Python list filtering: remove subsets from list of lists.

Question

Using Python how do you reduce a list of lists by an ordered subset match [[..],[..],..]?

In the context of this question a list L is a subset of list M if M contains all members of L, and in the same order. For example, the list [1,2] is a subset of the list [1,2,3], but not of the list [2,1,3].

Example input:

a. [[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]
b. [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [1], [1, 2, 3, 4], [1, 2], [17, 18, 19, 22, 41, 48], [2, 3], [1, 2, 3], [50, 69], [1, 2, 3], [2, 3, 21], [1, 2, 3], [1, 2, 4, 8], [1, 2, 4, 5, 6]]

Expected result:

a. [[1, 2, 4, 8], [2, 3, 21], [1, 2, 3, 4, 5, 6, 7]]
b. [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [17, 18, 19, 22, 41, 48], [50, 69],  [2, 3, 21], [1, 2, 4, 8], [1, 2, 4, 5, 6]]

Further Examples:

L = [[1, 2, 3, 4, 5, 6, 7], [1, 2, 5, 6]] - No reduce

L = [[1, 2, 3, 4, 5, 6, 7], [1, 2, 3], [1, 2, 4, 8]] - Yes reduce

L = [[1, 2, 3, 4, 5, 6, 7], [7, 6, 5, 4, 3, 2, 1]] - No reduce

(Sorry for causing confusion with the incorrect data set.)

Answer 1

Edit: I really need to improve my reading comprehension. Here's the answer to what was actually asked. It exploits the fact that "A is super of B" implies "len(A) > len(B) or A == B".

def advance_to(it, value):
    """Advances an iterator until it matches the given value. Returns False
    if not found."""
    for item in it:
        if item == value:
            return True
    return False

def has_supersequence(seq, super_sequences):
    """Checks if the given sequence has a supersequence in the list of
    supersequences.""" 
    candidates = map(iter, super_sequences)
    for next_item in seq:
        candidates = [seq for seq in candidates if advance_to(seq, next_item)]
    return len(candidates) > 0

def find_supersequences(sequences):
    """Finds the supersequences in the given list of sequences.

    Sequence A is a supersequence of sequence B if B can be created by removing
    items from A."""
    super_seqs = []
    for candidate in sorted(sequences, key=len, reverse=True):
        if not has_supersequence(candidate, super_seqs):
            super_seqs.append(candidate)
    return super_seqs

print(find_supersequences([[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3],
    [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]))
#Output: [[1, 2, 3, 4, 5, 6, 7], [1, 2, 4, 8], [2, 3, 21]]

If you need to also preserve the original order of the sequences, then the find_supersequences() function needs to keep track of the positions of the sequences and sort the output afterwards.

Answer 2

list0=[[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]

for list1 in list0[:]:
    for list2 in list0:
        if list2!=list1:
            len1=len(list1)
            c=0
            for n in list2:
                if n==list1[c]:
                    c+=1
                if c==len1:
                    list0.remove(list1)
                    break

This filters list0 in place using a copy of it. This is good if the result is expected to be about the same size as the original, there is only a few "subset" to remove.

If the result is expected to be small and the original is large, you might prefer this one who is more memory freindly as it doesn't copy the original list.

list0=[[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]
result=[]

for list1 in list0:
    subset=False
    for list2 in list0:
        if list2!=list1:
            len1=len(list1)
            c=0
            for n in list2:
                if n==list1[c]:
                    c+=1
                if c==len1:
                    subset=True
                    break
            if subset:
                break
    if not subset:
        result.append(list1)

Answer 3

This seems to work:

original=[[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]

target=[[1, 2, 4, 8], [2, 3, 21], [1, 2, 3, 4, 5, 6, 7]]

class SetAndList:
    def __init__(self,aList):
        self.list=aList
        self.set=set(aList)
        self.isUnique=True
    def compare(self,aList):
        s=set(aList)
        if self.set.issubset(s):
            #print self.list,'superceded by',aList
            self.isUnique=False

def listReduce(lists):
    temp=[]
    for l in lists:
        for t in temp:
            t.compare(l)
        temp.append( SetAndList(l) )

    return [t.list for t in temp if t.isUnique]

print listReduce(original)
print target

This prints the calculated list and the target for visual comparison.

