Notes from Claw Dev March 2015

See https://github.com/clawpack/clawpack/wiki/Claw-Dev-Hackathon,-March-2015

Visualization

Donna's google earth tools:

• https://github.com/donnaaboise/visclaw/tree/googleearth
• plotfigure.kml_xlimits etc. specified in setplot.py
• Why is this attribute of plotfigure rather than plotaxes? Can only have one axes for a kml figure.
• Would be nice to have transparent where wave amplitude below some tolerance.
• Progress made by Donna Calhoun and Tommaso Buvoli, see https://github.com/clawpack/visclaw/pull/150

Griddle

• Data may be in memory rather than reading from disk.
• Create figure and keep handle, not easily available in visclaw.
• Uses dictionary with attributes data, field, etc. rather than defining new custom class.
• Less hierarchical -- each plotitem tells what axis it's on.
• Gallery generation using https://github.com/saimn/sigal -- possible replacement for plotpages.py with make plots?
• Order in which things get plotted: finer patches get plotted on top of coarser, can specify custom order items.

Proposal: Make clawpack/griddle a separate repository, which eventually would include new I/O routines and other stuff that's not Clawpack-specific.

Pandas:

• New data analysis package Used extensively, especially in finance.
• dataframe: each entry could be results from one run, with metadata about parameters used to produce them.
• time series: data with time stamps or other sequential. Might be useful for gauges. Built in statistical tools.
• Would be nice to make notebook with various examples.

Plotting slices of 3d results

• Progress made by Donsub Rim and Chris Vogl.

Parameter sweeps:

• Kyle has some nice tools at https://github.com/mandli/tohoku-pc

Installation

• Install all Python code to site-packages

  • Need to separate out install for users vs. developers where users do not modify source and developers do.
  • Decided for user installs we will install ALL python source to site-packages (including the Fortran).
  • Need to clean up setup.py and write better documentation for installing in user and developer modes

• Want to have clawpack/clawpack updated automatically when a PR is issued

• clawpack/clawpack#44 assigned to Kyle

• Move all Python code so it gets installed in site-packages, so never set PYTHONPATH.

• Refer users to code online if they want to look at it.

• Improve instructions for installation.

• Do not move geoclaw/python stuff into pyclaw because separate repos much better for development.

• Create mechanism so that any merge into master in any repo creates new PR on clawpack/clawpack #58

Run-time, compile-time options

• Run-Time and Compile-Time Option Handling for Fortran Code
  • Agreed to start to use pre-processor flags for use of packages such as NetCDF while trying to retain run-time options where at all possible
• Use ifdef instead of different versions of code with / without NetCDF, for example.
• Storm surge -- some places where only a small amount of new code needed in some routines.
• Need documentation.
• See https://github.com/clawpack/clawutil/pull/84

Riemann solvers

• Pointwise a good idea. Do not see a need for passing in row above/below to normal solvers. @brisadavis is starting to work on this.
• geometry popular but not backward compatible --- defer for now, along with other CEPs.
• See https://github.com/clawpack/riemann/pull/101 for some new pointwise solvers.
• New repository to collect notebooks for a planned book on Riemann solvers: https://github.com/clawpack/riemann_book

Third order methods

• Christiane Helzel's recent work implementing 3rd order L-W for systems.
  • Uses f-wave formulation since involves (f'(q)f(q))_x, etc.
  • Need to distinguish between cell averages and point-wise values. In 1d, simple formula to compute ptwise from averages. In multi-dim case, need to do multi-dim reconstruction. This should be done outside or Riemann solver, e.g. in flux1.
  • Flux limiter currently used - is there a good way to limit Z-waves instead?
  • Works well for scalar problems. Some improvement for systems, but not fully 3rd order.
• Scott Moe's work with David Seal, James Rossmanith:
• Evolve more higher order terms explicitly as in DG.
• Flux function is augmented with additional terms for high-order derivatives before applying DG.

Time stepping

• How to use SharpClaw as the "right-hand-side" within your own time-stepping code? There is an example here: https://github.com/Parallel-in-Time/pySDC/tree/master/examples/sharpclaw

GeoClaw

• Better support for other topo types: see https://github.com/clawpack/geoclaw/pull/137

Misc.

• Fixed recursive Makefile calls so e.g. make .plots -f <makefilename> works properly: https://github.com/clawpack/clawutil/pull/83