#!/usr/bin/env python

import tensorflow as tf
input_num = 64
output_num = 2
def create_file(path,output_num):
    #write = tf.python_io.TFRecordWriter('train.tfrecords')
    with open(path,'r') as file:
        lines = file.readlines()
    # print lines.__len__()
        count = 0
        data = []
        featuresList = []
        labelList = []
        label = []
        label3 = []
        days = []
        for line in lines:
            word = line.split("\t")
            features = []
            #label3=[]
            for i in range(2, len(word)):
                if i < (len(word) - 1):
                    features.append(word[i].split(":")[1])
                else:
                    features.append(word[len(word) - 1].split(":")[1].split("\n")[0])
            label.append(int(word[1]))
            days.append(str(word[0]))
            count = count + 1
            #print(count)
            featuresList.append(features)
            labelList.append(label)
        for m in labelList[0]:
            label2 = []
            for k in range(output_num):
                k+=1
                #print(k,m)
                if m == k:
                    label2.append(1)
                else:
                    label2.append(0)
            label3.append(label2)
        data.append(featuresList)
        data.append(label3)
        data.append(days)
        #write.close()
        return data[0],data[1]

data = create_file("train03.txt",output_num)
#test = create_file("train04.txt",output_num)
d = tf.convert_to_tensor(data[0])#訓練集
d1 = tf.convert_to_tensor(data[1])

# 初始化權值
def weight_variable(shape):
    initial = tf.truncated_normal(shape, stddev=0.1)  # 生成一個截斷的正態分佈
    return tf.Variable(initial)


# 初始化偏置
def bias_variable(shape):
    initial = tf.constant(0.1, shape=shape)
    return tf.Variable(initial)


# 卷積層
def conv2d(x, W):
    # x input tensor of shape `[batch, in_height, in_width, in_channels]`
    # W filter / kernel tensor of shape [filter_height, filter_width, in_channels, out_channels]
    # `strides[0] = strides[3] = 1`. strides[1]代表x方向的步長，strides[2]代表y方向的步長
    # padding: A `string` from: `"SAME", "VALID"`
    return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')


# 池化層
def max_pool_2x2(x):
    # ksize [1,x,y,1]
    return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
# 定義兩個placeholder
x = tf.placeholder(tf.float32, [None, input_num])
y = tf.placeholder(tf.float32, [None, output_num])
# 改變x的格式轉為4D的向量[batch, in_height, in_width, in_channels]`
x_image = tf.reshape(x, [-1, 8, 8, 1])
# 初始化第一個卷積層的權值和偏置
W_conv1 = weight_variable([2, 2, 1, 16])  # 2*2的取樣視窗，16個卷積核從1個平面抽取特徵
b_conv1 = bias_variable([16])  # 每一個卷積核一個偏置值
# 把x_image和權值向量進行卷積，再加上偏置值，然後應用於relu啟用函式
conv2d_1 = conv2d(x_image, W_conv1) + b_conv1
h_conv1 = tf.nn.relu(conv2d_1)
h_pool1 = max_pool_2x2(h_conv1)  # 進行max-pooling output size 4x4x16
#     # 初始化第二個卷積層的權值和偏置
W_conv2 = weight_variable([2, 2, 16, 32])  # 5*5的取樣視窗，32個卷積核從16個平面抽取特徵
b_conv2 = bias_variable([32])  # 每一個卷積核一個偏置值
#     # 把h_pool1和權值向量進行卷積，再加上偏置值，然後應用於relu啟用函式
conv2d_2 = conv2d(h_pool1, W_conv2) + b_conv2
h_conv2 = tf.nn.relu(conv2d_2)
h_pool2 = max_pool_2x2(h_conv2)  # 進行max-pooling
# 8*8的圖片第一次卷積後還是8*8，第一次池化後變為4*4
# 第二次卷積後為4*4，第二次池化後變為了2*2
# 進過上面操作後得到32張2*2的平面
    # 初始化第一個全連線層的權值
W_fc1 = weight_variable([2 * 2 * 32, 512])  # 上一場有2*2*32個神經元，全連線層有512個神經元
b_fc1 = bias_variable([512])  # 512個節點
    # 把池化層2的輸出扁平化為1維
h_pool2_flat = tf.reshape(h_pool2, [-1, 2 * 2 * 32])
    # 求第一個全連線層的輸出
wx_plus_b1 = tf.matmul(h_pool2_flat, W_fc1) + b_fc1
h_fc1 = tf.nn.relu(wx_plus_b1)
    # keep_prob用來表示神經元的輸出概率
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
    # 初始化第二個全連線層
W_fc2 = weight_variable([512, output_num])
b_fc2 = bias_variable([output_num])
wx_plus_b2 = tf.matmul(h_fc1_drop, W_fc2) + b_fc2
        # 計算輸出
prediction = tf.nn.softmax(wx_plus_b2)
# 交叉熵代價函式
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(labels=y, logits=prediction))
# 使用AdamOptimizer進行優化
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
        # 結果存放在一個布林列表中
correct_prediction = tf.equal(tf.argmax(prediction, 1), tf.argmax(y, 1))  # argmax返回一維張量中最大的值所在的位置
        # 求準確率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
saver = tf.train.Saver(max_to_keep=4)
with tf.Session() as sess:
    try :
        saver.restore(sess, tf.train.latest_checkpoint("model/"))
    except:
        sess.run(tf.global_variables_initializer())
    for epoch in range(100):
        batch_xs = sess.run(d)
        batch_ys = sess.run(d1)
        sess.run(train_step, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 0.7})
        if epoch % 1 ==0:
            acc = sess.run(accuracy, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 1.0})
            print("Iter " + str(epoch) + ", Testing Accuracy= " + str(acc))
            saver.save(sess, "model/my-model2", global_step=epoch)