Uncomment the print line in the compare method to see how various lists get superceded.

Tested with python 2.6.2

Answer 4

I implemented a different issubseq because yours doesn't say that [1, 2, 4, 5, 6] is a subsequence of [1, 2, 3, 4, 5, 6, 7], for example (besides being painfully slow). The solution I came up with looks like this:

 def is_subseq(a, b):
    if len(a) > len(b): return False
    start = 0
    for el in a:
     while start < len(b):
      if el == b[start]:
       break
      start = start + 1
     else:
      return False
    return True

def filter_partial_matches(sets):
     return [s for s in sets if all([not(is_subseq(s, ss)) for ss in sets if s != ss])]

A simple test case, given your inputs and outputs:

>>> test = [[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]
>>> another_test = [[1, 2, 3, 4], [2, 4, 3], [3, 4, 5]]
>>> filter_partial_matches(test)
[[1, 2, 4, 8], [2, 3, 21], [1, 2, 3, 4, 5, 6, 7]]
>>> filter_partial_matches(another_test)
[[1, 2, 3, 4], [2, 4, 3], [3, 4, 5]]

Hope it helps!

Answer 5

This code should be rather memory efficient. Beyond storing your initial list of lists, this code uses negligible extra memory (no temporary sets or copies of lists are created).

def is_subset(needle,haystack):
   """ Check if needle is ordered subset of haystack in O(n)  """

   if len(haystack) < len(needle): return False

   index = 0
   for element in needle:
      try:
         index = haystack.index(element, index) + 1
      except ValueError:
         return False
   else:
      return True

def filter_subsets(lists):
   """ Given list of lists, return new list of lists without subsets  """

   for needle in lists:
      if not any(is_subset(needle, haystack) for haystack in lists
         if needle is not haystack):
         yield needle

my_lists = [[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], 
            [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]    
print list(filter_subsets(my_lists))

>>> [[1, 2, 4, 8], [2, 3, 21], [1, 2, 3, 4, 5, 6, 7]]

And, just for fun, a one-liner:

def filter_list(L):
    return [x for x in L if not any(set(x)<=set(y) for y in L if x is not y)]

Answer 6

A list is a superlist if it is not a subset of any other list. It's a subset of another list if every element of the list can be found, in order, in another list.

Here's my code:

def is_sublist_of_any_list(cand, lists):
    # Compare candidate to a single list
    def is_sublist_of_list(cand, target):
        try:
            i = 0
            for c in cand:
                i = 1 + target.index(c, i)
            return True
        except ValueError:
            return False
    # See if candidate matches any other list
    return any(is_sublist_of_list(cand, target) for target in lists if len(cand) <= len(target))

# Compare candidates to all other lists
def super_lists(lists):
    return [cand for i, cand in enumerate(lists) if not is_sublist_of_any_list(cand, lists[:i] + lists[i+1:])]

if __name__ == '__main__':
    lists = [[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]
    superlists = super_lists(lists)
    print superlists

Here are the results:

[[1, 2, 4, 8], [2, 3, 21], [1, 2, 3, 4, 5, 6, 7]]

Edit: Results for your later data set.

>>> lists = [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [1], [1, 2, 3, 4], [1, 2], [17,
 18, 19, 22, 41, 48], [2, 3], [1, 2, 3], [50, 69], [1, 2, 3], [2, 3, 21], [1, 2,
 3], [1, 2, 4, 8], [1, 2, 4, 5, 6]]
>>> superlists = super_lists(lists)
>>> expected = [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [17, 18, 19, 22, 41, 48], [5
0, 69],  [2, 3, 21], [1, 2, 4, 8]]
>>> assert(superlists == expected)
>>> print superlists
[[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [17, 18, 19, 22, 41, 48], [50, 69], [2, 3,
21], [1, 2, 4, 8]]

Answer 7

This could be simplified, but:

l = [[1, 2, 4, 8], [1, 2, 4, 5, 6], [1, 2, 3], [2, 3, 21], [1, 2, 3, 4], [1, 2, 3, 4, 5, 6, 7]]
l2 = l[:]

for m in l:
    for n in l:
        if set(m).issubset(set(n)) and m != n:
            l2.remove(m)
            break

print l2
[[1, 2, 4, 8], [2, 3, 21], [1, 2, 3, 4, 5, 6, 7]]

