// Copyright 2026 International Digital Economy Academy
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
// An implementation of HAMT (Hash Array Mapped Trie) in MoonBit.
//
// Hash-Array-Mapped-Trie (HAMT) is a persistent hash-table data structure.
// It is a trie over the hash of keys (i.e. strings of binary digits)
//
// Every level in a HAMT can have up to 32 branches (5 digits),
// so HAMT has a tree height of at most 7,
// and is more efficient compared to most other tree data structures.
//
// HAMT uses bitmap-based sparse array to avoid space waste
//
// Some references:
// -
// -
///|
priv struct BuildEntry[A] {
value : A
path : @path.Path // Full path; bulk construction tracks consumed segments by depth.
}
///|
let bulk_build_threshold = 64
///|
/// Create a new instance.
#as_free_fn
pub fn[A] HashSet::new() -> HashSet[A] {
{ data: None }
}
///|
/// Lookup a value from the hash set
pub fn[A : Eq + Hash] HashSet::contains(self : HashSet[A], key : A) -> Bool {
self.data is Some(node) && node.contains(key, @path.of(key))
}
///|
fn[A : Eq] Node::contains(self : Node[A], key : A, path : @path.Path) -> Bool {
for s = self, p = path {
match (s, p) {
(Leaf(key1, bucket), _) => break key == key1 || bucket.contains(key)
(Flat(key1, path1), path) => break path == path1 && key == key1
(Branch(children), path) => {
let idx = path.idx()
if children.get(idx) is Some(child) {
continue child, path.next()
}
break false
}
}
}
}
///|
/// require: key1 != key2, path1 and path2 has the same length
#owned(key1, key2)
fn[A] join_2(
key1 : A,
path1 : @path.Path,
key2 : A,
path2 : @path.Path,
) -> Node[A] {
let idx1 = path1.idx()
let idx2 = path2.idx()
if idx1 == idx2 {
let node = if path1.is_last() {
Leaf(key2, @list.singleton(key1))
} else {
join_2(key1, path1.next(), key2, path2.next())
}
Branch(@sparse_array.singleton(idx1, node))
} else {
let (node1, node2) = if path1.is_last() {
(Leaf(key1, @list.empty()), Leaf(key2, @list.empty()))
} else {
(Flat(key1, path1.next()), Flat(key2, path2.next()))
}
Branch(@sparse_array.doubleton(idx1, node1, idx2, node2))
}
}
///|
fn[A : Eq] Node::add_with_path(
self : Node[A],
key : A,
path : @path.Path,
) -> Node[A] {
match self {
Leaf(key1, bucket) =>
if key == key1 || bucket.contains(key) {
self
} else {
Leaf(key, bucket.add(key1))
}
Flat(key1, path1) =>
if path == path1 && key == key1 {
self
} else {
join_2(key1, path1, key, path)
}
Branch(children) => {
let idx = path.idx()
match children.get(idx) {
Some(child) => {
let child = child.add_with_path(key, path.next())
Branch(children.replace(idx, child))
}
None => {
let child = Flat(key, path.next())
Branch(children.add(idx, child))
}
}
}
}
}
///|
fn[A : Eq] BuildEntry::add_to_node(
self : BuildEntry[A],
node : Node[A]?,
depth : Int,
) -> Node[A] {
let path = self.path.advance(depth)
match node {
None => Flat(self.value, path)
Some(node) => node.add_with_path(self.value, path)
}
}
///|
fn[A : Eq] build_hashset_node_by_add(
entries : Array[BuildEntry[A]],
start : Int,
end : Int,
depth : Int,
reverse : Bool,
) -> Node[A] {
if reverse {
for i = end, node = (None : Node[A]?) {
if i == start {
break node.unwrap()
}
let node = entries[i - 1].add_to_node(node, depth)
continue i - 1, Some(node)
