2021-01-18 15:20:06 +03:00
|
|
|
// Copyright (c) 2019-2021 Alexander Medvednikov. All rights reserved.
|
2019-06-23 05:21:30 +03:00
|
|
|
// Use of this source code is governed by an MIT license
|
|
|
|
// that can be found in the LICENSE file.
|
2019-06-22 21:20:28 +03:00
|
|
|
module builtin
|
|
|
|
|
2020-12-06 02:24:24 +03:00
|
|
|
// import hash.wyhash as hash
|
2020-07-18 14:49:00 +03:00
|
|
|
import hash
|
2020-04-08 01:02:15 +03:00
|
|
|
|
2020-03-19 08:52:34 +03:00
|
|
|
/*
|
2020-05-17 14:51:18 +03:00
|
|
|
This is a highly optimized hashmap implementation. It has several traits that
|
2020-04-12 15:10:06 +03:00
|
|
|
in combination makes it very fast and memory efficient. Here is a short expl-
|
|
|
|
anation of each trait. After reading this you should have a basic understand-
|
|
|
|
ing of how it functions:
|
|
|
|
|
|
|
|
1. Hash-function: Wyhash. Wyhash is the fastest hash-function for short keys
|
|
|
|
passing SMHasher, so it was an obvious choice.
|
|
|
|
|
|
|
|
2. Open addressing: Robin Hood Hashing. With this method, a hash-collision is
|
|
|
|
resolved by probing. As opposed to linear probing, Robin Hood hashing has a
|
|
|
|
simple but clever twist: As new keys are inserted, old keys are shifted arou-
|
|
|
|
nd in a way such that all keys stay reasonably close to the slot they origin-
|
|
|
|
ally hash to. A new key may displace a key already inserted if its probe cou-
|
2020-05-17 14:51:18 +03:00
|
|
|
nt is larger than that of the key at the current position.
|
2020-04-12 15:10:06 +03:00
|
|
|
|
|
|
|
3. Memory layout: key-value pairs are stored in a `DenseArray`. This is a dy-
|
|
|
|
namic array with a very low volume of unused memory, at the cost of more rea-
|
|
|
|
llocations when inserting elements. It also preserves the order of the key-v-
|
|
|
|
alues. This array is named `key_values`. Instead of probing a new key-value,
|
|
|
|
this map probes two 32-bit numbers collectively. The first number has its 8
|
|
|
|
most significant bits reserved for the probe-count and the remaining 24 bits
|
|
|
|
are cached bits from the hash which are utilized for faster re-hashing. This
|
|
|
|
number is often referred to as `meta`. The other 32-bit number is the index
|
|
|
|
at which the key-value was pushed to in `key_values`. Both of these numbers
|
|
|
|
are stored in a sparse array `metas`. The `meta`s and `kv_index`s are stored
|
|
|
|
at even and odd indices, respectively:
|
|
|
|
|
|
|
|
metas = [meta, kv_index, 0, 0, meta, kv_index, 0, 0, meta, kv_index, ...]
|
|
|
|
key_values = [kv, kv, kv, ...]
|
|
|
|
|
|
|
|
4. The size of metas is a power of two. This enables the use of bitwise AND
|
2020-05-26 18:59:52 +03:00
|
|
|
to convert the 64-bit hash to a bucket/index that doesn't overflow metas. If
|
2020-04-12 15:10:06 +03:00
|
|
|
the size is power of two you can use "hash & (SIZE - 1)" instead of "hash %
|
|
|
|
SIZE". Modulo is extremely expensive so using '&' is a big performance impro-
|
|
|
|
vement. The general concern with this approach is that you only make use of
|
|
|
|
the lower bits of the hash which can cause more collisions. This is solved by
|
|
|
|
using a well-dispersed hash-function.
|
|
|
|
|
|
|
|
5. The hashmap keeps track of the highest probe_count. The trick is to alloc-
|
|
|
|
ate `extra_metas` > max(probe_count), so you never have to do any bounds-che-
|
2020-05-17 14:51:18 +03:00
|
|
|
cking since the extra meta memory ensures that a meta will never go beyond
|
2020-03-21 15:55:07 +03:00
|
|
|
the last index.
|
2020-03-19 08:52:34 +03:00
|
|
|
|
2020-04-12 15:10:06 +03:00
|
|
|
6. Cached rehashing. When the `load_factor` of the map exceeds the `max_load_
|
|
|
|
factor` the size of metas is doubled and all the key-values are "rehashed" to
|
|
|
|
find the index for their meta's in the new array. Instead of rehashing compl-
|
|
|
|
etely, it simply uses the cached-hashbits stored in the meta, resulting in
|
|
|
|
much faster rehashing.
|
2020-03-19 08:52:34 +03:00
|
|
|
*/
|
2020-01-24 22:13:59 +03:00
|
|
|
const (
|
2020-05-26 18:59:52 +03:00
|
|
|
// Number of bits from the hash stored for each entry
|
2020-05-09 13:42:01 +03:00
|
|
|
hashbits = 24
|
2020-03-21 15:55:07 +03:00
|
|
|
// Number of bits from the hash stored for rehashing
|
2020-04-05 23:09:52 +03:00
|
|
|
max_cached_hashbits = 16
|
2020-02-20 22:04:06 +03:00
|
|
|
// Initial log-number of buckets in the hashtable
|
2020-05-09 13:42:01 +03:00
|
|
|
init_log_capicity = 5
|
2020-02-20 22:04:06 +03:00
|
|
|
// Initial number of buckets in the hashtable
|
2020-05-09 13:42:01 +03:00
|
|
|
init_capicity = 1 << init_log_capicity
|
2020-06-21 17:51:02 +03:00
|
|
|
// Maximum load-factor (len / capacity)
|
2020-05-09 13:42:01 +03:00
|
|
|
max_load_factor = 0.8
|
2020-03-21 15:55:07 +03:00
|
|
|
// Initial highest even index in metas
|
2020-12-21 07:34:35 +03:00
|
|
|
init_even_index = init_capicity - 2
|
2020-03-19 08:52:34 +03:00
|
|
|
// Used for incrementing `extra_metas` when max
|
|
|
|
// probe count is too high, to avoid overflow
|
2020-05-09 13:42:01 +03:00
|
|
|
extra_metas_inc = 4
