nc5ng / nadcon5-ng Goto Github PK

currently we read converage and vector data into python, need to expand scope to include grid files and final transformations

*.pyf files under nc5ng/core where generated but not modified.

There is a need to go through these generated sources to create input/output mappings, array info, etc. for the functions we have imported.

Additionally, docstrings should be created or remapped from the module to the function

Targeting functions and submodules first, experiment with wrapping fortran files in python

F2PY has been chosen over Cython for one primary reason, it more cleanly seperates the fortran code from python code, and makes it easier for us to maintain the python linkage as a seperate external module, as opposed to an integrated one.

For the time being, in order to maintain continuity with upstream there is no real desire to modify the root fortran code (except superficially), as such f2py even with its custom wrapping syntax appears to be preferable.

As part of #14

Release the two tightly couple packages nc5ng.core and nc5ng.nc5data as seperate pypi published packages

Need to fix regression in src/Makefile where the helper binaries (under BinSource) are no longer buind built

Gating #4 , The following core library submodules link other submodules and require resolution

• nc5ng.core.bwplotvc
• nc5ng.core.bwplotcv
• nc5ng.core.coplot
• nc5ng.core.coplotwcv

these all depend on plotcoast.f

Current solution is to pass f2py_options per module in setup.py

Not compile fortran core library but inform user, it is unnecessary dependency for what is a potential future feature

To be safe this requires import check for nc5ng.core but there are no consumers in any project yet

Need to finish NADCON5 Manual section of the documentation and to update the generated docs to include the python packaged

TBD: To Package Python Docs in Doxygen or seperate website (e.g. hosted on readthedocs)

Master Make Fails to Complete, this is due to reintroduced forrtan compiler options and particularly

-O5

Roughly there are two sets of programs that generate the final output and data.

There is some amount of transformation (pruning, mean calculation from the vector data, etc.) that happens in the fortran doer programs (e.g. makeplotfile01 makeplotfiles02 myrms)

The other part of transformation is handled by Generic Mapping Tools using batch files generated by the fortran doer programs, the GMT tools act on the processed data and produced additional datasets

This issue focuses on taking over the transformations run by the fortran programs and not the generated batch file.

Components

List

Todo

Wrap the NADCON5 data files, input, and output files in a python inspection and plotting layer.

Tentative Branch : develop/python-nc5data-wrappers

Our makefile structure is very autoconf like, with most variables being definable. need to take the leap to autoconf to take care of certain configuration stuff for us