}
} else {
for i = start, node = (None : Node[A]?) {
if i == end {
break node.unwrap()
}
let node = entries[i].add_to_node(node, depth)
continue i + 1, Some(node)
}
}
}
///|
fn[A : Eq] build_hashset_node_range(
entries : Array[BuildEntry[A]],
start : Int,
end : Int,
depth : Int,
reverse : Bool,
) -> Node[A] {
if end - start == 1 {
let entry = entries[start]
return Flat(entry.value, entry.path.advance(depth))
}
if depth == 5 {
return build_hashset_node_by_add(entries, start, end, depth, reverse)
}
let starts = FixedArray::make(32, 0)
for i in start.. Flat(value, path.push(first_idx))
_ => Branch(@sparse_array.singleton(first_idx, first_child))
}
}
let indices = FixedArray::make(child_count, first_idx)
let children = FixedArray::make(child_count, first_child)
for idx = first_idx + 1, out = 1; idx < 32; {
if nexts[idx] == starts[idx] {
continue idx + 1, out
}
indices[out] = idx
children[out] = build_hashset_node_range(
partitioned,
starts[idx],
nexts[idx],
depth + 1,
reverse,
)
continue idx + 1, out + 1
}
Branch(@sparse_array.from_sorted_fixed_array(indices, children))
}
///|
fn[A : Eq + Hash] hash_set_from_iter_by_add(iter : Iter[A]) -> HashSet[A] {
iter.fold(init=new(), (s, e) => s.add(e))
}
///|
fn[A : Eq + Hash] hash_set_from_array_by_add(arr : ArrayView[A]) -> HashSet[A] {
for n = arr.length(), set = new() {
match (n, set) {
(0, set) => break set
(n, set) => {
let k = arr[n - 1]
continue n - 1, set.add(k)
}
}
}
}
///|
fn[A : Eq + Hash] hash_set_from_array(arr : ArrayView[A]) -> HashSet[A] {
if arr.length() <= bulk_build_threshold {
return hash_set_from_array_by_add(arr)
}
let entries = Array::makei(arr.length(), i => {
let value = arr[i]
{ value, path: @path.of(value) }
})
{
data: Some(build_hashset_node_range(entries, 0, entries.length(), 0, true)),
}
}
///|
/// Add a key to the hashset.
pub fn[A : Eq + Hash] HashSet::add(self : HashSet[A], key : A) -> HashSet[A] {
{
data: match self.data {
None => Some(Flat(key, @path.of(key)))
Some(node) => Some(node.add_with_path(key, @path.of(key)))
},
}
}
///|
/// Remove an element from a set
pub fn[A : Eq + Hash] HashSet::remove(self : HashSet[A], key : A) -> HashSet[A] {
{
data: match self.data {
None => None
Some(node) => node.remove_with_path(key, @path.of(key))
},
}
}
///|
fn[A : Eq] Node::remove_with_path(
self : Node[A],
key : A,
path : @path.Path,
) -> Node[A]? {
match self {
Leaf(key1, bucket) =>
if key1 == key {
match bucket {
Empty => None
More(key2, tail=xs) => Some(Leaf(key2, xs))
}
} else if bucket.find_index(x => key.equal(x)) is Some(index) {
Some(Leaf(key1, bucket.remove_at(index)))
} else {
Some(self)
}
Flat(key1, path1) =>
if path == path1 && key == key1 {
None
} else {
Some(self)
}
Branch(children) => {
let idx = path.idx()
match children.get(idx) {
None => Some(self)
Some(child) => {
let new_child = child.remove_with_path(key, path.next())
let new_children = match (children.length(), new_child) {
(1, None) => return None
(_, None) => children.remove(idx)
(_, Some(new_child)) => children.replace(idx, new_child)
}
match new_children.data {
[Flat(key1, path1)] =>
Some(Flat(key1, path1.push(new_children.elem_info.first_idx())))
_ => Some(Branch(new_children))
}
}
}
}
}
}
///|
/// Calculate the size of a set.
///
/// WARNING: this operation is `O(N)` in set size
#alias(size, deprecated)
pub fn[A] HashSet::length(self : HashSet[A]) -> Int {
fn node_size(node) {
match node {
Leaf(_, bucket) => 1 + bucket.length()
Flat(_) => 1
Branch(children) =>
for child in children.data; total_size = 0 {
continue total_size + node_size(child)
} nobreak {
total_size
}
}
}
match self.data {
None => 0
Some(node) => node_size(node)
}
}
///|
/// Union two hashsets
pub fn[K : Eq] HashSet::union(
self : HashSet[K],
other : HashSet[K],
) -> HashSet[K] {
fn go(node1, node2) {
match (node1, node2) {
(node, Flat(key, path)) | (Flat(key, path), node) =>
node.add_with_path(key, path)
(Branch(children1), Branch(children2)) =>
Branch(children1.union(children2, go))
(Leaf(key1, bucket1), Leaf(key2, bucket2)) => {
let keys1 = bucket1.add(key1)
let keys2 = bucket2.add(key2)
match keys1.filter(k => !keys2.contains(k)) {
Empty => node2
More(head, tail~) => Leaf(key2, bucket2 + tail.add(head))
}
}
_ => abort("Unreachable")
}
}
{
data: match (self.data, other.data) {
(None, x) | (x, None) => x
(Some(a), Some(b)) => Some(go(a, b))
},
}
}
///|
/// Intersect two hashsets
pub fn[K : Eq] HashSet::intersection(
self : HashSet[K],
other : HashSet[K],
) -> HashSet[K] {
fn go(node1, node2) {
match (node1, node2) {
(node, Flat(key, path) as flat) | (Flat(key, path) as flat, node) =>
if node.contains(key, path) {
Some(flat)
} else {
None
}
(Branch(children1), Branch(children2)) =>
match children1.intersection(children2, go) {
None => None
Some({ data: [Flat(key, path)], elem_info }) =>
Some(Flat(key, path.push(elem_info.first_idx())))
Some(children) => Some(Branch(children))
}
(Leaf(key1, bucket1), Leaf(key2, bucket2)) => {
let keys1 = bucket1.add(key1)
let keys2 = bucket2.add(key2)
match keys1.filter(k => keys2.contains(k)) {
Empty => None
More(head, tail~) => Some(Leaf(head, tail))
}
}
_ => abort("Unreachable")
}
}
{
data: match (self.data, other.data) {
(None, _) | (_, None) => None
(Some(a), Some(b)) => go(a, b)
},
}
}
///|
/// Difference of two hashsets: elements in `self` but not in `other`
pub fn[K : Eq] HashSet::difference(
self : HashSet[K],
other : HashSet[K],
) -> HashSet[K] {
fn go(node1 : Node[_], node2) {
match (node1, node2) {
(node, Flat(k, path)) => node.remove_with_path(k, path)
(Flat(key, path) as flat, node) =>
if node.contains(key, path) {
None
} else {
Some(flat)
}
(Branch(children1), Branch(children2)) =>
match children1.difference(children2, go) {
None => None
Some({ data: [Flat(key, path)], elem_info }) =>
Some(Flat(key, path.push(elem_info.first_idx())))
Some(children) => Some(Branch(children))
}
(Leaf(key1, bucket1), Leaf(key2, bucket2)) => {
let keys1 = bucket1.add(key1)
let keys2 = bucket2.add(key2)
match keys1.filter(k => !keys2.contains(k)) {
Empty => None
More(head, tail~) => Some(Leaf(head, tail))
}
}
_ => abort("Unreachable")
}
}
match (self.data, other.data) {
(None, _) => { data: None }
(_, None) => self
(Some(a), Some(b)) => { data: go(a, b) }
}
}
///|
/// Returns true if the hash set is empty.
pub fn[A] HashSet::is_empty(self : HashSet[A]) -> Bool {
self.data is None
}
///|
/// Iterate through the elements in a hash set
pub fn[A] HashSet::each(
self : HashSet[A],
f : (A) -> Unit raise?,
) -> Unit raise? {
fn go(node) raise? {
match node {
Leaf(k, bucket) => {
f(k)
bucket.each(f)
}
Flat(k, _) => f(k)
Branch(children) => children.each(go)
}
}
if self.data is Some(node) {
go(node)
}
}
///|
priv enum CurrNode[A] {
Tree(Node[A])
Bucket(@list.List[A])
}
///|
/// Converted to Iter
#alias(iterator, deprecated)
pub fn[A] HashSet::iter(self : HashSet[A]) -> Iter[A] {
let empty = Bucket(@list.new())
let mut curr_node = match self.data {
Some(tree) => Tree(tree)
None => empty
}
let mut curr_index = 0
let parents = []
Iter::new(fn() {
for cn = curr_node {
match cn {
Tree(Flat(x, _)) => {
curr_node = empty
break Some(x)
}
Tree(Leaf(x, bucket)) => {
curr_node = Bucket(bucket)
break Some(x)
}
Bucket(More(x, tail~)) => {
curr_node = Bucket(tail)
break Some(x)
}
Tree(Branch(children)) as n if curr_index < children.length() => {
let child = children.data[curr_index]
parents.push((n, curr_index + 1))
curr_index = 0
continue Tree(child)
}
Bucket(Empty) | Tree(Branch(_)) if parents.pop()
is Some((parent, parent_index)) => {
curr_node = parent
curr_index = parent_index
continue parent
}
Bucket(Empty) | Tree(Branch(_)) => break None
}
}
})
}
///|
/// Creates a hash set from an iterator of values.
