import time

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
import stumpy
from scipy.signal import stft
from sklearn.preprocessing import MinMaxScaler
from sklearn.decomposition import PCA,NMF
from sklearn.manifold import Isomap,LocallyLinearEmbedding,MDS,TSNE
import umap
from skimage.feature import match_template


data = pd.read_csv('TaxiData')
data.index = pd.to_datetime(data['timestamp'])
data = data.resample('30T').mean()
data = data.dropna(axis=0)
m = 48
COLUMN = 'value'
data[COLUMN] = data[COLUMN].astype(np.float)

# data = pd.read_csv('StumpyData')
# m = 640
# COLUMN = 'steam flow'



start = time.time()
mp = stumpy.stump(data[COLUMN], m)
print(time.time() - start)

start = time.time()

freq = data[COLUMN].values

freq = stft(freq, nperseg=m, noverlap=m - 1, boundary=None)[2][0:]
freq = freq / np.linalg.norm(freq, axis=0, keepdims=True)


freq = 1 - np.abs(np.dot(freq.T, freq.conj()).imag)

# freq = np.abs(freq).T



# freq = PCA().fit_transform(freq)
# dims = umap.UMAP().fit_transform(freq)
# plt.scatter(dims[:,0],dims[:,1],c=np.arange(len(dims)))
# plt.show()

print(time.time() - start)

sub_gen = (data[i:i + m] for i in range(len(data) - m))
data = data[COLUMN].values
start = time.time()
tot = np.ones((len(data) - m + 1, len(data) - m + 1))
for i, subsequence in enumerate(sub_gen):
    cross = match_template(data[i:].reshape(1,-1), subsequence.reshape(1,-1))
    tot[i, i:] = cross.flatten()
    tot[i:, i] = cross.flatten()

tot = np.sqrt(1 - tot)
np.fill_diagonal(tot,2)

print(time.time() - start)

# tot = tot / tot.diagonal()
# tot = 1 - np.abs(tot)

# freq = MinMaxScaler().fit_transform(freq.min(axis=1).reshape(-1, 1)).flatten()

mp = MinMaxScaler().fit_transform(mp[:, 0].reshape(-1, 1)).flatten()
tot = MinMaxScaler().fit_transform(tot.min(axis=1).reshape(-1, 1)).flatten()

# tot = tot.min(axis=1)
plt.plot(mp)
# plt.plot(freq)
plt.plot(tot)
plt.show()

final = pd.DataFrame(np.concatenate((mp.reshape(-1, 1),freq.reshape(-1, 1),  tot.reshape(-1, 1)),axis=1))

sns.pairplot(final,alpha=0.1)
plt.show()
print()