adrn / gala Goto Github PK

All subpackages need much better documentation!

Using the axes= kwarg, not plotting on the specified axes.

I need to add Returns blocks to all docstrings.

Right now, all potentials are subclasses of Cartesian potentials so that, for example, even in spherical potentials, the user has to specify a full 3D position for each. Is there a better way to handle this so that a spherical potential can be used as 1D? Or, should the classes support Astropy coordinate representations so the class evaluations can accept, e.g., spherical and cartesian?

And should return times

Should return an array of energies.

Sum the potential values...

For example, inference and parts of observation should not be in here.

In SCFReader

Integrators should accept a memory-mapped array to dump coordinates to. This is helpful for integrating many particles, or long integration windows so that memory usage doesn't explode.

This would add support for, e.g., misaligned disks, point masses at positions other than the origin, rotated or rotating halos.

Passing in an ICRS vs. Galactic coordinate leads to output velocities that are transposed from one another?

Right now they just return the best fitting parameters.

Need a way to get back the components of a CCompositePotential. Also, value and gradient work differently in the Cython class.

Would be very useful to use adaptive integrators, and still be able to FFT within NAFF.

Maybe a bug in Scipy's ode?

Or, leave in package with sphinx warnings that it is experimental and API will change

E.g., pericenter, apocenter, eccentricity,

orbit.pericenter() to get mean
orbit.pericenter(np.min) to get minimum pericenter
etc.

e.g.,

ix = np.array([0,1,2,3,7])
w[ix]

Using 'triangle=True' kwarg to .plot() seems to fail to plot first panel

• Update core dynamics routines to use orbits
• Update angle-action / NAFF routines to use orbits
• Update integrate subpackage to return Orbit instances, accept initial conditions
• Update io code to read in files as Orbit instances
• Update potentials to work with orbits

Write tests for everything, and use coverage.py to check test coverage.

If a potential is a CPotential subclass, pass the c_instance and integration arguments in to the C-implementation of Leapfrog.

e.g.,

ERROR: TypeError: F() argument after * must be a sequence, not float [streamteam.integrate.dopri853]

Should probably support Python 3 and Python 2.6

Should internally promote to a (3,n) array

Right now, some docstrings aren't using proper Sphinx references.

Need to add support to the SCF class to read the other simulation output files. Currently only supports SCFPAR (to get units) and SNAP files.

The issue is not about docstrings -- class documentation, e.g., the built-in potentials -- are not showing members inherited from superclasses. It looks like this is because of how automodapi works. So, I either need to not use that and use

:members:
:inherited-members:

(see here )

See TODO here http://gary.readthedocs.org/en/latest/dynamics/index.html

Also add link to tact

e.g., if

p = OphiuchusPotential()
p['derp'] = ...

Then pickle and load will fail. For subclasses, this shouldn't allow adding a component that isn't a registered name.

Some stuff in the coordinates and units subpackages might be merged into the astropy core eventually.

Do the integral over τ from specified density -- only supports cartesian for now...

Right now, any plots made are saved to tmpdir. This should be the fallback behavior -- really the plot should be saved somewhere the user can look and organized in a sensible way.