|
2020-02-20 22:04:06 +03:00
|
|
|
// Bitmask to select all the hashbits
|
2020-05-09 13:42:01 +03:00
|
|
|
hash_mask = u32(0x00FFFFFF)
|
2020-02-20 22:04:06 +03:00
|
|
|
// Used for incrementing the probe-count
|
2020-05-09 13:42:01 +03:00
|
|
|
probe_inc = u32(0x01000000)
|
2020-01-24 22:13:59 +03:00
|
|
|
)
|
|
|
|
|
2021-02-06 05:07:42 +03:00
|
|
|
// fast_string_eq is intended to be fast when
|
2020-04-12 15:10:06 +03:00
|
|
|
// the strings are very likely to be equal
|
|
|
|
// TODO: add branch prediction hints
|
|
|
|
[inline]
|
2020-10-15 13:32:28 +03:00
|
|
|
fn fast_string_eq(a string, b string) bool {
|
2020-04-12 15:10:06 +03:00
|
|
|
if a.len != b.len {
|
|
|
|
return false
|
|
|
|
}
|
2020-07-20 20:06:41 +03:00
|
|
|
unsafe {
|
|
|
|
return C.memcmp(a.str, b.str, b.len) == 0
|
|
|
|
}
|
2020-04-12 15:10:06 +03:00
|
|
|
}
|
|
|
|
|
2021-02-06 05:07:42 +03:00
|
|
|
// DenseArray represents a dynamic array with very low growth factor
|
2020-03-12 14:19:13 +03:00
|
|
|
struct DenseArray {
|
2020-12-06 00:53:50 +03:00
|
|
|
key_bytes int
|
2021-03-27 20:59:51 +03:00
|
|
|
key_bytes_ int
|
2020-05-04 15:56:19 +03:00
|
|
|
value_bytes int
|
2020-12-06 00:53:50 +03:00
|
|
|
slot_bytes int // sum of 2 fields above
|
2020-03-12 14:19:13 +03:00
|
|
|
mut:
|
2021-01-12 06:38:43 +03:00
|
|
|
cap int
|
|
|
|
len int
|
|
|
|
deletes u32 // count
|
2020-12-13 04:05:56 +03:00
|
|
|
// array allocated (with `cap` bytes) on first deletion
|
|
|
|
// has non-zero element when key deleted
|
|
|
|
all_deleted &byte
|
|
|
|
data byteptr // array of interleaved key data and value data
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
[inline]
|
2020-12-06 00:53:50 +03:00
|
|
|
fn new_dense_array(key_bytes int, value_bytes int) DenseArray {
|
2021-03-27 20:59:51 +03:00
|
|
|
mut key_bytes_ := key_bytes
|
|
|
|
mut value_bytes_ := value_bytes
|
2021-03-24 20:49:16 +03:00
|
|
|
$if gcboehm ? {
|
|
|
|
align, mask := $if x64 { 7, int(0xfffffff8) } $else { 3, int(0xfffffffc) }
|
2021-03-27 20:59:51 +03:00
|
|
|
key_bytes_ = (key_bytes + align) & mask
|
|
|
|
value_bytes_ = (value_bytes + align) & mask
|
2021-03-24 20:49:16 +03:00
|
|
|
}
|
2021-03-27 20:59:51 +03:00
|
|
|
slot_bytes := key_bytes_ + value_bytes_
|
2020-12-06 00:53:50 +03:00
|
|
|
cap := 8
|
2020-05-04 15:56:19 +03:00
|
|
|
return DenseArray{
|
2020-12-06 00:53:50 +03:00
|
|
|
key_bytes: key_bytes
|
2021-03-27 20:59:51 +03:00
|
|
|
key_bytes_: key_bytes_
|
2020-05-04 15:56:19 +03:00
|
|
|
value_bytes: value_bytes
|
2020-12-06 00:53:50 +03:00
|
|
|
slot_bytes: slot_bytes
|
|
|
|
cap: cap
|
2020-06-21 17:51:02 +03:00
|
|
|
len: 0
|
2020-05-04 15:56:19 +03:00
|
|
|
deletes: 0
|
2020-12-13 04:05:56 +03:00
|
|
|
all_deleted: 0
|
2021-02-14 21:31:42 +03:00
|
|
|
data: unsafe { malloc(cap * slot_bytes) }
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-12-03 22:12:53 +03:00
|
|
|
[inline]
|
|
|
|
fn (d &DenseArray) key(i int) voidptr {
|
2020-12-23 21:13:42 +03:00
|
|
|
return unsafe { d.data + i * d.slot_bytes }
|
2020-12-03 22:12:53 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
// for cgen
|
|
|
|
[inline]
|
|
|
|
fn (d &DenseArray) value(i int) voidptr {
|
2021-03-27 20:59:51 +03:00
|
|
|
return unsafe { d.data + i * d.slot_bytes + d.key_bytes_ }
|
2020-12-03 22:12:53 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
[inline]
|
|
|
|
fn (d &DenseArray) has_index(i int) bool {
|
2020-12-23 21:13:42 +03:00
|
|
|
return d.deletes == 0 || unsafe { d.all_deleted[i] } == 0
|
2020-12-03 22:12:53 +03:00
|
|
|
}
|
|
|
|
|
2020-12-27 16:18:46 +03:00
|
|
|
// Make space to append an element and return index
|
2020-04-12 15:10:06 +03:00
|
|
|
// The growth-factor is roughly 1.125 `(x + (x >> 3))`
|
2020-03-19 08:52:34 +03:00
|
|
|
[inline]
|
2020-12-27 16:18:46 +03:00
|
|
|
fn (mut d DenseArray) expand() int {
|
2021-03-14 15:54:35 +03:00
|
|
|
old_cap := d.cap
|
|
|
|
old_size := d.slot_bytes * old_cap
|
2020-06-21 17:51:02 +03:00
|
|
|
if d.cap == d.len {
|
2020-05-04 15:56:19 +03:00
|
|
|
d.cap += d.cap >> 3
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
2021-03-14 15:54:35 +03:00
|
|
|
d.data = realloc_data(d.data, old_size, d.slot_bytes * d.cap)
|
2020-12-13 04:05:56 +03:00
|
|
|
if d.deletes != 0 {
|
2021-03-14 15:54:35 +03:00
|
|
|
d.all_deleted = realloc_data(d.all_deleted, old_cap, d.cap)
|
2020-12-13 04:05:56 +03:00
|
|
|
C.memset(d.all_deleted + d.len, 0, d.cap - d.len)
|
|
|
|
}
|
2020-07-15 22:56:50 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
2020-06-21 17:51:02 +03:00
|
|
|
push_index := d.len
|
2020-07-03 19:10:10 +03:00
|
|
|
unsafe {
|
2020-12-13 04:05:56 +03:00
|
|
|
if d.deletes != 0 {
|
|
|
|
d.all_deleted[push_index] = 0
|
|
|
|
}
|
2020-07-03 19:10:10 +03:00
|
|
|
}
|
2020-06-21 17:51:02 +03:00
|
|
|
d.len++
|
2020-03-19 08:52:34 +03:00
|
|
|
return push_index
|
|
|
|
}
|
|
|
|
|
2020-06-25 00:31:19 +03:00
|
|
|
// Move all zeros to the end of the array and resize array
|
2020-05-17 14:51:18 +03:00
|
|
|
fn (mut d DenseArray) zeros_to_end() {
|
2020-12-06 00:53:50 +03:00
|
|
|
// TODO alloca?
