Python 按日期列出分组

Question

假设我有一个如下所示的列表：

[(datetime.datetime(2013, 8, 8, 1, 20, 15), 2060), (datetime.datetime(2013, 8, 9, 1, 6, 14), 2055), (datetime.datetime(2013, 8, 9, 1, 21, 1), 2050), (datetime.datetime(2013, 8, 10, 1, 5, 49), 2050), (datetime.datetime(2013, 8, 10, 1, 19, 51), 2050), (datetime.datetime(2013, 8, 11, 2, 4, 53), 2050), (datetime.datetime(2013, 8, 12, 0, 29, 45), 2050), (datetime.datetime(2013, 8, 12, 0, 44, 13), 2050), (datetime.datetime(2013, 8, 13, 0, 34, 13), 2050), (datetime.datetime(2013, 8, 13, 0, 47, 29), 2050), (datetime.datetime(2013, 8, 14, 1, 30, 39), 2050), (datetime.datetime(2013, 8, 14, 1, 33, 51), 2050), (datetime.datetime(2013, 8, 15, 0, 41, 1), 2050), (datetime.datetime(2013, 8, 15, 0, 54, 45), 2050), (datetime.datetime(2013, 8, 16, 0, 29, 57), 1950), (datetime.datetime(2013, 8, 16, 0, 43, 11), 1950), (datetime.datetime(2013, 8, 17, 0, 27, 4), 1950), (datetime.datetime(2013, 8, 17, 0, 42, 30), 1950), (datetime.datetime(2013, 8, 18, 0, 26, 26), 1950), (datetime.datetime(2013, 8, 18, 0, 43, 11), 1950), (datetime.datetime(2013, 8, 19, 0, 41, 49), 1950), (datetime.datetime(2013, 8, 20, 1, 10, 23), 1950), (datetime.datetime(2013, 8, 20, 1, 23, 44), 1950), (datetime.datetime(2013, 8, 21, 0, 47, 25), 1950), (datetime.datetime(2013, 8, 21, 1, 0, 12), 1950), (datetime.datetime(2013, 8, 22, 0, 45, 21), 1950), (datetime.datetime(2013, 8, 22, 1, 4, 33), 1950), (datetime.datetime(2013, 8, 23, 0, 51, 27), 1950), (datetime.datetime(2013, 8, 23, 1, 6, 36), 1950), (datetime.datetime(2013, 8, 24, 0, 41, 3), 1950), (datetime.datetime(2013, 8, 24, 0, 53, 14), 1950), (datetime.datetime(2013, 8, 25, 0, 29, 24), 1950), (datetime.datetime(2013, 8, 25, 0, 42, 40), 1950), (datetime.datetime(2013, 8, 26, 0, 28, 13), 1950), (datetime.datetime(2013, 8, 26, 0, 43, 30), 1950), (datetime.datetime(2013, 8, 27, 0, 30, 1), 1950), (datetime.datetime(2013, 8, 27, 0, 43, 43), 1950), (datetime.datetime(2013, 8, 28, 0, 33, 19), 1950), (datetime.datetime(2013, 8, 28, 0, 49, 11), 1950), (datetime.datetime(2013, 8, 29, 0, 26, 49), 1950), (datetime.datetime(2013, 8, 29, 0, 41, 21), 1950), (datetime.datetime(2013, 8, 30, 0, 26, 13), 1950), (datetime.datetime(2013, 8, 30, 0, 42, 9), 1950), (datetime.datetime(2013, 8, 31, 0, 23, 40), 1950), (datetime.datetime(2013, 8, 31, 0, 39, 49), 1950), (datetime.datetime(2013, 9, 1, 0, 22, 2), 1950), (datetime.datetime(2013, 9, 1, 0, 38, 16), 1950), (datetime.datetime(2013, 9, 2, 0, 21, 2), 1950), (datetime.datetime(2013, 9, 2, 0, 36, 19), 1950), (datetime.datetime(2013, 9, 3, 0, 22, 16), 1950), (datetime.datetime(2013, 9, 3, 0, 39, 2), 1900)]

很明显，您可以看到这是一个元组列表，每个元组中的第一个元素是一个时间戳。已采用良好格式，由以下人员生成：

datetime.strptime(record[0], timeFormat)

第二个元素是监控值。但是，每天可能有多个记录。例如，datetime.datetime(2013, 8, 9..) 上有两条记录，它们有两个不同的值 2055 和 2050。我想要的是实际上每天的最大值。 所以在这种情况下。 2055 将是 (2013, 8, 9) 的唯一记录。

我想知道 Python 中是否有一种方便的方法可以做到这一点。类似mysql的东西：

select 
    date(timestamp), 
    max(value)
from table 
group by date(timestamp)

mysql 语句只是为了展示想法，我绝对想要一个 python 解决方案。

Answer 1

使用itertools.groupby:

>>> records = [(datetime.datetime(2013, 8, 8, 1, 20, 15), 2060), ....]
>>> import itertools
>>> [(dt, max(v for d, v in grp)) for dt, grp in itertools.groupby(records, key=lambda x: x[0].date())]
[(datetime.date(2013, 8, 8), 2060),
 (datetime.date(2013, 8, 9), 2055),
 (datetime.date(2013, 8, 10), 2050),
 ...
]

