collectiondict - Create multidicts more easily

This package simplifies the creation of multimappings from various sources. It provides the functions collectiondict, reverse_mapping and reverse_multimapping. In all three cases the return value will be of type dict[_ValueT, Collection[_KeyT]].

All three functions expect the target collection to be provided as an argument. The supported collections are fixed. Only the built-in collections Counter, frozenset, list, set, and tuple as well as their subclasses are supported. If a unsupported collection is passed, an exception is raised. However, mypy will warn about it.

Due to the limits of Pythons type annotations, it is not possible to specify the correct return type for the custom classes. Thus, custom classes are supported but the return type is inferred to be the parent class (e.g. set), as opposed to be the class passed in (e.g. class MySet(set)).

In order to have the best type inference, it is recommended to cast clct_t to specify the value type. Passing a specialised collection class is not supported currently. The examples show how to use a cast.

collectiondict

Given any stream of key-value tuples, this function creates a multi-dictionary which maps all values to the corresponding key. Thus, collectiondict(clct, stream) is similar to dict(stream) but does not discard values. It is conceptually similar to multidict but much broader with respect to the supported types.

The implementation tries to be memory efficient and performant. It is possible to use it on extremely large streams, as long as the end result fits in memory. Thus, if a list of the stream consumes more than half of the available memory, collectiondict can still be used. For deduplicating collections, e.g. set, the stream could exceed available memory, as long as the key-value pairs do not. One of the examples covers this scenario.

Simple usage using set:

>>> from collectiondict import collectiondict
>>> collectiondict(set, [("a", 1), ("b", 2), ("a", 3)])
{'a': {1, 3}, 'b': {2}}

Usage using frozenset and a cast to have the best type inference:

>>> import typing as t
>>> from collectiondict import collectiondict
>>> clct = t.cast(t.Type[frozenset[int]], frozenset)
>>> collectiondict(clct, [("a", 1), ("b", 2), ("a", 3)])
{'a': frozenset({1, 3}), 'b': frozenset({2})}

Scenario that might exceed memory:

>>> from collectiondict import collectiondict
>>> N=1000  # could be humongous, e.g. 10**20
>>> collectiondict(set, ((str(n%2), n%3) for n in range(N)))
{'0': {0, 1, 2}, '1': {0, 1, 2}}

reverse_mapping

Given a mapping, e.g. a dictionary, from keys to values, this function reverses the mapping so it maps from values to keys. The keys are collected in a collection specified by passing a constructor as argument.

Simple usage using set:

>>> from collectiondict import reverse_mapping
>>> reverse_mapping(set, {1: "foobar", 2: "blablubb", 3: "foobar"})
{'foobar': {1, 3}, 'blablubb': {2}}

Usage using frozenset and a cast to have the best type inference:

>>> import typing as t
>>> from collectiondict import reverse_mapping
>>> clct = t.cast(t.Type[frozenset[int]], frozenset)
>>> reverse_mapping(clct, {1: "foobar", 2: "blablubb", 3: "foobar"})
{'foobar': frozenset({1, 3}), 'blablubb': frozenset({2})}

reverse_multimapping

Given a mapping, e.g. a dictionary, from keys to an iterable of values, this function reverses the mapping so it maps from values to keys. The keys are collected in a collection specified by passing a constructor as argument.

Simple usage using set:

>>> from collectiondict import reverse_multimapping
>>> reverse_multimapping(set, {1: "abc", 2: "bcd", 3: "a"})
{'a': {1, 3}, 'b': {1, 2}, 'c': {1, 2}, 'd': {2}}

Usage using frozenset and a cast to have the best type inference:

>>> import typing as t
>>> from collectiondict import reverse_multimapping
>>> clct = t.cast(t.Type[frozenset[int]], frozenset)
>>> reverse_multimapping(clct, {1: [13, 37], 2: [13, 42], 3: [42]})
{13: frozenset({1, 2}), 37: frozenset({1}), 42: frozenset({2, 3})}

Since reverse_multimapping is its own inverse, there is also a nice roundtrip behaviour:

>>> import typing as t
>>> from collectiondict import reverse_multimapping
>>> start = {1: {13, 37}, 2: {13, 42}, 3: {42}}
>>> reversed_ = reverse_multimapping(set, start)
>>> roundtripped = reverse_multimapping(set, reversed_)
>>> start == roundtripped
True