Prepare PyCAMA release 0.10.0
- Update configuration
- Test cases
#7 Updated by Maarten Sneep about 1 year ago
- Status changed from In Progress to Feedback
- Assignee changed from Maarten Sneep to PDGS Operations
- % Done changed from 0 to 100
PyCAMA 0.10.0 is ready, and has been placed on the DMSF of this project. The user manual has been updated to reflect the current status.
- Distribution package: https://dev.knmi.nl/dmsf/files/11031/view or alternatively:
hg clone https://dev.knmi.nl/hg/pycama.
- User manual: https://dev.knmi.nl/dmsf/files/1314/view
- Configuration file: https://dev.knmi.nl/dmsf/files/3381/view
The last two items are included in the distribution package.
Test data is not included in the package because of volume concerns (about 110 GB in total). When testing on the sandbox for CentOS 7.7, the test case failed, but the reason is unclear. The comparison tool reports differences, and while some insignificant rounding differences can be observed, other reported differences are not understood - I do not see how an empty string doesn't compare to an empty string. I ran the same test on my laptop, and did not find any differences. We're likely dealing with some version difference of the python packages, but I do not think it is worth to get the test to reproduce, the code itself works fine. Using Python version 3.7.6 (final), numpy 1.18.1, netCDF4 18.104.22.168 (with netCDF4 library 4.6.2), and h5py 2.9.0 (with hdf5 library 1.10.4) will likely resolve the differences.
All 22 tests (offline and NRTI across all products) have been executed at KNMI, and test data is available on request. Before uploading I need to know how we want to continue. A smaller package for a single test case is available as well.
#10 Updated by Maarten Sneep about 1 year ago
As an extra test, I transferred the file generated in the sandbox to KNMI, and generated the report locally. This is working as expected. I do not see an issue here. Of course, if updates to the python modules are required for other processors then this issue may disappear as well.
Locally tested with python version 3.7.7 (final), numpy 1.18.1, netCDF4 1.5.3 (using netCDF4 library version 4.7.3), h5py 2.10.0 (with hdf-5 library version 1.10.4).