import mca
import pandas as pd
import numpy as np

np.set_printoptions(formatter={'float': '{: 0.4f}'.format})
pd.set_option('display.precision', 5)
pd.set_option('display.max_columns', 25)

data = pd.read_table('../data/burgundies.csv',
                     sep=',', skiprows=1, index_col=0, header=0)
X = data.drop('oak_type', axis=1)
j_sup = data.oak_type
i_sup = np.array([0, 1, 0, 1, 0, .5, .5, 1, 0, 1, 0, 0, 1, 0, .5, .5, 1, 0, .5, .5, 0, 1])
ncols = 10

X.shape, j_sup.shape, i_sup.shape

src_index = (['Expert 1'] * 7 + ['Expert 2'] * 9 + ['Expert 3'] * 6)
var_index = (['fruity'] * 2 + ['woody'] * 3 + ['coffee'] * 2 + ['fruity'] * 2
             + ['roasted'] * 2 + ['vanillin'] * 3 + ['woody'] * 2 + ['fruity'] * 2
             + ['butter'] * 2 + ['woody'] * 2)
yn = ['y','n']; rg = ['1', '2', '3']; val_index = yn + rg + yn*3 + rg + yn*4
col_index = pd.MultiIndex.from_arrays([src_index, var_index, val_index], 
                                      names=['source', 'variable', 'value'])

table1 = pd.DataFrame(data=X.values, index=X.index, columns=col_index)
table1.loc['W?'] = i_sup
table1['','Oak Type',''] = j_sup

table1

mca_ben = mca.MCA(X, ncols=ncols)
mca_ind = mca.MCA(X, ncols=ncols, benzecri=False)

print(mca.MCA.__doc__)

data = {'Iλ': pd.Series(mca_ind.L),
        'τI': mca_ind.expl_var(greenacre=False, N=4),
        'Zλ': pd.Series(mca_ben.L),
        'τZ': mca_ben.expl_var(greenacre=False, N=4),
        'cλ': pd.Series(mca_ben.L),
        'τc': mca_ind.expl_var(greenacre=True, N=4)}

# 'Indicator Matrix', 'Benzecri Correction', 'Greenacre Correction'
columns = ['Iλ', 'τI', 'Zλ', 'τZ', 'cλ', 'τc']
table2 = pd.DataFrame(data=data, columns=columns).fillna(0)
table2.index += 1
table2.loc['Σ'] = table2.sum()
table2.index.name = 'Factor'

table2

mca_ind.inertia, mca_ind.L.sum(), mca_ben.inertia, mca_ben.L.sum()

data = np.array([mca_ben.L[:2], 
                 mca_ben.expl_var(greenacre=True, N=2) * 100]).T
df = pd.DataFrame(data=data, columns=['cλ','%c'], index=range(1,3))
df

fs, cos, cont = 'Factor score','Squared cosines', 'Contributions x 1000'
table3 = pd.DataFrame(columns=X.index, index=pd.MultiIndex
                      .from_product([[fs, cos, cont], range(1, 3)]))

table3.loc[fs,    :] = mca_ben.fs_r(N=2).T
table3.loc[cos,   :] = mca_ben.cos_r(N=2).T
table3.loc[cont,  :] = mca_ben.cont_r(N=2).T * 1000
table3.loc[fs, 'W?'] = mca_ben.fs_r_sup(pd.DataFrame([i_sup]), N=2)[0]

np.round(table3.astype(float), 2)

table4 = pd.DataFrame(columns=col_index, index=pd.MultiIndex
                      .from_product([[fs, cos, cont], range(1, 3)]))
table4.loc[fs,  :] = mca_ben.fs_c(N=2).T
table4.loc[cos, :] = mca_ben.cos_c(N=2).T
table4.loc[cont,:] = mca_ben.cont_c(N=2).T * 1000

fs_c_sup = mca_ben.fs_c_sup(mca.dummy(pd.DataFrame(j_sup)), N=2)
table4.loc[fs, ('Oak', '', 1)] = fs_c_sup[0]
table4.loc[fs, ('Oak', '', 2)] = fs_c_sup[1]

np.round(table4.astype(float), 2)

%matplotlib inline
import matplotlib.pyplot as plt

points = table3.loc[fs].values
labels = table3.columns.values

plt.figure()
plt.margins(0.1)
plt.axhline(0, color='gray')
plt.axvline(0, color='gray')
plt.xlabel('Factor 1')
plt.ylabel('Factor 2')
plt.scatter(*points, s=120, marker='o', c='r', alpha=.5, linewidths=0)
for label, x, y in zip(labels, *points):
    plt.annotate(label, xy=(x, y), xytext=(x + .03, y + .03))
plt.show()

noise = 0.05 * (np.random.rand(*table4.T[fs].shape) - 0.5)
fs_by_source = table4.T[fs].add(noise).groupby(level=['source'])

fig, ax = plt.subplots()
plt.margins(0.1)
plt.axhline(0, color='gray')
plt.axvline(0, color='gray')
plt.xlabel('Factor 1')
plt.ylabel('Factor 2')
ax.margins(0.1)
markers = '^', 's', 'o', 'o'
colors = 'r', 'g', 'b', 'y'
for fscore, marker, color in zip(fs_by_source, markers, colors):
    label, points = fscore
    ax.plot(*points.T.values, marker=marker, color=color, label=label, linestyle='', alpha=.5, mew=0, ms=12)
ax.legend(numpoints=1, loc=4)
plt.show()