Opening GDAS BUFR files with the python API

[annavaughan]

Is it possible to open GDAS files using the python API?

We are trying to open GDAS AMSU-A files (e.g https://data.rda.ucar.edu/ds735.0/1bamua/2005/1bamua.20050109.tar.gz) and extract data

We have tried:
git clone https://github.com/EXWEXs/py-ncepbufr.git
cd py-ncepbufr
python setup.py build
python setup.py install
And then running (from the notebook found at https://github.com/NOAA-EMC/NCEPLIBS-bufr/blob/develop/python/test/Python_tutorial_bufr.ipynb)
import ncepbufr
bufr = ncepbufr.open("gdas.1bamua.t12z.20210107.bufr")
bufr.advance()
bufr.load_subset()
Which fails on bufr.load_subset() with exit code -1. It looks like py-ncepbufr is unmaintained, is there another way to do this within NCEPLIBS-bufr? Thanks!

[rmclaren]

The python wrapper you are using is now included in nceplib-bufr itself, you should probably use that version as the code you are using is very old. If you are feeling very patient you could try this one: https://github.com/NOAA-EMC/bufr-query

[annavaughan]

Thank you very much for the quick response Ron!

I've tried both nceplib-bufr and the new bufr-query repo.

For nceplib-bufr I tried downloading the release from https://github.com/NOAA-EMC/NCEPLIBS-bufr/releases then
mkdir build && cd build
cmake -DCMAKE_INSTALL_PREFIX=path1 -DENABLE_PYTHON=ON
make -j4
ctest
make install
Which completes, however when I try import ncepbufr in a python session I still get ModuleNotFoundError. Is there any documentation available for how to add this to the python path?

For bufr-query is there any documentation available for installation?

[stratisMarkou]

Hi @rmclaren, and thanks for the quick reply! Also copying in @jbathegit who discussed a similar issue (#530) and may know more about this. Just to add to @annavaughan's issue: it seems that (as discussed in issue #530, see this comment and following comments) when running ctest following the installation some of the tests fail with ModuleNotFoundError. Following the suggestions by @jbathegit in that issue, I've tried cloning the develop branch and following the installation instructions but am still having a few issues. Here are the steps I'm using and the issues there are as follows. First, clone the develop branch:

git clone -b develop git@github.com:NOAA-EMC/NCEPLIBS-bufr.git
cd NCEPLIBS-bufr

With my conda virtual environment enabled, I then run

cmake -DCMAKE_INSTALL_PREFIX=~/ -DENABLE_PYTHON=ON

At this point, one of the messages from cmake complains that it Could not download bufr test files, not building tests:

CMake Warning:
  No source or binary directory provided.  Both will be assumed to be the
  same as the current working directory, but note that this warning will
  become a fatal error in future CMake releases.


-- The C compiler identification is GNU 4.8.5
-- The Fortran compiler identification is GNU 4.8.5
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc - works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working Fortran compiler: /usr/bin/gfortran
-- Check for working Fortran compiler: /usr/bin/gfortran - works
-- Detecting Fortran compiler ABI info
-- Detecting Fortran compiler ABI info - done
-- Checking whether /usr/bin/gfortran supports Fortran 90
-- Checking whether /usr/bin/gfortran supports Fortran 90 - yes
-- Finding test data files in directory ..
-- Setting build type to 'Release' as none was specified.
-- Found Python3: /home/em626/miniconda3/envs/npw/bin/python3.8 (found version "3.8.18") found components: Interpreter 
-- Check if the system is big endian
-- Searching 16 bit integer
-- Looking for sys/types.h
-- Looking for sys/types.h - found
-- Looking for stdint.h
-- Looking for stdint.h - found
-- Looking for stddef.h
-- Looking for stddef.h - found
-- Check size of unsigned short
-- Check size of unsigned short - done
-- Searching 16 bit integer - Using unsigned short
-- Check if the system is big endian - little endian
-- Downloading bufr test files...
-- Could not download bufr test files, not building tests
-- Configuring done
-- Generating done
-- Build files have been written to: /home/em626/rds/hpc-work/NCEPLIBS-bufr

In my local machine, the download works fine, and I can see the files under test/testfiles, but in our cluster (where I'm running the above commands) the download seems to fail without giving an explanation in the error. In any case, running

make -j4
make install

seems to work fine except make gives the warning WARN: Could not locate executable armflang, which I'm guessing might not be important. Lastly, I download and unzip a BUFR file from GDAS data:

wget https://data.rda.ucar.edu/ds735.0/1bamua/2005/1bamua.20050109.tar.gz
tar -xvf 1bamua.20050109.tar.gz

And then try to load it in python:

import ncepbufr  # import works fine
bufr = ncepbufr.open("20050109.1bamua/gdas.1bamua.t00z.20050109.bufr")  # also works fine
bufr.advance()  # returns 0
bufr.load_subset()  # returns -1

So load_subset does not seem to work as expected. I've also tried this on one of the test bufr files (which I donwloaded on my local machine and copied over) but that does not work either. @rmclaren @jbathegit, do you have any thoughts on what might be going wrong? We'd really appreciate your help!

[rmclaren]

@annavaughan The python path should be extended with the the site-packages directory that appears in the NCEPLIB-bufr install directory. So for example /lib/python3.12/site-packages (path will vary depending on the version of python you are using). You can either extend it on the command line (export PYTHONPATH = <your path>:$PYTHONPATH) or you can extend the sys path in your python script (import sys; sys.path.insert(0, <your path>)).

The bufr-query thing has a lot of dependencies which might be hard for you to build (JCSDA oops, ioda, etc...) which I'm working to remove (uuuugggg)... So might want to hold off for now.

[rmclaren]

@stratisMarkou I'm going to let the NCEPLIB-bufr maintainers answer your question.

[jbathegit]

Sorry, but for my part I'm afraid I don't have much to contribute. The python binding and associated modules is the part of the library I'm least knowledgeable about personally (I'm more of a Fortran and C person myself w.r.t. code development, in addition to being the overall library manager :-) But what I can do is bring a few other folks into this discussion (@AlexanderRichert-NOAA, @jswhit, @climbfuji, @edwardhartnett, @aerorahul) who might be able to help.

That said, please note that #530 was specifically related to the new Intel oneAPI compilers, and Alex in particular has been working further on that in #538, but it's still not resolved. Otherwise, the only way I can see that either of those might be related to this issue is in the aspect of being able to download the test tarfile, which was really just a side part of the discussion in #530. So if you're still having that problem(?), could you please specify which particular release version of the code you were trying to download from https://github.com/NOAA-EMC/NCEPLIBS-bufr/releases? If it was 11.5.0 or older, then that might well explain that issue, and I would respectfully suggest you to try downloading a newer release as noted in #530.

[jbathegit]

No further responses received, so closing this issue.

[vimal295sen]

In HPC You can use following procedure -:

ncepbufr installation -:
module purge
source install_Application/miniconda3/etc/profile.d/conda.sh
conda activate obsmon
python -m site
export PYTHONPATH=/home/vimal/install_Application/miniconda3/envs/myenv/lib/python3.9/site-packages:$PYTHONPATH
git clone https://github.com/EXWEXs/py-ncepbufr.git
cd py-ncepbufr/
python setup.py build
conda install numpy
python setup.py build
python setup.py install