|
2021-02-14 21:31:42 +03:00
|
|
|
mut tmp_buf := unsafe { malloc(d.slot_bytes) }
|
2020-12-06 00:53:50 +03:00
|
|
|
mut count := 0
|
|
|
|
for i in 0 .. d.len {
|
|
|
|
if d.has_index(i) {
|
|
|
|
// swap (TODO: optimize)
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
2020-12-06 00:53:50 +03:00
|
|
|
C.memcpy(tmp_buf, d.key(count), d.slot_bytes)
|
|
|
|
C.memcpy(d.key(count), d.key(i), d.slot_bytes)
|
|
|
|
C.memcpy(d.key(i), tmp_buf, d.slot_bytes)
|
2020-07-03 19:10:10 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
count++
|
|
|
|
}
|
|
|
|
}
|
2021-02-14 21:31:42 +03:00
|
|
|
unsafe {
|
|
|
|
free(tmp_buf)
|
|
|
|
d.deletes = 0
|
|
|
|
// TODO: reallocate instead as more deletes are likely
|
|
|
|
free(d.all_deleted)
|
|
|
|
}
|
2020-06-21 17:51:02 +03:00
|
|
|
d.len = count
|
2021-03-14 15:54:35 +03:00
|
|
|
old_cap := d.cap
|
2020-12-06 00:53:50 +03:00
|
|
|
d.cap = if count < 8 { 8 } else { count }
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
2021-03-14 15:54:35 +03:00
|
|
|
d.data = realloc_data(d.data, d.slot_bytes * old_cap, d.slot_bytes * d.cap)
|
2020-07-15 22:56:50 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
|
|
|
|
2020-12-27 16:18:46 +03:00
|
|
|
type MapHashFn = fn (voidptr) u64
|
|
|
|
|
|
|
|
type MapEqFn = fn (voidptr, voidptr) bool
|
|
|
|
|
|
|
|
type MapCloneFn = fn (voidptr, voidptr)
|
|
|
|
|
|
|
|
type MapFreeFn = fn (voidptr)
|
|
|
|
|
2021-02-06 05:07:42 +03:00
|
|
|
// map is the internal representation of a V `map` type.
|
2020-03-19 08:52:34 +03:00
|
|
|
pub struct map {
|
2020-12-14 15:57:53 +03:00
|
|
|
// Number of bytes of a key
|
2021-03-27 20:59:51 +03:00
|
|
|
key_bytes int
|
|
|
|
key_bytes_ int
|
2020-06-25 00:31:19 +03:00
|
|
|
// Number of bytes of a value
|
2021-01-12 06:38:43 +03:00
|
|
|
value_bytes int
|
2020-03-19 08:52:34 +03:00
|
|
|
mut:
|
2020-06-25 00:31:19 +03:00
|
|
|
// Highest even index in the hashtable
|
2021-01-12 06:38:43 +03:00
|
|
|
even_index u32
|
2021-02-12 03:02:33 +03:00
|
|
|
// Number of cached hashbits left for rehashing
|
2020-05-09 13:42:01 +03:00
|
|
|
cached_hashbits byte
|
2020-03-19 08:52:34 +03:00
|
|
|
// Used for right-shifting out used hashbits
|
2021-01-12 06:38:43 +03:00
|
|
|
shift byte
|
2020-03-21 15:55:07 +03:00
|
|
|
// Array storing key-values (ordered)
|
2021-01-12 06:38:43 +03:00
|
|
|
key_values DenseArray
|
2020-03-21 15:55:07 +03:00
|
|
|
// Pointer to meta-data:
|
2020-06-25 00:31:19 +03:00
|
|
|
// - Odd indices store kv_index.
|
|
|
|
// - Even indices store probe_count and hashbits.
|
2021-01-12 06:38:43 +03:00
|
|
|
metas &u32
|
2020-03-19 08:52:34 +03:00
|
|
|
// Extra metas that allows for no ranging when incrementing
|
|
|
|
// index in the hashmap
|
2020-05-09 13:42:01 +03:00
|
|
|
extra_metas u32
|
2020-12-27 16:18:46 +03:00
|
|
|
has_string_keys bool
|
|
|
|
hash_fn MapHashFn
|
|
|
|
key_eq_fn MapEqFn
|
|
|
|
clone_fn MapCloneFn
|
|
|
|
free_fn MapFreeFn
|
2020-03-19 08:52:34 +03:00
|
|
|
pub mut:
|
2020-04-12 15:10:06 +03:00
|
|
|
// Number of key-values currently in the hashmap
|
2021-01-12 06:38:43 +03:00
|
|
|
len int
|
2020-03-12 14:19:13 +03:00
|
|
|
}
|
|
|
|
|
2020-12-27 16:18:46 +03:00
|
|
|
fn map_hash_string(pkey voidptr) u64 {
|
2021-01-05 21:07:45 +03:00
|
|
|
key := *unsafe { &string(pkey) }
|
2020-12-27 16:18:46 +03:00
|
|
|
return hash.wyhash_c(key.str, u64(key.len), 0)
|
|
|
|
}
|
|
|
|
|
|
|
|
fn map_hash_int_1(pkey voidptr) u64 {
|
2021-01-05 21:07:45 +03:00
|
|
|
return hash.wyhash64_c(*unsafe { &byte(pkey) }, 0)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_hash_int_2(pkey voidptr) u64 {
|
2021-01-05 21:07:45 +03:00
|
|
|
return hash.wyhash64_c(*unsafe { &u16(pkey) }, 0)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_hash_int_4(pkey voidptr) u64 {
|
2021-01-05 21:07:45 +03:00
|
|
|
return hash.wyhash64_c(*unsafe { &u32(pkey) }, 0)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_hash_int_8(pkey voidptr) u64 {
|
2021-01-05 21:07:45 +03:00
|
|
|
return hash.wyhash64_c(*unsafe { &u64(pkey) }, 0)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_eq_string(a voidptr, b voidptr) bool {
|
2021-01-05 21:07:45 +03:00
|
|
|
return fast_string_eq(*unsafe { &string(a) }, *unsafe { &string(b) })
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_eq_int_1(a voidptr, b voidptr) bool {
|
2021-03-18 17:23:29 +03:00
|
|
|
return unsafe { C.memcmp(a, b, 1) == 0 }
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_eq_int_2(a voidptr, b voidptr) bool {
|
2021-03-18 17:23:29 +03:00
|
|
|
return unsafe { C.memcmp(a, b, 2) == 0 }
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_eq_int_4(a voidptr, b voidptr) bool {
|
2021-03-18 17:23:29 +03:00
|
|
|
return unsafe { C.memcmp(a, b, 4) == 0 }
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_eq_int_8(a voidptr, b voidptr) bool {
|
2021-03-18 17:23:29 +03:00
|
|
|
return unsafe { C.memcmp(a, b, 8) == 0 }
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_clone_string(dest voidptr, pkey voidptr) {
|
|
|
|
unsafe {
|
|
|
|
s := *&string(pkey)
|
|
|
|
(*&string(dest)) = s.clone()
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fn map_clone_int_1(dest voidptr, pkey voidptr) {
|
|
|
|
unsafe {
|
2021-03-18 17:23:29 +03:00
|
|
|
C.memcpy(dest, pkey, 1)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fn map_clone_int_2(dest voidptr, pkey voidptr) {
|
|
|
|
unsafe {
|
2021-03-18 17:23:29 +03:00
|
|
|
C.memcpy(dest, pkey, 2)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fn map_clone_int_4(dest voidptr, pkey voidptr) {
|
|
|
|
unsafe {
|
2021-03-18 17:23:29 +03:00
|
|
|
C.memcpy(dest, pkey, 4)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fn map_clone_int_8(dest voidptr, pkey voidptr) {
|
|
|
|
unsafe {
|
2021-03-18 17:23:29 +03:00
