"""
et = dset[:, 7:13]
ep = dset[:, 13:19]
# check that et is traceless
et_tr = et[:, 0] + et[:, 1] + et[:, 2]
et_tr_max = np.max(np.abs(et_tr))
if et_tr_max > 1E-4:
print ii
print et_tr_max
# calculate the norm of et
et_norm = np.sqrt(et[:, 0]**2 + et[:, 1]**2 + et[:, 2]**2 + et[:, 3]**2 +
et[:, 4]**2 + et[:, 5]**2)
# check that this curve is represented well with legendre polynomial
ep11 = ep[:, 0]
nodes, weights, rootsamp = lif.get_nodes(et_norm, ep11, a, b, N)
eindx = np.int16(np.round((180. / (3. * np.pi)) * euler[ii, :]))
ep_set[eindx[0], eindx[1], eindx[2], :] = rootsamp
f_mwp.close()
f_nhp.close()
# numbers of nodes
leg_set = np.array([4, 5, 6, 7, 8, 9, 10, 15, 20])
leg_error = np.zeros([leg_set.size, 9])
c = 0
for N in leg_set:
mean_err = np.zeros(N_samp * n_th)
max_err = np.zeros(N_samp * n_th)
for ii in xrange(N_samp * n_th):
xvar = alldata[ii, 4:45]
yvar = alldata[ii, 45:86]
xnode, ynode, weights = leg.get_nodes(etvec, yvar, a, b, N)
coeff_set = leg.get_coeff(xnode, ynode, weights, N)
ytest = leg.get_interp(xvar, coeff_set, a, b)
# calculate error in this approach based on sampled values
error = 100 * np.abs((yvar - ytest) / xvar)
# if N == 6 and ii > .3*N_samp*n_th:
# print np.mean(error)
# plt.figure(num=1, figsize=[10, 6])
# plt.plot(xvar, yvar, 'bx')
# plt.plot(etvec, ytest, 'r')
# plt.show()
