Question

I have the following hierarchy:

- A
    - X : [1, 2, 3]
    - Y : [4, 5]
    - Z : [10, 11]
- B
    - X : [6, 7]
    - Y : [8]

And what I want is to have following queries give me following results:

get(A) ==> [1,2,3,4,5,10,11]
get(A,Y) ==> [4,5]
get(B) ==> [6,7,8]
get(B,X) ==> [6,7]

So far, it seems easy. I can accomplish this by having a Dictionary> which can be a defaultdict(lambda : defaultdict(list)) in Python.
However, what if I need to make it more generic and have another level, or another 2 levels?
Something like :

- A
    - X
        - i  : [1]
        - ii : [2,3]
    - Y
        - i  : [4, 5]
    - Z
        - ii : [10, 11]
- B
    - X
        - ii  : [6]
        - iii : [7]
    - Y
        - i   : [8]

In this example, the first hierarchy is a “projection” of the second hierarchy where the last level is merged into the parent. So, all queries for the first hierarchy should give the same results.

Some sample queries for new level:

get(B, X, ii) ==> [6]
get(B,X) ==> [6,7]          (same query and result as before)

Please note that, data is only in leaf nodes. So, for insertion, whole path must be given:

insert(A, X, i, 20)

That also means, we can give the depth of the tree in constructor of the data structure.

EDIT: I realized that I need validation of depth:

• Insert operation : whole path must be given and the len(path) must be equal to depth
• Get operation : a path “deeper” than the depth of the structure is not allowed

Answer 1

I wrote this class based on the idea of @black_dragon and with the support for the validation of depth.
Here is how to use it (copied from the test case):

def test_index_with_sample_case_for_depth_2(self):
    idx = HierarchicalIndex(2)

    # A
    idx.insert(1, 'A', 'X')
    idx.insert(2, 'A', 'X')
    idx.insert(3, 'A', 'X')

    idx.insert(4, 'A', 'Y')
    idx.insert(5, 'A', 'Y')

    idx.insert(10, 'A', 'Z')
    idx.insert(11, 'A', 'Z')

    #B
    idx.insert(6, 'B', 'X')
    idx.insert(7, 'B', 'X')

    idx.insert(8, 'B', 'Y')

    assert_that(idx.get('A'), equal_to([1, 2, 3, 4, 5, 10, 11]))
    assert_that(idx.get('A', 'Y'), equal_to([4, 5]))
    assert_that(idx.get('B'), equal_to([6, 7, 8]))
    assert_that(idx.get('B', 'X'), equal_to([6, 7]))


def test_index_with_sample_case_for_depth_3(self):
    idx = HierarchicalIndex(3)

    # A
    idx.insert(1, 'A', 'X', 'i')
    idx.insert(2, 'A', 'X', 'ii')
    idx.insert(3, 'A', 'X', 'ii')

    idx.insert(4, 'A', 'Y', 'i')
    idx.insert(5, 'A', 'Y', 'ii')

    idx.insert(10, 'A', 'Z', 'ii')
    idx.insert(11, 'A', 'Z', 'iii')

    #B
    idx.insert(6, 'B', 'X', 'ii')
    idx.insert(7, 'B', 'X', 'iii')

    idx.insert(8, 'B', 'Y', 'i')

    #same queries with case for depth 2
    assert_that(idx.get('A'), equal_to([1, 2, 3, 4, 5, 10, 11]))
    assert_that(idx.get('A', 'Y'), equal_to([4, 5]))
    assert_that(idx.get('B'), equal_to([6, 7, 8]))
    assert_that(idx.get('B', 'X'), equal_to([6, 7]))

    #new queries
    assert_that(idx.get('B', 'X', 'ii'), equal_to([6]))
    assert_that(idx.get('A', 'X', 'ii'), equal_to([2, 3]))

And validation of depth:

def test_index_should_validate_depth_in_operations(self):
    # ....
    # depth=3
    idx = HierarchicalIndex(3)

    assert_that(idx.get('A'), has_length(0))
    assert_that(idx.get('A', 'X'), has_length(0))
    assert_that(idx.get('A', 'X', 'i'), has_length(0))
    self.assertRaises(AssertionError, lambda: idx.get('A', 'X', 'i', '1'))

    self.assertRaises(AssertionError, lambda: idx.insert(1))
    self.assertRaises(AssertionError, lambda: idx.insert(1, 'A'))
    self.assertRaises(AssertionError, lambda: idx.insert(1, 'A', 'X'))
    idx.insert(1, 'A', 'X', 'i')        # should not raise anything
    self.assertRaises(AssertionError, lambda: idx.insert(1, 'A', 'X', 'i', 'a'))

    assert_that(idx.get('A', 'X', 'i'), equal_to([1]))