|
|
|
C.memcpy(dest, pkey, 8)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fn map_free_string(pkey voidptr) {
|
2021-02-14 21:31:42 +03:00
|
|
|
unsafe {
|
|
|
|
(*&string(pkey)).free()
|
|
|
|
}
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
fn map_free_nop(_ voidptr) {
|
|
|
|
}
|
|
|
|
|
2020-12-27 20:31:50 +03:00
|
|
|
fn new_map_2(key_bytes int, value_bytes int, hash_fn MapHashFn, key_eq_fn MapEqFn, clone_fn MapCloneFn, free_fn MapFreeFn) map {
|
|
|
|
metasize := int(sizeof(u32) * (init_capicity + extra_metas_inc))
|
|
|
|
// for now assume anything bigger than a pointer is a string
|
|
|
|
has_string_keys := key_bytes > sizeof(voidptr)
|
2021-03-27 20:59:51 +03:00
|
|
|
mut key_bytes_ := key_bytes
|
|
|
|
$if gcboehm ? {
|
|
|
|
align, mask := $if x64 { 7, int(0xfffffff8) } $else { 3, int(0xfffffffc) }
|
|
|
|
key_bytes_ = (key_bytes + align) & mask
|
|
|
|
}
|
2020-12-27 20:31:50 +03:00
|
|
|
return map{
|
|
|
|
key_bytes: key_bytes
|
2021-03-27 20:59:51 +03:00
|
|
|
key_bytes_: key_bytes_
|
2020-12-27 20:31:50 +03:00
|
|
|
value_bytes: value_bytes
|
|
|
|
even_index: init_even_index
|
|
|
|
cached_hashbits: max_cached_hashbits
|
|
|
|
shift: init_log_capicity
|
|
|
|
key_values: new_dense_array(key_bytes, value_bytes)
|
|
|
|
metas: &u32(vcalloc(metasize))
|
|
|
|
extra_metas: extra_metas_inc
|
|
|
|
len: 0
|
|
|
|
has_string_keys: has_string_keys
|
|
|
|
hash_fn: hash_fn
|
|
|
|
key_eq_fn: key_eq_fn
|
|
|
|
clone_fn: clone_fn
|
|
|
|
free_fn: free_fn
|
|
|
|
}
|
2020-12-21 08:03:59 +03:00
|
|
|
}
|
|
|
|
|
2020-12-27 20:31:50 +03:00
|
|
|
fn new_map_init_2(hash_fn MapHashFn, key_eq_fn MapEqFn, clone_fn MapCloneFn, free_fn MapFreeFn, n int, key_bytes int, value_bytes int, keys voidptr, values voidptr) map {
|
|
|
|
mut out := new_map_2(key_bytes, value_bytes, hash_fn, key_eq_fn, clone_fn, free_fn)
|
|
|
|
// TODO pre-allocate n slots
|
|
|
|
mut pkey := byteptr(keys)
|
|
|
|
mut pval := byteptr(values)
|
|
|
|
for _ in 0 .. n {
|
|
|
|
unsafe {
|
|
|
|
out.set_1(pkey, pval)
|
|
|
|
pkey += key_bytes
|
|
|
|
pval += value_bytes
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return out
|
2020-12-21 08:03:59 +03:00
|
|
|
}
|
|
|
|
|
2021-02-12 03:02:33 +03:00
|
|
|
pub fn (mut m map) move() map {
|
|
|
|
r := *m
|
|
|
|
unsafe {
|
|
|
|
C.memset(m, 0, sizeof(map))
|
|
|
|
}
|
|
|
|
return r
|
|
|
|
}
|
|
|
|
|
2020-12-14 15:57:53 +03:00
|
|
|
[inline]
|
|
|
|
fn (m &map) key_to_index(pkey voidptr) (u32, u32) {
|
2020-12-27 16:18:46 +03:00
|
|
|
hash := m.hash_fn(pkey)
|
2020-12-21 07:34:35 +03:00
|
|
|
index := hash & m.even_index
|
2020-05-09 13:42:01 +03:00
|
|
|
meta := ((hash >> m.shift) & hash_mask) | probe_inc
|
2020-12-06 02:24:24 +03:00
|
|
|
return u32(index), u32(meta)
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
[inline]
|
2020-12-06 02:24:24 +03:00
|
|
|
fn (m &map) meta_less(_index u32, _metas u32) (u32, u32) {
|
2020-04-05 23:09:52 +03:00
|
|
|
mut index := _index
|
|
|
|
mut meta := _metas
|
2020-12-23 21:13:42 +03:00
|
|
|
for meta < unsafe { m.metas[index] } {
|
2020-03-19 08:52:34 +03:00
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
2020-02-20 22:04:06 +03:00
|
|
|
}
|
2020-12-06 02:24:24 +03:00
|
|
|
return index, meta
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
[inline]
|
2020-05-17 14:51:18 +03:00
|
|
|
fn (mut m map) meta_greater(_index u32, _metas u32, kvi u32) {
|
2020-04-05 23:09:52 +03:00
|
|
|
mut meta := _metas
|
|
|
|
mut index := _index
|
2020-03-19 08:52:34 +03:00
|
|
|
mut kv_index := kvi
|
2020-12-23 21:13:42 +03:00
|
|
|
for unsafe { m.metas[index] } != 0 {
|
|
|
|
if meta > unsafe { m.metas[index] } {
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
|
|
|
tmp_meta := m.metas[index]
|
|
|
|
m.metas[index] = meta
|
|
|
|
meta = tmp_meta
|
2020-12-06 02:24:24 +03:00
|
|
|
tmp_index := m.metas[index + 1]
|
2020-07-15 22:56:50 +03:00
|
|
|
m.metas[index + 1] = kv_index
|
2020-12-06 02:24:24 +03:00
|
|
|
kv_index = tmp_index
|
2020-07-15 22:56:50 +03:00
|
|
|
}
|
2020-01-24 22:13:59 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
2020-02-20 22:04:06 +03:00
|
|
|
}
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
|
|
|
m.metas[index] = meta
|
|
|
|
m.metas[index + 1] = kv_index
|
|
|
|
}
|
2020-05-09 13:42:01 +03:00
|
|
|
probe_count := (meta >> hashbits) - 1
|
2020-05-26 18:59:52 +03:00
|
|
|
m.ensure_extra_metas(probe_count)
|
|
|
|
}
|
|
|
|
|
|
|
|
[inline]
|
|
|
|
fn (mut m map) ensure_extra_metas(probe_count u32) {
|
2020-05-09 13:42:01 +03:00
|
|
|
if (probe_count << 1) == m.extra_metas {
|
2021-03-14 15:54:35 +03:00
|
|
|
size_of_u32 := sizeof(u32)
|
|
|
|
old_mem_size := (m.even_index + 2 + m.extra_metas)
|
2020-03-19 08:52:34 +03:00
|
|
|
m.extra_metas += extra_metas_inc
|
2020-12-21 07:34:35 +03:00
|
|
|
mem_size := (m.even_index + 2 + m.extra_metas)
|
2020-07-03 19:10:10 +03:00
|
|
|
unsafe {
|
2021-03-14 15:54:35 +03:00
|
|
|
x := realloc_data(byteptr(m.metas), int(size_of_u32 * old_mem_size), int(size_of_u32 * mem_size))
|
2020-07-25 19:20:47 +03:00
|
|
|
m.metas = &u32(x)
|
2020-12-02 16:35:11 +03:00
|
|
|
C.memset(m.metas + mem_size - extra_metas_inc, 0, int(sizeof(u32) * extra_metas_inc))
|
2020-07-03 19:10:10 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
// Should almost never happen
|
|
|
|
if probe_count == 252 {
|
2020-03-21 15:55:07 +03:00
|
|
|
panic('Probe overflow')
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
2020-03-19 09:05:20 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
}
|
|
|
|
|
2020-06-27 14:58:07 +03:00
|
|
|
// Insert new element to the map. The element is inserted if its key is
|
2020-06-25 00:31:19 +03:00
|
|
|
// not equivalent to the key of any other element already in the container.
