Options
All
  • Public
  • Public/Protected
  • All
Menu

@thi.ng/associative

associative

npm versionnpm downloads Twitter Follow

This project is part of the @thi.ng/umbrella monorepo.

About

Alternative Map and Set implementations with customizable equality semantics & supporting operations.

  • Array based ArraySet, Linked List based LLSet, Skiplist based SortedMap & SortedSet and customizable EquivMap implement the full ES6 Map/Set APIs and additional features:
    • range query iterators (via entries(), keys(), values()) (sorted types only)
    • ICopy, IEmpty & IEquiv implementations
    • ICompare implementation for sorted types
    • multiple value additions / updates / deletions via into(), dissoc() (maps) and disj() (sets)
    • configurable key equality & comparison (incl. default implementations)
    • getters w/ optional "not-found" default value
    • fromObject() converters (for maps only)
  • TrieMap for string-based keys and MultiTrie for array-like keys and multiple values per key
  • SparseSet implementations for numeric values
  • Polymorphic set operations (union, intersection, difference) - works with both native and custom Sets and retains their types
  • Natural & selective joins (incl. key renaming, ported from Clojure)
  • Key-value pair inversion for maps and vanilla objects
    • i.e. swaps K => V to V => K
  • Single or multi-property index generation for maps and objects
  • Key selection & renaming for maps and objects

Why?

Please see these packages for some example use cases:

The native ES6 implementations use object reference identity to determine key containment, but often it's more practical and useful to use equivalent value semantics for this purpose, especially when keys are structured data (arrays / objects).

Note: It's the user's responsibility to ensure the inserted keys are kept immutable (even if technically they're not).

Comparison with ES6 native types

// first two objects w/ equal values
a = [1, 2];
b = [1, 2];

Using native implementations

set = new Set();
set.add(a);
set.has(b);
// false

map = new Map();
map.set(a, "foo");
map.get(b);
// undefined

Using custom implementations:

set = defArraySet();
set.add(a);
set.add({a: 1});
// ArraySet { [ 1, 2 ], { a: 1 } }
set.has(b);
// true
set.has({a: 1});
// true

set = defLLSet();
set.add(a);
set.add({a: 1});
// LLSet { [ 1, 2 ], { a: 1 } }
set.has(b);
// true
set.has({a: 1});
// true

// by default EquivMap uses ArraySet for its canonical keys
map = defEquivMap();

// with custom implementation
map = defEquivMap(null, { keys: assoc.ArraySet });
map.set(a, "foo");
// EquivMap { [ 1, 2 ] => 'foo' }
map.get(b);
// "foo"

// Hash map w/ user supplied hash code function
// (here using `hash` function for arrays)
import { hash } from "@thi.ng/vectors"

m = defHashMap([], { hash })
m.set([1, 2], "a");
m.set([3, 4, 5], "b");
m.set([1, 2], "c");
// HashMap { [ 1, 2 ] => 'c', [ 3, 4, 5 ] => 'b' }

set = defSortedSet([a, [-1, 2], [-1, -2]]);
// SortedSet { [ -1, -2 ], [ -1, 2 ], [ 1, 2 ] }
set.has(b);
// true

map = defSortedMap([[a, "foo"], [[-1,-2], "bar"]]);
// SortedMap { [ -1, -2 ] => 'bar', [ 1, 2 ] => 'foo' }
map.get(b);
// "foo"

// key lookup w/ default value
map.get([3,4], "n/a");
// "n/a"

Status

STABLE - used in production

Search or submit any issues for this package

Installation

yarn add @thi.ng/associative
// ES module
<script type="module" src="https://unpkg.com/@thi.ng/associative?module" crossorigin></script>

// UMD
<script src="https://unpkg.com/@thi.ng/associative/lib/index.umd.js" crossorigin></script>

Package sizes (gzipped, pre-treeshake): ESM: 6.24 KB / CJS: 6.43 KB / UMD: 6.21 KB

Dependencies

API

Generated API docs

IEquivSet

All Set implementations in this package implement the IEquivSet interface, an extension of the native ES6 Set API.

ArraySet

Simple array based Set implementation which by default uses @thi.ng/equiv for value equivalence checking.

LLSet

Similar to ArraySet, but uses @thi.ng/dcons linked list as backing storage for values.

EquivMap

This Map implementation uses a native ES6 Map as backing storage for its key-value pairs and an additional IEquivSet implementation for canonical keys. By default uses ArraySet for this purpose.

