Python read_geqdsk function interpolating incorrectly

[shaunhaskey]

The python read_geqdsk function seems to be interpolating incorrectly (probably since the change to using RegularGridInterpolator:

import fidasim
import matplotlib.pyplot as plt
import numpy as np
geqdsk_fname=f"/home/shaskey/code/FSIM_NN/sim_files/g164988.03525"
rmin, rmax= 100., 250.
zmin, zmax = -150., 150.
nr = 500
nz = 300
grid = fidasim.utils.rz_grid(rmin, rmax, nr, zmin, zmax, nz)
equil, rhop, btipsign = fidasim.utils.read_geqdsk(geqdsk_fname, grid, poloidal=True)
fig, ax = plt.subplots()
im = ax.pcolormesh(grid['r2d']/100, grid['z2d']/100, rhop)
im.set_clim([0, 1.5])
ax.contour(grid['r2d']/100, grid['z2d']/100, rhop, 8, colors='k')
ax.set_xlabel('R [m]')
ax.set_ylabel('Z [m]')
fig.canvas.draw();fig.show()

Produces an image that doesn't look correct:

I think this is the source of the problem, due to a recent switch from interp2d to RegularGridInterpolator:
https://github.com/D3DEnergetic/FIDASIM/blame/d727862a6b897ecca70610a4964583079968f812/lib/python/fidasim/utils.py#L934

psirz_itp = RectBivariateSpline(r, z, g["psirz"])

RectBivariateSpline expects the third argument to be x.size, y.size so [R_index, Z_index], so a .T is missing for g["psirz"]

This makes sense according to this where we expect to need to include a .T
https://scipy.github.io/devdocs/tutorial/interpolate/interp_transition_guide.html

Screenshot:

[shaunhaskey]

@prechelg @hayashiw
PR for fix is here: #332

I also have some changes to speed this function up.

These types of single calls to the interpolation function are very inefficient:

rhogrid = np.array([psirz_itp(rr,zz) for (rr,zz) in zip(r_pts,z_pts)]).reshape(dims)

I'll push those changes after this fix is accepted/rejected.

[JoseRuedaRueda]

I do agree, this went under my radar, but yes, all densities and other equilibrium inputs from my runs are affected by this :(