|
|
|
|
// If the key already exists, its value is changed to the value of the new element.
|
2020-12-18 09:37:59 +03:00
|
|
|
fn (mut m map) set_1(key voidptr, value voidptr) {
|
2020-12-21 07:34:35 +03:00
|
|
|
load_factor := f32(m.len << 1) / f32(m.even_index)
|
2020-03-21 15:55:07 +03:00
|
|
|
if load_factor > max_load_factor {
|
2020-03-19 08:52:34 +03:00
|
|
|
m.expand()
|
2020-02-20 22:30:34 +03:00
|
|
|
}
|
2020-12-18 09:37:59 +03:00
|
|
|
mut index, mut meta := m.key_to_index(key)
|
2020-12-06 02:24:24 +03:00
|
|
|
index, meta = m.meta_less(index, meta)
|
2020-03-19 08:52:34 +03:00
|
|
|
// While we might have a match
|
2020-12-23 21:13:42 +03:00
|
|
|
for meta == unsafe { m.metas[index] } {
|
|
|
|
kv_index := int(unsafe { m.metas[index + 1] })
|
|
|
|
pkey := unsafe { m.key_values.key(kv_index) }
|
2020-12-27 16:18:46 +03:00
|
|
|
if m.key_eq_fn(key, pkey) {
|
2020-07-03 19:10:10 +03:00
|
|
|
unsafe {
|
2021-03-27 20:59:51 +03:00
|
|
|
pval := byteptr(pkey) + m.key_bytes_
|
2020-12-06 00:53:50 +03:00
|
|
|
C.memcpy(pval, value, m.value_bytes)
|
2020-07-03 19:10:10 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
return
|
2020-02-20 19:58:13 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
2019-12-30 22:01:24 +03:00
|
|
|
}
|
2020-12-27 16:18:46 +03:00
|
|
|
kv_index := m.key_values.expand()
|
|
|
|
unsafe {
|
|
|
|
pkey := m.key_values.key(kv_index)
|
|
|
|
m.clone_fn(pkey, key)
|
2021-03-27 20:59:51 +03:00
|
|
|
C.memcpy(byteptr(pkey) + m.key_bytes_, value, m.value_bytes)
|
2020-12-27 16:18:46 +03:00
|
|
|
}
|
2020-12-06 00:53:50 +03:00
|
|
|
m.meta_greater(index, meta, u32(kv_index))
|
2020-06-21 17:51:02 +03:00
|
|
|
m.len++
|
2020-02-20 19:58:13 +03:00
|
|
|
}
|
|
|
|
|
2020-03-19 08:52:34 +03:00
|
|
|
// Doubles the size of the hashmap
|
2020-05-17 14:51:18 +03:00
|
|
|
fn (mut m map) expand() {
|
2020-12-21 07:34:35 +03:00
|
|
|
old_cap := m.even_index
|
|
|
|
m.even_index = ((m.even_index + 2) << 1) - 2
|
2020-03-19 08:52:34 +03:00
|
|
|
// Check if any hashbits are left
|
2020-04-05 23:09:52 +03:00
|
|
|
if m.cached_hashbits == 0 {
|
|
|
|
m.shift += max_cached_hashbits
|
|
|
|
m.cached_hashbits = max_cached_hashbits
|
2020-03-21 15:55:07 +03:00
|
|
|
m.rehash()
|
2020-12-06 02:24:24 +03:00
|
|
|
} else {
|
2020-03-19 08:52:34 +03:00
|
|
|
m.cached_rehash(old_cap)
|
2020-05-04 15:56:19 +03:00
|
|
|
m.cached_hashbits--
|
2019-08-29 01:52:32 +03:00
|
|
|
}
|
2020-02-20 19:58:13 +03:00
|
|
|
}
|
|
|
|
|
2020-06-27 14:58:07 +03:00
|
|
|
// A rehash is the reconstruction of the hash table:
|
|
|
|
// All the elements in the container are rearranged according
|
|
|
|
// to their hash value into the newly sized key-value container.
|
2020-06-25 00:31:19 +03:00
|
|
|
// Rehashes are performed when the load_factor is going to surpass
|
|
|
|
// the max_load_factor in an operation.