Answer 8

Refined answer after new test case:

original= [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [1], [1, 2, 3, 4], [1, 2], [17, 18, 19, 22, 41, 48], [2, 3], [1, 2, 3], [50, 69], [1, 2, 3], [2, 3, 21], [1, 2, 3], [1, 2, 4, 8], [1, 2, 4, 5, 6]]

class SetAndList:
    def __init__(self,aList):
        self.list=aList
        self.set=set(aList)
        self.isUnique=True
    def compare(self,other):
        if self.set.issubset(other.set):
            #print self.list,'superceded by',other.list
            self.isUnique=False

def listReduce(lists):
    temp=[]
    for l in lists:
        s=SetAndList(l)
        for t in temp:
            t.compare(s)
            s.compare(t)
        temp.append( s )
        temp=[t for t in temp if t.isUnique]

    return [t.list for t in temp if t.isUnique]

print listReduce(original)

You didn't give the required output, but I'm guessing this is right, as [1,2,3] does not appear in the output.

Answer 9

Thanks to all who suggested solutions and coping with my sometimes erroneous data sets. Using @hughdbrown solution I modified it to what I wanted:

The modification was to use a sliding window over the target to ensure the subset sequence was found. I think I should have used a more appropriate word than 'Set' to describe my problem.

def is_sublist_of_any_list(cand, lists):
    # Compare candidate to a single list
    def is_sublist_of_list(cand, target):
        try:
            i = 0            
            try:
                start = target.index(cand[0])
            except:
                return False

            while start < (len(target) + len(cand)) - start:
                if cand == target[start:len(cand)]:
                    return True
                else:
                    start = target.index(cand[0], start + 1)
        except ValueError:
            return False

    # See if candidate matches any other list
    return any(is_sublist_of_list(cand, target) for target in lists if len(cand) <= len(target))

# Compare candidates to all other lists
def super_lists(lists):
    a = [cand for i, cand in enumerate(lists) if not is_sublist_of_any_list(cand, lists[:i] + lists[i+1:])]
    return a

lists = [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [1], [1, 2, 3, 4], [1, 2], [17, 18, 19, 22, 41, 48], [2, 3], [1, 2, 3], [50, 69], [1, 2, 3], [2, 3, 21], [1, 2, 3], [1, 2, 4, 8], [1, 2, 4, 5, 6]]
expect = [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [17, 18, 19, 22, 41, 48], [50, 69],  [2, 3, 21], [1, 2, 4, 8], [1, 2, 4, 5, 6]]

def test():
    out = super_lists(list(lists))

    print "In  : ", lists
    print "Out : ", out

    assert (out == expect)

Result:

In  :  [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [1], [1, 2, 3, 4], [1, 2], [17, 18, 19, 22, 41, 48], [2, 3], [1, 2, 3], [50, 69], [1, 2, 3], [2, 3, 21], [1, 2, 3], [1, 2, 4, 8], [1, 2, 4, 5, 6]]
Out :  [[2, 16, 17], [1, 2, 3, 4, 5, 6, 7], [17, 18, 19, 22, 41, 48], [50, 69], [2, 3, 21], [1, 2, 4, 8], [1, 2, 4, 5, 6]]

Answer 10

So what you really wanted was to know if a list was a substring, so to speak, of another, with all the matching elements consecutive. Here is code that converts the candidate and the target list to comma-separated strings and does a substring comparison to see if the candidate appears within the target list

def is_sublist_of_any_list(cand, lists):
    def comma_list(l):
        return "," + ",".join(str(x) for x in l) + ","
    cand = comma_list(cand)
    return any(cand in comma_list(target) for target in lists if len(cand) <= len(target))


def super_lists(lists):
    return [cand for i, cand in enumerate(lists) if not is_sublist_of_any_list(cand, lists[:i] + lists[i+1:])]

The function comma_list() puts leading and trailing commas on the list to ensure that integers are fully delimited. Otherwise, [1] would be a subset of [100], for example.