Question

到目前为止，我正在将CSV读入numpy数组

示例行：

20041207,7.04,7.18,6.88,7.10,25981485

代码：

import datetime
import numpy as np
import matplotlib.dates as dt


def mkdate(text):
    return dt.date2num(datetime.datetime.strptime(text, '%Y%m%d'))


np.genfromtxt(
    filename,
    delimiter=',',
    skip_header=1,
    usecols=[1, 2, 5, 3, 4, 6],
    names=('date', 'open', 'close', 'high', 'low', 'volume'),
    converters={'date': mkdate},
    dtype=(
        np.float64,
        np.float64,
        np.float64,
        np.float64,
        np.float64,
        np.int64
    )
)

现在我必须切换到数据库。从数据库中获取相关值后，它看起来像这样（元组列表）：

[(datetime.datetime(2004, 12, 7, 0, 0), Decimal('7.04000'), Decimal('7.10000'), Decimal('7.18000'), Decimal('6.88000'), 25981485L), (and so on), ... ]

现在我需要转换成与以前相同的numpy数组，我想象它会是：

def mkdate(date):
    return dt.date2num(date)

np.somefunction(
    list_of_tuples,
    names=('date', 'open', 'close', 'high', 'low', 'volume'),
    converters={
        'date': mkdate,
        'open': float,
        'close': float,
        'high': float,
        'low': float,
        'volume': int,
    },
    dtype=(
        np.float64,
        np.float64,
        np.float64,
        np.float64,
        np.float64,
        np.int64
    )
)

所以要总结一下：我需要将元组列表转换为带有命名列的numpy数组。

Answer 1

如果np.readtxt()不可行，也许你可以用numpy的1D数组进行dict：

tofloat = lambda w: float(w)  # must be function not type
converters={
    'date': mkdate,
    'open': tofloat,
    'close': tofloat,
    'high': tofloat,
    'low': tofloat,
    'volume': lambda w: int(w),
}
# dict with 1D lists per column: 
dd = dict([(k,[]) for k in converters.keys()])  
with open(fname, 'r') as f:
    for l in f:  # read line
       ww = l.split(',') # split line into strings
       for w,k in zip(ww,dd.keys()):  # iterate over strings and column names
           f = converters[k]
           dd[k].append(f(w)) # append to proper dict-entry list
# convert to dict with numpy arrays:
dd_a = dict([(k,np.asarray(v)) for k,v in dd.items()])

将元组列表转换为具有命名列的numpy数组（genfromtxt用于元组列表）

1 个答案: