按新的日期范围重新索引数据框 [英] Re-index dataframe by new range of dates

问题描述

我有一个包含许多观察结果的数据框:

I have a data frame containing a number of observations:

date         colour     orders
2014-10-20   red        7
2014-10-21   red        10
2014-10-20   yellow     3

我想重新索引数据框并标准化日期.

I would like to re-index the data frame and standardise the dates.

date         colour     orders
2014-10-20   red        7
2014-10-21   red        10
2014-10-22   red        NaN
2014-10-20   yellow     3
2014-10-21   yellow     NaN
2014-10-22   yellow     NaN

我想按colour和date对数据框进行排序，然后尝试重新索引它.

I though to order the data frame by colour and date, and then try to re-index it.

index = pd.date_range('20/10/2014', '22/10/2014')
test_df = df.sort(['colour', 'date'], ascending=(True, True))
ts = test_df.reindex(index)
ts

但它返回一个新的数据框，该数据框具有正确的索引但所有 NaN 值.

But it returns a new data frame with the right index but all NaN values.

date         colour     orders
2014-10-20   NaN        NaN
2014-10-21   NaN        NaN
2014-10-22   NaN        NaN

推荐答案

从您的示例数据框开始:

Starting from your exampe dataframe:

In [51]: df
Out[51]:
        date  colour  orders
0 2014-10-20     red       7
1 2014-10-21     red      10
2 2014-10-20  yellow       3

如果您想在日期"和颜色"上重新索引，一种可能性是将两者都设置为索引(多索引):

If you want to reindex on both 'date' and 'colour', one possibility is to set both as the index (a multi-index):

In [52]: df = df.set_index(['date', 'colour'])

In [53]: df
Out[53]:
                   orders
date       colour
2014-10-20 red          7
2014-10-21 red         10
2014-10-20 yellow       3

您现在可以在构建所需索引后重新索引此数据框:

You can now reindex this dataframe, after you constructed to desired index:

In [54]: index = pd.date_range('20/10/2014', '22/10/2014')

In [55]: multi_index = pd.MultiIndex.from_product([index, ['red', 'yellow']])

In [56]: df.reindex(multi_index)
Out[56]:
                   orders
2014-10-20 red          7
           yellow       3
2014-10-21 red         10
           yellow     NaN
2014-10-22 red        NaN
           yellow     NaN

要获得与示例输出相同的输出，索引应在第二级排序(level=1，因为它是从 0 开始的):

To have the same output as your example output, the index should be sorted in the second level (level=1 as it is 0-based):

In [60]: df2 = df.reindex(multi_index)

In [64]: df2.sortlevel(level=1)
Out[64]:
                   orders
2014-10-20 red          7
2014-10-21 red         10
2014-10-22 red        NaN
2014-10-20 yellow       3
2014-10-21 yellow     NaN
2014-10-22 yellow     NaN

自动生成多索引的一种可能方法是(使用您的原始框架):

A possible way to generate the multi-index automatically would be (with your original frame):

pd.MultiIndex.from_product([pd.date_range(df['date'].min(), df['date'].max(), freq='D'), 
                            df['colour'].unique()])

<小时>

另一种方法是对每组颜色使用resample:

In [77]: df = df.set_index('date')

In [78]: df.groupby('colour').resample('D')

这更简单，但这不会为您提供每种颜色的完整日期范围，只会提供该颜色组可用的日期范围.

This is simpler, but this does not give you the full range of dates for each colour, only the range of dates that is available for that colour group.

这篇关于按新的日期范围重新索引数据框的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！