself. At the very least, python-numpy-dev should be in the Recommends of python-numpy, and the errors one gets when trying to use f2py without it should be documented in python-numpy's README.Debian such that users know what is going on. -- Robert Kern "I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth." -- Umberto Eco -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#389118: [Deb-scipy-devel] Bug#389118: python-numpy: will not import unless python-numpy-ext is installed
Robert Kern writes: > Matthias Klose wrote: > > Robert Kern writes: > >> Thank you. As an upstream numpy developer, I can verify that the contents > >> of > >> python-numpy-ext cannot be separated from python-numpy. > > > > I assume the separation of the base package and the extension packages > > comes from the Numeric and numarray packages in debian. The extension > > did make sense to avoid the installation of the blas and lapack > > libraries; basic functionality can be used without having these > > libraries be installed. In which ways did numpy change, so that this > > is not possible anymore? > > numpy is a unitary package, now. You can't remove subpackages and still have a > functional numpy. Look at the __init__.py: > > import linalg > import fft > import random > > The core has the same optional dependency on BLAS as the linalg subpackage for > the numpy.dot() function, and the Debian package python-numpy already has the > dependency on LAPACK and BLAS Debian packages, so folding python-numpy-ext > into > it will not add any more dependencies than already exist. ok, do the Debian maintainers agree, if I prepare such an upload? > Also, for f2py to be functional, the two files in python-numpy-dev also need > to > be present. It would save everyone some grief if those files were in > python-numpy itself. At the very least, python-numpy-dev should be in the > Recommends of python-numpy, and the errors one gets when trying to use f2py > without it should be documented in python-numpy's README.Debian such that > users > know what is going on. Or put f2py into the -dev package? Debian has the distinction between runtime and development packages; to make the package work out of the box, you would have to depend on compilers as well, which is not desired when you install the package just as a dependency of another package. Maybe another option could be a renaming of python-numpy to python-numpy-runtime and python-numpy-dev to python-numpy, so that you get what upstream wants you to get. Matthias -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#389118: [Deb-scipy-devel] Bug#389118: python-numpy: will not import unless python-numpy-ext is installed
Matthias Klose wrote: > Robert Kern writes: >> Also, for f2py to be functional, the two files in python-numpy-dev also need >> to >> be present. It would save everyone some grief if those files were in >> python-numpy itself. At the very least, python-numpy-dev should be in the >> Recommends of python-numpy, and the errors one gets when trying to use f2py >> without it should be documented in python-numpy's README.Debian such that >> users >> know what is going on. > > Or put f2py into the -dev package? Debian has the distinction between > runtime and development packages; to make the package work out of the > box, you would have to depend on compilers as well, which is not > desired when you install the package just as a dependency of another > package. Maybe another option could be a renaming of python-numpy to > python-numpy-runtime and python-numpy-dev to python-numpy, so that you > get what upstream wants you to get. If you want to move the f2py script (though not the numpy.f2py subpackage) to another Debian package, I have no objection. However, the fortranobject.[ch] files are package data, and IMO should stay with the numpy.f2py subpackage; I don't see how anyone benefits by removing them. -- Robert Kern "I have come to believe that the whole world is an enigma, a harmless enigma that is made terrible by our own mad attempt to interpret it as though it had an underlying truth." -- Umberto Eco -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]