#as_free_fn
#alias(from_iterator, deprecated)
#as_free_fn(from_iterator, deprecated)
pub fn[A : Eq + Hash] HashSet::from_iter(iter : Iter[A]) -> HashSet[A] {
if iter.size_hint() is Some(len) && len <= bulk_build_threshold {
return hash_set_from_iter_by_add(iter)
}
let entries = match iter.size_hint() {
Some(len) => Array::new(capacity=len)
None => []
}
iter.each(value => entries.push({ value, path: @path.of(value) }))
if entries.is_empty() {
new()
} else {
{
data: Some(
build_hashset_node_range(entries, 0, entries.length(), 0, false),
),
}
}
}
///|
#deprecated("Use @debug.Debug instead of Show for debugging purposes. See https://github.com/moonbitlang/core/blob/main/debug/README.mbt.md")
pub impl[A : Show] Show for HashSet[A]
///|
pub impl[A : Show] Show for HashSet[A] with fn output(self, logger) {
logger.write_iter(
self.iter(),
prefix="@immut/hashset.from_array([",
suffix="])",
)
}
///|
/// Creates a hash set from an array of values.
#as_free_fn(deprecated="Use @immut/hashset.HashSet([...]) instead")
#alias(of, deprecated="Use @immut/hashset.HashSet([...]) instead")
#as_free_fn(of, deprecated="Use @immut/hashset.HashSet([...]) instead")
#deprecated("Use @immut/hashset.HashSet([...]) instead")
pub fn[A : Eq + Hash] HashSet::from_array(arr : ArrayView[A]) -> HashSet[A] {
hash_set_from_array(arr)
}
///|
/// Creates a hash set from an array of values.
///
/// # Example
///
/// ```mbt check
/// test {
/// let set = @hashset.HashSet([3, 1, 2, 3])
/// @test.assert_eq(set.contains(1), true)
/// @test.assert_eq(set.contains(4), false)
/// }
/// ```
pub fn[A : Eq + Hash] HashSet::HashSet(arr : ArrayView[A]) -> HashSet[A] {
hash_set_from_array(arr)
}
///|
pub impl[A : Hash] Hash for HashSet[A] with fn hash_combine(self, hasher) {
hasher.combine(self.iter().fold(init=0, (x, y) => x ^ y.hash()))
}
///|
impl[A : Eq] Eq for Node[A] with fn equal(self, other) {
if physical_equal(self, other) {
return true
}
match (self, other) {
(Leaf(x, xs), Leaf(y, ys)) =>
xs.length() == ys.length() &&
({
let keys1 = xs.add(x)
let keys2 = ys.add(y)
keys1.iter().all(k => keys2.contains(k))
})
(Flat(x, pathx), Flat(y, pathy)) => pathx == pathy && x == y
(Branch(xs), Branch(ys)) => xs == ys
_ => false
}
}
///|
test "hash" {
@test.assert_eq(
Hash::hash(HashSet([1, 2, 3, 4])),
Hash::hash(HashSet([3, 2]).add(1).add(4)),
)
@test.assert_not_eq(
Hash::hash(HashSet([1, 2, 3])),
Hash::hash(HashSet([1, 2, 4])),
)
}
///|
test "eq" {
@test.assert_eq(HashSet([1, 2, 3, 4]), HashSet([3, 2]).add(1).add(4))
@test.assert_not_eq(HashSet([1, 2, 3]), HashSet([1, 2, 4]))
}