HashMap

Map implementation w/ standard ES6 Map API, supporting any key type via hash codes computed via user supplied hash function. Uses Open Addressing / Linear Probing to resolve key collisions. Customizable via HashMapOpts constructor argument. Hash function MUST be given.

SortedMap

Alternative implementation of the ES6 Map API using a Skip list as backing store and support for configurable key equality and sorting semantics. Like with sets, uses @thi.ng/equiv & @thi.ng/compare by default.

William Pugh's (creator of this data structure) description:

"Skip lists are probabilistic data structures that have the same asymptotic expected time bounds as balanced trees, are simpler, faster and use less space."

Data structure description:

Ranged queries

map = defSortedMap([
    ["c", 3], ["a", 1], ["d", 4], ["b", 2]
]);
// SortedMap { 'a' => 1, 'b' => 2, 'c' => 3, 'd' => 4 }

// all entries
[...map.entries()]
// [ [ 'd', 4 ], [ 'c', 3 ], [ 'b', 2 ], [ 'a', 1 ] ]

// range query w/ given start key
// also works with `keys()` and `values()`
[...map.entries("c")]
// [ [ 'c', 3 ], [ 'd', 4 ] ]

// unknown start keys are ok
[...map.entries("cc")]
// [ [ 'd', 4 ] ]

// range query w/ given MAX key
[...map.entries("c", true)]
// [ [ 'a', 1 ], [ 'b', 2 ], [ 'c', 3 ] ]

SortedSet

Sorted set implementation with standard ES6 Set API, customizable value equality and comparison semantics and additional functionality:

  • range queries (via entries, keys, values)
  • multiple value addition/deletion via into() and disj()

Furthermore, this class implements the ICopy, IEmpty, ICompare and IEquiv interfaces defined by @thi.ng/api. The latter two allow instances to be used as keys themselves in other data types defined in this (and other) package(s).

This set uses a SortedMap as backing store.

SparseSet8/16/32

Sparse sets provide super fast (approx. 4x faster than the native Set impl) insertion & lookups for numeric values in the interval [0..n) . The implementation in this package provides most of the ES6 Set API and internally relies on 2 uint typed arrays, with the actual backing type dependent on n.

Furthermore, unless (or until) values are being removed from the set, they retain their original insertion order. For some use cases (e.g. deduplication of values), this property can be very useful.

// create sparse set for value range 0 - 99 (uint8 backed)
const a = defSparseSet(100);
a.into([99, 42, 66, 23, 66, 42]);
// SparseSet8 { 99, 42, 66, 23 }

a.has(66)
// true

// sparse sets are iterable
[...a]
// [ 99, 42, 66, 23 ]

// attempting to add out-of-range values will fail
a.add(100)
// SparseSet8 { 99, 42, 66, 23 }

// create sparse set for 16 bit value range 0 - 0xffff (uint16 backed)
const b = defSparseSet(0x10000);
// SparseSet16 {}

TrieMap

Tries (also called Prefix maps) are useful data structures for search based use cases, auto-complete, text indexing etc. and provide partial key matching (prefixes), suffix iteration for a common prefix, longest matching prefix queries etc.

The implementations here too feature ES6 Map-like API, similar to other types in this package, with some further trie-specific additions.

const trie = defTrieMap([
  ["hey", "en"],
  ["hello", "en"],
  ["hallo", "de"],
  ["hallo", "de-at"],
  ["hola", "es"],
  ["hold", "en"],
  ["hej", "se"],
]);

trie.knownPrefix("hole")
// "hol"

[...trie.suffixes("he")]
// [ "j", "llo", "y" ]

// w/ prefix included
[...trie.suffixes("he", true)]
// [ "hej", "hello", "hey" ]

MultiTrie

The MultiTrie is similar to TrieMap, but supports array-like keys and multiple values per key. Values are stored in sets whose implementation can be configured via ctor options.

// init w/ custom value set type (here only for illustration)
const t = defMultiTrie<string[], string>(null, { vals: () => new ArraySet() });

t.add("to be or not to be".split(" "), 1);
t.add("to be or not to be".split(" "), 2);
t.add("to be and to live".split(" "), 3);

t.get("to be or not to be".split(" "))
// Set(2) { 1, 2 }

t.knownPrefix(["to", "be", "not"]);
// [ "to", "be" ]

// auto-complete w/ custom separator between words
[...t.suffixes(["to", "be"], false, "/")]
// [ "and/to/live", "or/not/to/be" ]

Authors

Karsten Schmidt

License

© 2017 - 2020 Karsten Schmidt // Apache Software License 2.0

Generated using TypeDoc