Question

我有一个使用np.genfromtxt()函数从csv文件加载的元组数组。

import numpy as np
import re
from matplotlib.dates import strpdate2num
def convert_string_to_bigint(x):
    p = re.compile(r'(\d{4})/(\d{1,2})/(\d{1,2}) (\d{1,2}):(\d{2}):\d{2}')
    m = p.findall(x)
    l = list(m[0])
    l[1] = ('0' + l[1])[-2:]
    l[2] = ('0' + l[2])[-2:]
    return long("".join(l))

#print convert_string_to_bigint("2012/7/2 14:07:00")
csv = np.genfromtxt ('sr00-1min.txt', delimiter=',', converters={0:convert_string_to_bigint})

csv文件中的数据样本：

2015/9/2 14:54:00,5169,5170,5167,5168
2015/9/2 14:55:00,5168,5169,5166,5166
2015/9/2 14:56:00,5167,5170,5165,5169
2015/9/2 14:57:00,5168,5173,5167,5172
2015/9/2 14:58:00,5172,5187,5171,5182
2015/9/2 14:59:00,5182,5183,5171,5176
2015/9/2 15:00:00,5176,5183,5174,5182

加载后，它看起来像这样：

[(201509021455L, 5168.0, 5169.0, 5166.0, 5166.0)
 (201509021456L, 5167.0, 5170.0, 5165.0, 5169.0)
 (201509021457L, 5168.0, 5173.0, 5167.0, 5172.0)
 (201509021458L, 5172.0, 5187.0, 5171.0, 5182.0)
 (201509021459L, 5182.0, 5183.0, 5171.0, 5176.0)
 (201509021500L, 5176.0, 5183.0, 5174.0, 5182.0)]

我想将它转换为numpy 2d数组。它应该是这样的：

[[201509021455L, 5168.0, 5169.0, 5166.0, 5166.0]
 [201509021456L, 5167.0, 5170.0, 5165.0, 5169.0]
 [201509021457L, 5168.0, 5173.0, 5167.0, 5172.0]
 [201509021458L, 5172.0, 5187.0, 5171.0, 5182.0]
 [201509021459L, 5182.0, 5183.0, 5171.0, 5176.0]
 [201509021500L, 5176.0, 5183.0, 5174.0, 5182.0]]

我使用下面的代码来解决这个问题，但它看起来非常难看。有人能告诉我如何以优雅的方式转换它吗？

pool = np.asarray([x for x in csv if x[0] > 201508010000])
sj = np.asarray([x[0] for x in pool])
kpj = np.asarray([x[1] for x in pool])
zgj = np.asarray([x[2] for x in pool])
zdj = np.asarray([x[3] for x in pool])
spj = np.asarray([x[4] for x in pool])
output = np.column_stack((sj,kpj,zgj,zdj,spj))
print output.shape

Answer 1

在convert_string_to_bigint中，更改

return long("".join(l))

到

return float("".join(l))

然后genfromtxt会将所有值识别为浮点数，并返回一个浮点数dtype的二维数组：

In [23]: np.genfromtxt ('sr00-1min.txt', delimiter=',', converters={0:convert_string_to_bigint}).shape
Out[23]: (7, 5)

而不是混合dtype的1D 结构化数组。

将元组数组转换为二维数组

1 个答案: