Future Work

This section describes the features and fixes I plan to work on in this package. The most urgent items are listed closer to the begining of the lists.

• Automate the installation using Python's distutils utilities.

• Describe a way of using job control (either via the operating system or IPython's jobctrl module) to do a quick-and-dirty parallelization of multiple Qtcm instance run sessions. Or use some sort of threading to fire up two simulataneously running models. Check that the simultaneously running models have different memory space.

• Add capability for create_benchmark.py to overwrite existing benchmark files.

• Make compiled_form set to 'parts' as the default instantiation. Change documentation accordingly.

• Currently, the Qtcm plotm method works only on 3-D output (time, latitude, longitude). Some of the fields in the netCDF output files are 2-D. Add the capability to plot_netcdf_output in the plot submodule to handle 2-D fields.

• Add documentation about removing temporary files. Add documentation in Section 4.2 of details of what occurs during instantiation of a Qtcm instance.

• Add the units and long names for all field variables in the defaults module.

• Create a keyword to automatically change precipitation and evaporation units to mm/day (or similar).

• Add ability to calculate and plot fields at different pressure levels. Create another module like defaults that specifies the vertical fields and gives the equation to use to calculate those fields; call the module ``derivfields'' or something similar.

• Throughout the qtcm package I use the condition N.rank(arg)=0 to test whether arg is a scalar. This works fine for numpy objects, but it does not work properly for Numeric and numarray arrays. In those array packages, rank('abc') returns the value 1. This is not a problem, as long as everyone has numpy, but in order to make the package interoperable, I need to find a better way of testing for scalars. The definitions of isscalar need to be changed in num_settings.

• num_settings needs to be changed to truly enable me to test whether qtcm works for numarray and Numeric arrays. The tests do not do this right now, because num_settings defaults to numpy, if it exists.

• Create makefiles for other platforms.

• A few fields (e.g., u1) have data for extra latitude bands, due to the use of ``ghost latitudes'' as part of the implementation of the numerics. Details are found in the QTCM1 manual [4].

  Though adjusting to this idiosyncracy is not that difficult, in the future I hope to implement a method of handing fields with ghost latitudes so that they have the same dimensions as the other gridded output variables. In order to do this, I plan to write a Python method to read the Fortran generated binary restart file.

• Change the set_qtcm_item method so that it can automatically accomodate setting Fortran real variables if integer values are input.

• Currently, the get_item_qtcm and set_item_qtcm methods will not work on integer and character arrays, only scalars and real arrays. Add that missing functionality to those methods.

• Currently, the make_snapshot method duplicates the functionality of the pure-Fortran QTCM1 restart file mechanism. However, the restart file mechanism itself does not do a true restart. A continuous run does not provide the same results as two runs over the same period, joined by the restart file.

  To see whether saving more variables would do the trick, I altered make_snapshot to store all Python level variables (i.e., self._qtcm_fields_ids). However, the restart failing described above still continued. In the future, I hope to figure out exactly how many variables are needed in order to make the restart feature do a true restart.

• Add a test of using the mrestart=1 restart option. Does the qtcm.restart file need to be in the current working directory or another?

• Add a test in the unit test scripts to confirm that the init_with_instance_state attribute setting only has an effect if compiled_form='parts'.

• Document tmppreview keyword in plot.plot_ncdf_output.

• Confirm and document that for netCDF output, time is model time since dd-mm-yyyy.

• Add to the plotm method the ability to plot as text onto the figure the runname string and the calling line for the plotm method.

• Couple with the CliMT climate modeling toolkit.

• Enable Python to set arr1name, etc., which are string variables at the Python level. I haven't really thought through how arr1 variables work with the Python Qtcm instance.

• Possible: In the Qtcm method __setattr__, add a test to raise an exception if the instance tries to set viscU, viscT, or viscQ as attributes. Also create a method isotropic_visc that will set all viscosity parameters non-dependent on direction. See Section 6.4.1 for details.

• Go through the manual and create HTML-only versions of tables that have table numbers (use a similar construct as in figure environments).

• Go through documentation to check that output variable names are capitalized consistently.

• Create way to redirect stdout.

• Create a step method to run an arbitrary number of timesteps at the atmosphere level.