Question

删除所有引号和引号之间的特殊字符。

我有这行："The quick "brown-fox" "jumps" over the lazy dog"，我想做的是将引号放在两端，但是要删除它们之间的所有引号和特殊字符，因此例如，该行将变为：{{1 }}。

我有一点RegEx可以找到带引号的字符串："The quick brown fox jumps over the lazy dog"

但是；我想做的实际上是删除引号之间的特殊字符。我只是最近才发现正则表达式，它可以完成很多工作，但我仍在学习。

Answer 1

您可以尝试：

import keras
import numpy as np


def generate_data(a, nt):
    ny = len(a)
    max_lag = -1
    for iy in range(ny):
        for jy in range(ny):
            max_lag = max(max_lag, len(a[iy][jy]))
    d2 = np.zeros(shape=(nt, ny, max_lag), dtype=float)
    # seed random number generator
    np.random.seed(1)
    d2[0] = np.random.rand(ny, max_lag)  # initalize lagged values for time 0
    d2[:, :, 0] = np.nan  # Set all the y values to nan
    for it in range(nt):
        for iy in range(ny):
            da = 0
            for jy in range(ny):
                lag_coeff = a[iy][jy][1:]
                n_terms = len(lag_coeff)
                if n_terms > 0:
                    lags = d2[it][jy][1:(n_terms + 1)]
                    this_term = np.matmul(lags, lag_coeff)
                    da = da - this_term
            d2[it][iy][0] = da
            if it < (nt - 1):
                d2[it + 1][iy][1:max_lag] = d2[it][iy][:(max_lag - 1)]
    x2 = d2[:, :, 1:]
    x2_1 = x2.shape[1]
    x2_2 = x2.shape[2]
    x2_product = x2_1 * x2_2
    x2 = x2.reshape((nt, x2_product))
    y2 = d2[:, :, 0].reshape((nt, ny))
    return max_lag, ny, x2, y2


def data_split(fraction_train, fraction_validate, x2, y2):
    nt = y2.shape[0]
    n_train = round(fraction_train * nt)
    n_validation = round(fraction_validate * n_train)
    x_train_val = x2[:(n_train + n_validation)]
    y_train_val = y2[:(n_train + n_validation)]
    x_train = x_train_val[:n_train]
    x_validation = x_train_val[n_train:]
    y_train = y_train_val[:n_train]
    y_validation = y_train_val[n_train:]
    x_test = x2[(n_train + n_validation):]
    y_test = y2[(n_train + n_validation):]
    return x_train, y_train, x_test, y_test, x_validation, y_validation


nt = 1000
a = [[[1, -1.1, 0.7, .169], [0, -.5, .1]],
     [[0, -.04, .104], [1, -1.2, .6, .1, -.531]]]

max_lag, ny, x_data, y_data = generate_data(a, nt)

fraction_train = 0.8
fraction_validate = 0.1
x_train, y_train, x_test, y_test, x_validation, y_validation = \
    data_split(fraction_train, fraction_validate, x_data, y_data)

x_input = keras.Input(shape=(1, None, x_train.shape[1]), name='x_input')
print('x_input.shape: ', x_input.shape)
ny_output = keras.layers.Dense(ny, name='dense_a', use_bias=False)(x_input)
print('ny_output.shape:', ny_output.shape)
print('max_lag: ', max_lag)
output_list = list()
ky_start = 0
stored_lags = max_lag - 1
for iy in range(ny):
    ky_end = ky_start + stored_lags - 1
    print('append output, {}:{}'.format(iy, iy + 1))
    output_list.append(ny_output[:, :, :, iy:(iy + 1)])  # Add the next time step
    print('append input, {}:{}'.format(ky_start, ky_end))
    output_list.append(x_input[:, :, :, ky_start:ky_end])
    ky_start = ky_end + 1  # Move to the next set of lags

outputs = keras.layers.Concatenate(axis=3)(output_list)
model = keras.Model(inputs=x_input, outputs=outputs)
print(model.summary())

JS演示：https://jsfiddle.net/jorgesanure/vca2gsxm/

要从崇高的角度使用它，请仅复制此let string = `"The quick "brown-fox" "jumps" over the lazy dog"`.replace(/(?<=.)"(?=.)/g, ""); console.log(string);

Answer 2

可以肯定的是，有很多有效的方法/表达式可以做到这一点，但我想不到一个，也许我们会尝试三种表达式，其中一种类似于：

((?:[^"\r\n-])*)|((?!(^"|"$)).)+?

并替换为$1，后跟一个空格，它将打印：

"The quick   brown  fox     jumps   over the lazy dog "

Demo 1

然后用第二个表达式

\s{2,}

并替换为一个空格：

 "The quick brown fox jumps over the lazy dog "

Demo 2

并带有第三个表达式，

\s*"\s*

替换"，我们可能会得到：

"The quick brown fox jumps over the lazy dog"

Demo 3

不需要的字符进入字符类：[^"\r\nUndesired_Chars_Goes_here-]

删除引号之间的引号

2 个答案:

Demo 1

Demo 2

Demo 3