Package Details: python-netcdf4 1.2.7-1

Git Clone URL: https://aur.archlinux.org/python-netcdf4.git (read-only)
Package Base: python-netcdf4
Description: A python3 library for reading, manpiulating and creating netcdf files
Upstream URL: https://unidata.github.io/netcdf4-python/
Licenses: MIT
Submitter: None
Maintainer: Foucault
Last Packager: Foucault
Votes: 9
Popularity: 0.044252
First Submitted: 2010-02-10 12:12
Last Updated: 2017-01-08 22:50

Dependencies (8)

Required by (3)

Sources (1)

Latest Comments

Foucault commented on 2017-01-08 22:50

Version 1.2.7 proper is out, so everything should work properly with python 3.6

Foucault commented on 2017-01-06 20:27

There is no easy way to extract the fix for the tests as the patch does not apply cleanly to v1.2.6rel. What I did is fetch the latest commit from the master branch into v1.2.6. This allows the package to build but it is essentially a "prerelease" for 1.2.7. Hopefully the official 1.2.7 will appear within the next days anyway so we can have a proper package again. Please do comment if you have any runtime issues.

richli commented on 2017-01-06 19:19

Note that the check() function in the PKGBUILD needs to be updated for Python 3.6. Further, there's a test case that fails. A fix has been merged, but upstream has not yet released a new version.

https://github.com/Unidata/netcdf4-python/issues/626

Foucault commented on 2016-12-25 16:11

You are right, dateutils has been dropped indeed. Curl is required for OPeNDAP [0] support but should be optional dependency instead of a hard one. I will see to moving the documentation to a separate package within the next days and drop the python2 optdep.

[0]: https://www.opendap.org/

giniu commented on 2016-12-25 14:58

Hello, I plan to move python-netcdf4 to community as a dependency of python-xarray (which is optional/recommended dependency of python-pandas).

I have a few questions/suggestions:

- is deteutil still a dependency? Looks like it was dropped around 1.1.6.
- is curl needed? it is a dependency of netcdf, it is transitive dependency so already satisfied. Same with python.
- what is libssh2 and openssl and zlib needed for? I cannot find any mentions in documentation.
- might be good to move docs and examples to separate split package - it is not good to pull in all python2 dependencies for docs and examples, especially that python 3 is default for ages now.

Polly commented on 2016-11-13 14:23

The build directory has 26M as expected. The nc files are created in /tmp. Do the tests dump temporary files there?

Foucault commented on 2016-11-13 11:56

This is very strange. In my case build directory after build is roughly 25 MB. In any case you might want to skip the tests to get it build for the time being by deleting the check() function in the PKGBUILD.

Polly commented on 2016-11-13 11:42

Something is going wrong with the tests. It creates gigaytes of temporary nc files and fills my drive until it crashes.
That's the setup:
curl 7.51.0-1
netcdf 4.4.1-2
hdf5 1.10.0_patch1-1
python 3.5.2-1
cython 0.25.1-1

Anyone else experiencing the problem?

richli commented on 2015-10-01 20:33

Note that only does the package need rebuilding for Python 3.5, but the check() function in the PKGBUILD needs to be updated to work successfully (s/-3.4/-3.5/).

richli commented on 2015-03-26 17:02

FYI: I had to rebuild this package after the recent update to Python 3.4.3-2. Same issue as here: [1]

[1] https://bugs.archlinux.org/task/44350

Foucault commented on 2015-03-18 22:57

The examples of v1.1.7 include a set of binary blobs that amount to more than 100 MB, so I disabled them from the package. Uncomment the relevant line if you want them.

Foucault commented on 2015-03-09 20:27

Thanks for the input! I updated the PKGBUILD.

richli commented on 2015-03-09 20:03

@Foucault: I adopted the python2-netcdf4 package [1] and based it off your PKGBUILD. I figured out a few improvements you may like for your version:

(1) You can avoid the second (redundant) build step in the package() function by passing "--skip-build" to setup.py

(2) Also in the call to setup.py in the package() function, you can't pass "--optimize" like you can with most Python packages. However "--optimize 2" works.

(3) You can run the test suite by setting the PYTHONPATH environment variable. You can look at my check() function to see how I did it.

Hope that helps! Thanks for keeping this package up to date.

[1] https://aur.archlinux.org/packages/python2-netcdf4/

richli commented on 2015-03-02 18:46

According to the changelog [1], it seems that python-dateutil [2] should be added to the depends array. I haven't tried running it without python-dateutil installed, but I imagine it would throw an error.

Thanks!

[1] https://github.com/Unidata/netcdf4-python/blob/v1.1.5rel/Changelog
[2] https://www.archlinux.org/packages/community/any/python-dateutil/

Foucault commented on 2014-09-08 19:38

Thanks! Updated.

richli commented on 2014-09-08 19:28

The project has moved to Github. Here is the new upstream URL:

https://github.com/Unidata/netcdf4-python

richli commented on 2014-03-05 05:12

Hey, not a big deal, but is there a reason to put the explicit version dependency for hdf5?

The "hdf5-cpp-fortran" package (which I use) conflicts with the "hdf5" package. It also is set so it provides hdf5. So any package (such as this one) that depends on hdf5 can work with either the "hdf5" or "hdf5-cpp-fortran" packages.

However, having the explicit version dependency here breaks this so that instead of being satisfied with "hdf5-cpp-fortran", it requires the "hdf5" package. If I remove the version dep, then it's fine with accepting "hdf5-cpp-fortran" again.

Anonymous comment on 2012-02-09 02:04

rebuild against latest netcdf

big_gie commented on 2012-01-21 20:23

0.9.9 is available.

Anonymous comment on 2011-04-06 23:53

This package is now the python3 version. The python2 version is now called (strangely) python2-netcdf4