注意：假设记录已排序。如果没有，您应该先按日期对它们进行排序。

Answer 2

您可以使用collections.defaultdict（这将适用于O(N)时间的已排序和未排序数据）：

>>> from collections import defaultdict
>>> lis = [(datetime.datetime(2013, 8, 8, 1, 20, 15), 2060), (datetime.datetime(2013, 8, 9, 1, 6, 14), 2055), (datetime.datetime(2013, 8, 9, 1, 21, 1), 2050), (datetime.datetime(2013, 8, 10, 1, 5, 49), 2050), (datetime.datetime(2013, 8, 10, 1, 19, 51), 2050), (datetime.datetime(2013, 8, 11, 2, 4, 53), 2050), (datetime.datetime(2013, 8, 12, 0, 29, 45), 2050), (datetime.datetime(2013, 8, 12, 0, 44, 13), 2050), (datetime.datetime(2013, 8, 13, 0, 34, 13), 2050), (datetime.datetime(2013, 8, 13, 0, 47, 29), 2050), (datetime.datetime(2013, 8, 14, 1, 30, 39), 2050), (datetime.datetime(2013, 8, 14, 1, 33, 51), 2050), (datetime.datetime(2013, 8, 15, 0, 41, 1), 2050), (datetime.datetime(2013, 8, 15, 0, 54, 45), 2050), (datetime.datetime(2013, 8, 16, 0, 29, 57), 1950), (datetime.datetime(2013, 8, 16, 0, 43, 11), 1950), (datetime.datetime(2013, 8, 17, 0, 27, 4), 1950), (datetime.datetime(2013, 8, 17, 0, 42, 30), 1950), (datetime.datetime(2013, 8, 18, 0, 26, 26), 1950), (datetime.datetime(2013, 8, 18, 0, 43, 11), 1950), (datetime.datetime(2013, 8, 19, 0, 41, 49), 1950), (datetime.datetime(2013, 8, 20, 1, 10, 23), 1950), (datetime.datetime(2013, 8, 20, 1, 23, 44), 1950), (datetime.datetime(2013, 8, 21, 0, 47, 25), 1950), (datetime.datetime(2013, 8, 21, 1, 0, 12), 1950), (datetime.datetime(2013, 8, 22, 0, 45, 21), 1950), (datetime.datetime(2013, 8, 22, 1, 4, 33), 1950), (datetime.datetime(2013, 8, 23, 0, 51, 27), 1950), (datetime.datetime(2013, 8, 23, 1, 6, 36), 1950), (datetime.datetime(2013, 8, 24, 0, 41, 3), 1950), (datetime.datetime(2013, 8, 24, 0, 53, 14), 1950), (datetime.datetime(2013, 8, 25, 0, 29, 24), 1950), (datetime.datetime(2013, 8, 25, 0, 42, 40), 1950), (datetime.datetime(2013, 8, 26, 0, 28, 13), 1950), (datetime.datetime(2013, 8, 26, 0, 43, 30), 1950), (datetime.datetime(2013, 8, 27, 0, 30, 1), 1950), (datetime.datetime(2013, 8, 27, 0, 43, 43), 1950), (datetime.datetime(2013, 8, 28, 0, 33, 19), 1950), (datetime.datetime(2013, 8, 28, 0, 49, 11), 1950), (datetime.datetime(2013, 8, 29, 0, 26, 49), 1950), (datetime.datetime(2013, 8, 29, 0, 41, 21), 1950), (datetime.datetime(2013, 8, 30, 0, 26, 13), 1950), (datetime.datetime(2013, 8, 30, 0, 42, 9), 1950), (datetime.datetime(2013, 8, 31, 0, 23, 40), 1950), (datetime.datetime(2013, 8, 31, 0, 39, 49), 1950), (datetime.datetime(2013, 9, 1, 0, 22, 2), 1950), (datetime.datetime(2013, 9, 1, 0, 38, 16), 1950), (datetime.datetime(2013, 9, 2, 0, 21, 2), 1950), (datetime.datetime(2013, 9, 2, 0, 36, 19), 1950), (datetime.datetime(2013, 9, 3, 0, 22, 16), 1950), (datetime.datetime(2013, 9, 3, 0, 39, 2), 1900)]
>>> dic = defaultdict(list)
for dt, val in lis:
    dic[dt.date()].append(val)
...     
>>> for k, v in dic.iteritems():
    print k, max(v)
...     
2013-08-20 1950
2013-08-15 2050
2013-08-22 1950
2013-08-09 2055
2013-08-16 1950
2013-08-11 2050
2013-08-18 1950
2013-09-03 1950
2013-09-01 1950
...

正如@hughdbrown 所说，更好的方法是：

>>> dic = {}
>>> for dt, val in lis:
...     dt = dt.date()
...     dic[dt] = max(dic.get(dt,0), val)
...     
>>> for k, v in dic.iteritems():
...     print k,v
...     
2013-08-20 1950
2013-08-15 2050
2013-08-22 1950
2013-08-09 2055
2013-08-16 1950
2013-08-11 2050
2013-08-18 1950
2013-09-03 1950
2013-09-01 1950
...