按最大日期列出分组

Question

Suppose I have a list of lists like so:

[   
    ['group1', 'type1', '2021-3-24'],
    ['group1', 'type1', '2021-3-25'],
    ['group1', 'type1', '2021-3-26'],
    ['group2', 'type2', '2022-5-21'],
    ['group2', 'type2', '2021-1-12'],
    ['group2', 'type2', '2021-3-26'],
]

and I want these results:

[   
    ['group1', 'type1', '2021-3-26'],
    ['group2', 'type2', '2022-5-21'],
]

where each list in the parent list is grouped by group and type and the function performed is a "max date" operation.

The SQL statement equivalent of what I'm looking for:

select
    group,
    type,
    max(date)
from my_list
group by
    group,
    type

I would like to avoid the overhead of Pandas as I think this can be done using itertools.groupby and my datasets are relatively small, but I just can't find a close enough example to understand how this would work.

Answer 1

You can use collections.defaultdict :

import collections, datetime
d = collections.defaultdict(list)
data = [['group1', 'type1', '2021-3-24'], ['group1', 'type1', '2021-3-25'], ['group1', 'type1', '2021-3-26'], ['group2', 'type2', '2022-5-21'], ['group2', 'type2', '2021-1-12'], ['group2', 'type2', '2021-3-26']]
for a, b, c in data:
  d[(a, b)].append(datetime.date(*map(int, c.split('-'))))

result = [[*a, str(max(b))] for a, b in d.items()]

Output:

[['group1', 'type1', '2021-03-26'], ['group2', 'type2', '2022-05-21']]

Answer 2

@Ajax answer is good, but for completeness I add version with groupby :

lst = [
    ["group1", "type1", "2021-3-24"],
    ["group1", "type1", "2021-3-25"],
    ["group1", "type1", "2021-3-26"],
    ["group2", "type2", "2022-5-21"],
    ["group2", "type2", "2021-1-12"],
    ["group2", "type2", "2021-3-26"],
]

from itertools import groupby

out = []
# if list is not sorted:
# lst = sorted(lst, key=lambda k: (k[0], k[1]))
for c, g in groupby(lst, lambda k: (k[0], k[1])):
    out.append(
        [*c, "-".join(map(str, max([*map(int, v[-1].split("-"))] for v in g)))]
    )

print(out)

Prints:

[['group1', 'type1', '2021-3-26'], ['group2', 'type2', '2022-5-21']]