|
2020-05-17 14:51:18 +03:00
|
|
|
fn (mut m map) rehash() {
|
2020-12-21 07:34:35 +03:00
|
|
|
meta_bytes := sizeof(u32) * (m.even_index + 2 + m.extra_metas)
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
2021-03-14 15:54:35 +03:00
|
|
|
// TODO: use realloc_data here too
|
2020-12-02 16:35:11 +03:00
|
|
|
x := v_realloc(byteptr(m.metas), int(meta_bytes))
|
2020-07-25 19:20:47 +03:00
|
|
|
m.metas = &u32(x)
|
2020-07-15 22:56:50 +03:00
|
|
|
C.memset(m.metas, 0, meta_bytes)
|
|
|
|
}
|
2020-12-06 00:53:50 +03:00
|
|
|
for i := 0; i < m.key_values.len; i++ {
|
|
|
|
if !m.key_values.has_index(i) {
|
2020-03-19 08:52:34 +03:00
|
|
|
continue
|
2020-03-16 15:01:31 +03:00
|
|
|
}
|
2020-12-23 21:13:42 +03:00
|
|
|
pkey := unsafe { m.key_values.key(i) }
|
2020-12-14 15:57:53 +03:00
|
|
|
mut index, mut meta := m.key_to_index(pkey)
|
2020-12-06 02:24:24 +03:00
|
|
|
index, meta = m.meta_less(index, meta)
|
2020-12-06 00:53:50 +03:00
|
|
|
m.meta_greater(index, meta, u32(i))
|
2020-03-16 15:01:31 +03:00
|
|
|
}
|
2019-08-29 01:52:32 +03:00
|
|
|
}
|
|
|
|
|
2020-06-25 00:31:19 +03:00
|
|
|
// This method works like rehash. However, instead of rehashing the
|
2020-06-27 14:58:07 +03:00
|
|
|
// key completely, it uses the bits cached in `metas`.
|
2020-05-17 14:51:18 +03:00
|
|
|
fn (mut m map) cached_rehash(old_cap u32) {
|
2020-04-05 23:09:52 +03:00
|
|
|
old_metas := m.metas
|
2020-12-21 07:34:35 +03:00
|
|
|
metasize := int(sizeof(u32) * (m.even_index + 2 + m.extra_metas))
|
2020-09-14 16:20:54 +03:00
|
|
|
m.metas = &u32(vcalloc(metasize))
|
2020-03-19 08:52:34 +03:00
|
|
|
old_extra_metas := m.extra_metas
|
2020-03-21 15:55:07 +03:00
|
|
|
for i := u32(0); i <= old_cap + old_extra_metas; i += 2 {
|
2020-12-23 21:13:42 +03:00
|
|
|
if unsafe { old_metas[i] } == 0 {
|
2020-03-19 08:52:34 +03:00
|
|
|
continue
|
2020-03-16 15:01:31 +03:00
|
|
|
}
|
2020-12-23 21:13:42 +03:00
|
|
|
old_meta := unsafe { old_metas[i] }
|
2020-05-09 13:42:01 +03:00
|
|
|
old_probe_count := ((old_meta >> hashbits) - 1) << 1
|
2020-12-21 07:34:35 +03:00
|
|
|
old_index := (i - old_probe_count) & (m.even_index >> 1)
|
|
|
|
mut index := (old_index | (old_meta << m.shift)) & m.even_index
|
2020-03-19 08:52:34 +03:00
|
|
|
mut meta := (old_meta & hash_mask) | probe_inc
|
2020-12-06 02:24:24 +03:00
|
|
|
index, meta = m.meta_less(index, meta)
|
2020-12-23 21:13:42 +03:00
|
|
|
kv_index := unsafe { old_metas[i + 1] }
|
2020-04-05 23:09:52 +03:00
|
|
|
m.meta_greater(index, meta, kv_index)
|
2019-10-30 20:19:59 +03:00
|
|
|
}
|
2020-12-23 21:13:42 +03:00
|
|
|
unsafe { free(old_metas) }
|
2019-12-30 22:01:24 +03:00
|
|
|
}
|
|
|
|
|
2020-06-25 00:31:19 +03:00
|
|
|
// This method is used for assignment operators. If the argument-key
|
2020-12-06 00:53:50 +03:00
|
|
|
// does not exist in the map, it's added to the map along with the zero/default value.
|
2020-06-27 14:58:07 +03:00
|
|
|
// If the key exists, its respective value is returned.
|
2020-12-18 09:37:59 +03:00
|
|
|
fn (mut m map) get_and_set_1(key voidptr, zero voidptr) voidptr {
|
2020-06-24 21:41:26 +03:00
|
|
|
for {
|
2020-12-18 09:37:59 +03:00
|
|
|
mut index, mut meta := m.key_to_index(key)
|
2020-06-24 21:41:26 +03:00
|
|
|
for {
|
2020-12-23 21:13:42 +03:00
|
|
|
if meta == unsafe { m.metas[index] } {
|
|
|
|
kv_index := int(unsafe { m.metas[index + 1] })
|
|
|
|
pkey := unsafe { m.key_values.key(kv_index) }
|
2020-12-27 16:18:46 +03:00
|
|
|
if m.key_eq_fn(key, pkey) {
|
2021-03-27 20:59:51 +03:00
|
|
|
return unsafe { byteptr(pkey) + m.key_values.key_bytes_ }
|
2020-06-24 21:41:26 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
2020-12-23 21:13:42 +03:00
|
|
|
if meta > unsafe { m.metas[index] } {
|
2020-12-06 02:24:24 +03:00
|
|
|
break
|
|
|
|
}
|
2020-06-24 21:41:26 +03:00
|
|
|
}
|
2020-06-27 14:58:07 +03:00
|
|
|
// Key not found, insert key with zero-value
|
2020-12-18 09:37:59 +03:00
|
|
|
m.set_1(key, zero)
|
2020-06-24 21:41:26 +03:00
|
|
|
}
|
2020-12-02 16:35:11 +03:00
|
|
|
assert false
|
|
|
|
return voidptr(0)
|
2020-06-24 21:41:26 +03:00
|
|
|
}
|
2020-06-25 00:31:19 +03:00
|
|
|
|
2020-06-27 14:58:07 +03:00
|
|
|
// If `key` matches the key of an element in the container,
|
2020-06-25 00:31:19 +03:00
|
|
|
// the method returns a reference to its mapped value.
|
|
|
|
// If not, a zero/default value is returned.
|
2020-12-19 02:05:16 +03:00
|
|
|
fn (m &map) get_1(key voidptr, zero voidptr) voidptr {
|
2020-12-18 09:37:59 +03:00
|
|
|
mut index, mut meta := m.key_to_index(key)
|
2020-06-25 00:31:19 +03:00
|
|
|
for {
|
2020-12-23 21:13:42 +03:00
|
|
|
if meta == unsafe { m.metas[index] } {
|
|
|
|
kv_index := int(unsafe { m.metas[index + 1] })
|
|
|
|
pkey := unsafe { m.key_values.key(kv_index) }
|
2020-12-27 16:18:46 +03:00
|
|
|
if m.key_eq_fn(key, pkey) {
|
2021-03-27 20:59:51 +03:00
|
|
|
return unsafe { byteptr(pkey) + m.key_values.key_bytes_ }
|
2020-06-25 00:31:19 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
2020-12-23 21:13:42 +03:00
|
|
|
if meta > unsafe { m.metas[index] } {
|
2020-12-06 02:24:24 +03:00
|
|
|
break
|
|
|
|
}
|
2020-06-25 00:31:19 +03:00
|
|
|
}
|
|
|
|
return zero
|
|
|
|
}
|
2020-06-24 21:41:26 +03:00
|
|
|
|
2021-01-19 08:06:57 +03:00
|
|
|
// If `key` matches the key of an element in the container,
|
|
|
|
// the method returns a reference to its mapped value.
