In Python, I have a list of tuples and a list of integers with the same length, e.g.,
a = [
[1, 2],
[3, 2],
[4, 66],
[2, 3]
]
b = [
1,
31,
31,
44
]
The k-th entry in a can thought of as being associated with the k-th entry in b.
The entries [3, 2] and [2, 3] are really the same for me, and I'd like a uniquified with that in mind. Also, I would like a list of entries of belonging to the new unique list. For the above example,
a2 = [
[1, 2],
[3, 2], # or [2, 3]
[4, 66]
]
b2 = [
[1],
[31, 44],
[31]
]
b2[0] is [1] since [1, 2] is associated with only 1. b2[1] is [31, 44] since [2, 3] (which equals [3, 2] is associated with 31 and 44 in a.
It's possible to go through a entry by entry, make each 2-list a frozenset, sort it into a dictionary etc. Needless to say, this doesn't perform very well if a and b are large.
Any hints on how to handle this smarter? (List comprehensions?)
if you want to maintain order and group I don't think you won't get much better than grouping with an OrderedDict:
from collections import OrderedDict
a = [
[1, 2],
[3, 2],
[4, 66],
[2, 3]
]
b = [1, 31, 31, 44]
d = OrderedDict()
for ind, f in enumerate(map(frozenset, a)):
d.setdefault(f, []).append(b[ind])
print(list(d), list(d.values()))
Which would give you:
[frozenset({1, 2}), frozenset({2, 3}), frozenset({66, 4})] [[1], [31, 44], [31]]
if order seen is irrelevant, use a defaultdict:
from collections import defaultdict
a = [
[1, 2],
[3, 2],
[4, 66],
[2, 3]
]
b = [1, 31, 31, 44]
d = defaultdict(list)
for ind, f in enumerate(map(frozenset, a)):
d[f].append(b[ind])
print(list(d), list(d.values()))
Which would give you:
[frozenset({1, 2}), frozenset({2, 3}), frozenset({66, 4})] [[1], [31, 44], [31]]
If you really want lists or tuples:
print(list(map(list, d)), list(d.values()))
Which would give you:
[[1, 2], [2, 3], [66, 4]] [[1], [31, 44], [31]]
For python2, you should use itertools.izip and itertools.imap in place of map and zip.