# coding: utf-8

# This demonstrates the use of the humlicek, weideman, and sdv modules and how to generate the contour plots.
# Ideally you should have started the IPython interpreter with the --pylab option
# to automatically pre-load matplotlib and numpy for interactive use:
# 
# ipython --pylab 
# ipython qtconsole --pylab &
# ipython notebook --pylab &

from matplotlib import ticker   # possibly its already imported automtically

from ir import *
from humlicek import *
from weideman import *
from sdv import *       # this also imports the wofz reference from scipy!

# define the standard Voigt variables on a linear and logarithmic grid
x=np.linspace(0.,25.,251)
y=np.logspace(-6,2,81)
# kind of 'matrices' required for the contour plots (note Python is case sensitive!)
X, Y = np.meshgrid(x, y)
Z = X +1j*Y

# lets start with the Voigt function, just compare the Humlicek and wofz reference for y=1
wRef  = wofz(x+1j).real
wTest = humlicek(x+1j).real
plot (x, wTest, 'r', x, wRef, 'b--')
# possibly its better to close the figure before looking at the differences
semilogy(x,abs(wTest-wRef)/wRef)

# the same, but for all y
wRef = wofz(Z).real
wTest= humlicek(Z).real
contourf(X,log10(Y), abs(wTest-wRef)/wRef)  # first try!

# Use the levels as in the paper and change to the red-blue color map
contourLevels = logspace(-12,4,17)
set_cmap('RdBu_r')  # if you have already closed the figure this will give a RuntimeError, ignore it

# now try a Humlicek-Weideman combination
wTest = hum1wei32a(Z).real
contourf(X,log10(Y), abs(wTest-wRef)/wRef, contourLevels, locator=ticker.LogLocator())
colorbar()

# and now the speed-dependent Voigt
sdvRef  = sdv_axy_naive(X,Y,10.,wofz)
sdvTest = sdv_axy_naive(X,Y,10.,hum2wei32a)
contourf(X,log10(Y), abs(sdvTest-sdvRef)/sdvRef, contourLevels, locator=ticker.LogLocator()); colorbar()