|
|
|
|
// If not, a zero pointer is returned.
|
|
|
|
// This is used in `x := m['key'] or { ... }`
|
|
|
|
fn (m &map) get_1_check(key voidptr) voidptr {
|
|
|
|
mut index, mut meta := m.key_to_index(key)
|
|
|
|
for {
|
|
|
|
if meta == unsafe { m.metas[index] } {
|
|
|
|
kv_index := int(unsafe { m.metas[index + 1] })
|
|
|
|
pkey := unsafe { m.key_values.key(kv_index) }
|
|
|
|
if m.key_eq_fn(key, pkey) {
|
2021-03-27 20:59:51 +03:00
|
|
|
return unsafe { byteptr(pkey) + m.key_values.key_bytes_ }
|
2021-01-19 08:06:57 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
|
|
|
if meta > unsafe { m.metas[index] } {
|
|
|
|
break
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return 0
|
|
|
|
}
|
|
|
|
|
2020-12-18 09:37:59 +03:00
|
|
|
// Checks whether a particular key exists in the map.
|
2020-12-19 02:05:16 +03:00
|
|
|
fn (m &map) exists_1(key voidptr) bool {
|
2020-12-18 09:37:59 +03:00
|
|
|
mut index, mut meta := m.key_to_index(key)
|
2020-05-09 13:42:01 +03:00
|
|
|
for {
|
2020-12-23 21:13:42 +03:00
|
|
|
if meta == unsafe { m.metas[index] } {
|
|
|
|
kv_index := int(unsafe { m.metas[index + 1] })
|
|
|
|
pkey := unsafe { m.key_values.key(kv_index) }
|
2020-12-27 16:18:46 +03:00
|
|
|
if m.key_eq_fn(key, pkey) {
|
2020-12-06 00:53:50 +03:00
|
|
|
return true
|
2020-05-09 13:42:01 +03:00
|
|
|
}
|
2020-02-20 19:58:13 +03:00
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
2020-12-23 21:13:42 +03:00
|
|
|
if meta > unsafe { m.metas[index] } {
|
2020-12-06 02:24:24 +03:00
|
|
|
break
|
|
|
|
}
|
2020-02-20 19:58:13 +03:00
|
|
|
}
|
2020-02-20 22:04:06 +03:00
|
|
|
return false
|
2019-12-30 08:57:56 +03:00
|
|
|
}
|
|
|
|
|
2020-12-21 07:34:35 +03:00
|
|
|
[inline]
|
|
|
|
fn (mut d DenseArray) delete(i int) {
|
|
|
|
if d.deletes == 0 {
|
|
|
|
d.all_deleted = vcalloc(d.cap) // sets to 0
|
|
|
|
}
|
|
|
|
d.deletes++
|
|
|
|
unsafe {
|
|
|
|
d.all_deleted[i] = 1
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-12-27 20:31:50 +03:00
|
|
|
// delete this
|
2020-05-17 14:51:18 +03:00
|
|
|
pub fn (mut m map) delete(key string) {
|
2021-02-14 21:31:42 +03:00
|
|
|
unsafe {
|
|
|
|
m.delete_1(&key)
|
|
|
|
}
|
2020-12-18 09:37:59 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
// Removes the mapping of a particular key from the map.
|
2021-02-14 21:31:42 +03:00
|
|
|
[unsafe]
|
2020-12-18 09:37:59 +03:00
|
|
|
pub fn (mut m map) delete_1(key voidptr) {
|
|
|
|
mut index, mut meta := m.key_to_index(key)
|
2020-12-06 02:24:24 +03:00
|
|
|
index, meta = m.meta_less(index, meta)
|
2020-03-19 08:52:34 +03:00
|
|
|
// Perform backwards shifting
|
2020-12-23 21:13:42 +03:00
|
|
|
for meta == unsafe { m.metas[index] } {
|
|
|
|
kv_index := int(unsafe { m.metas[index + 1] })
|
|
|
|
pkey := unsafe { m.key_values.key(kv_index) }
|
2020-12-27 16:18:46 +03:00
|
|
|
if m.key_eq_fn(key, pkey) {
|
2020-12-23 21:13:42 +03:00
|
|
|
for (unsafe { m.metas[index + 2] } >> hashbits) > 1 {
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
|
|
|
m.metas[index] = m.metas[index + 2] - probe_inc
|
|
|
|
m.metas[index + 1] = m.metas[index + 3]
|
|
|
|
}
|
2020-03-19 08:52:34 +03:00
|
|
|
index += 2
|
|
|
|
}
|
2020-06-21 17:51:02 +03:00
|
|
|
m.len--
|
2020-12-21 07:34:35 +03:00
|
|
|
m.key_values.delete(kv_index)
|
2020-07-15 22:56:50 +03:00
|
|
|
unsafe {
|
2020-12-13 04:05:56 +03:00
|
|
|
m.metas[index] = 0
|
2020-12-27 16:18:46 +03:00
|
|
|
m.free_fn(pkey)
|
2020-12-13 04:05:56 +03:00
|
|
|
// Mark key as deleted
|
2020-12-18 09:37:59 +03:00
|
|
|
C.memset(pkey, 0, m.key_bytes)
|
2020-07-15 22:56:50 +03:00
|
|
|
}
|
2020-06-21 17:51:02 +03:00
|
|
|
if m.key_values.len <= 32 {
|
2020-03-21 15:55:07 +03:00
|
|
|
return
|
|
|
|
}
|
2020-04-05 23:09:52 +03:00
|
|
|
// Clean up key_values if too many have been deleted
|
2020-06-21 17:51:02 +03:00
|
|
|
if m.key_values.deletes >= (m.key_values.len >> 1) {
|
2020-03-19 08:52:34 +03:00
|
|
|
m.key_values.zeros_to_end()
|
|
|
|
m.rehash()
|
|
|
|
}
|
|
|
|
return
|
|
|
|
}
|
|
|
|
index += 2
|
|
|
|
meta += probe_inc
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-12-18 09:37:59 +03:00
|
|
|
// bootstrap
|
2020-12-27 20:31:50 +03:00
|
|
|
// delete this
|
2020-01-24 22:13:59 +03:00
|
|
|
pub fn (m &map) keys() []string {
|
2020-12-06 02:24:24 +03:00
|
|
|
mut keys := []string{len: m.len}
|
2020-12-23 21:13:42 +03:00
|
|
|
mut item := unsafe { byteptr(keys.data) }
|
2020-12-18 09:37:59 +03:00
|
|
|
for i := 0; i < m.key_values.len; i++ {
|
|
|
|
if !m.key_values.has_index(i) {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
unsafe {
|
|
|
|
pkey := m.key_values.key(i)
|
2020-12-27 16:18:46 +03:00
|
|
|
m.clone_fn(item, pkey)
|
2020-12-18 09:37:59 +03:00
|
|
|
item += m.key_bytes
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return keys
|
|
|
|
}
|
|
|
|
|
|
|
|
// Returns all keys in the map.
|
2021-01-03 17:55:06 +03:00
|
|
|
fn (m &map) keys_1() array {
|
2020-12-18 09:37:59 +03:00
|
|
|
mut keys := __new_array(m.len, 0, m.key_bytes)
|
2020-12-23 21:13:42 +03:00
|
|
|
mut item := unsafe { byteptr(keys.data) }
|
2020-12-06 02:24:24 +03:00
|
|
|
if m.key_values.deletes == 0 {
|
|
|
|
for i := 0; i < m.key_values.len; i++ {
|
2020-12-16 04:23:59 +03:00
|
|
|
unsafe {
|
|
|
|
pkey := m.key_values.key(i)
|
2020-12-27 16:18:46 +03:00
|
|
|
m.clone_fn(item, pkey)
|
2020-12-16 04:23:59 +03:00
|
|
|
item += m.key_bytes
|
|
|
|
}
|
2020-12-06 02:24:24 +03:00
|
|
|
}
|
|
|
|
return keys
|
|
|
|
}
|
2020-12-06 00:53:50 +03:00
|
|
|
for i := 0; i < m.key_values.len; i++ {
|
|
|
|
if !m.key_values.has_index(i) {
|
2020-03-19 08:52:34 +03:00
|
|
|
continue
|
2020-02-20 22:04:06 +03:00
|
|
|
}
|
2020-12-16 04:23:59 +03:00
|
|
|
unsafe {
|
|
|
|
pkey := m.key_values.key(i)
|
2020-12-27 16:18:46 +03:00
|
|
|
m.clone_fn(item, pkey)
|
2020-12-16 04:23:59 +03:00
|
|
|
item += m.key_bytes
|
|
|
|
}
|
2020-02-20 22:04:06 +03:00
|
|
|
}
|
2020-01-24 22:13:59 +03:00
|
|
|
return keys
|
2019-12-30 22:01:24 +03:00
|
|
|
}
|
|
|
|
|
2020-12-27 16:18:46 +03:00
|
|
|
// warning: only copies keys, does not clone
|
2020-08-09 12:22:11 +03:00
|
|
|
[unsafe]
|
2021-01-03 17:55:06 +03:00
|
|
|
fn (d &DenseArray) clone() DenseArray {
|
2020-12-06 02:24:24 +03:00
|
|
|
res := DenseArray{
|
|
|
|
key_bytes: d.key_bytes
|
2021-03-27 20:59:51 +03:00
|
|
|
key_bytes_: d.key_bytes_
|
2020-05-26 02:52:06 +03:00
|
|
|
value_bytes: d.value_bytes
|
2020-12-06 02:24:24 +03:00
|
|
|
slot_bytes: d.slot_bytes
|
|
|
|
cap: d.cap
|
|
|
|
len: d.len
|
|
|
|
deletes: d.deletes
|
2020-12-13 04:05:56 +03:00
|
|
|
all_deleted: 0
|
|
|
|
data: 0
|
|
|
|
}
|
|
|
|
unsafe {
|
|
|
|
if d.deletes != 0 {
|
|
|
|
res.all_deleted = memdup(d.all_deleted, d.cap)
|
|
|
|
}
|
|
|
|
res.data = memdup(d.data, d.cap * d.slot_bytes)
|
2020-05-26 02:52:06 +03:00
|
|
|
}
|
|
|
|
return res
|
|
|
|
}
|
|
|
|
|
2021-02-06 05:07:42 +03:00
|
|
|
// clone returns a clone of the `map`.
|
2020-08-09 12:22:11 +03:00
|
|
|
[unsafe]
|
2020-12-17 10:44:50 +03:00
|
|
|
pub fn (m &map) clone() map {
|
2020-12-21 07:34:35 +03:00
|
|
|
metasize := int(sizeof(u32) * (m.even_index + 2 + m.extra_metas))
|
2020-05-26 02:52:06 +03:00
|
|
|
res := map{
|
2020-12-14 15:57:53 +03:00
|
|
|
key_bytes: m.key_bytes
|
2021-03-27 20:59:51 +03:00
|
|
|
key_bytes_: m.key_bytes_
|
2020-12-06 02:24:24 +03:00
|
|
|
value_bytes: m.value_bytes
|
2020-12-21 07:34:35 +03:00
|
|
|
even_index: m.even_index
|
2020-05-26 02:52:06 +03:00
|
|
|
cached_hashbits: m.cached_hashbits
|
2020-12-06 02:24:24 +03:00
|
|
|
shift: m.shift
|
2020-12-23 21:13:42 +03:00
|
|
|
key_values: unsafe { m.key_values.clone() }
|
2021-02-14 21:31:42 +03:00
|
|
|
metas: unsafe { &u32(malloc(metasize)) }
|
2020-12-06 02:24:24 +03:00
|
|
|
extra_metas: m.extra_metas
|
|
|
|
len: m.len
|
2020-12-27 16:18:46 +03:00
|
|
|
has_string_keys: m.has_string_keys
|
|
|
|
hash_fn: m.hash_fn
|
|
|
|
key_eq_fn: m.key_eq_fn
|
|
|
|
clone_fn: m.clone_fn
|
|
|
|
free_fn: m.free_fn
|
2020-07-15 22:56:50 +03:00
|
|
|
}
|
2020-12-23 21:13:42 +03:00
|
|
|
unsafe { C.memcpy(res.metas, m.metas, metasize) }
|
2020-12-27 16:18:46 +03:00
|
|
|
if !m.has_string_keys {
|
|
|
|
return res
|
|
|
|
}
|
|
|
|
// clone keys
|
|
|
|
for i in 0 .. m.key_values.len {
|
|
|
|
if !m.key_values.has_index(i) {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
m.clone_fn(res.key_values.key(i), m.key_values.key(i))
|
|
|
|
}
|
2020-05-26 02:52:06 +03:00
|
|
|
return res
|
|
|
|
}
|
|
|
|
|
2021-02-06 05:07:42 +03:00
|
|
|
// free releases all memory resources occupied by the `map`.
|
2020-08-09 12:22:11 +03:00
|
|
|
[unsafe]
|
2020-05-06 19:03:44 +03:00
|
|
|
pub fn (m &map) free() {
|
2020-12-23 21:13:42 +03:00
|
|
|
unsafe { free(m.metas) }
|
2020-12-06 02:24:24 +03:00
|
|
|
if m.key_values.deletes == 0 {
|
|
|
|
for i := 0; i < m.key_values.len; i++ {
|
|
|
|
unsafe {
|
2020-12-18 09:37:59 +03:00
|
|
|
pkey := m.key_values.key(i)
|
2020-12-27 16:18:46 +03:00
|
|
|
m.free_fn(pkey)
|
2020-12-06 02:24:24 +03:00
|
|
|
}
|
2020-04-06 00:31:53 +03:00
|
|
|
}
|
2020-12-06 02:24:24 +03:00
|
|
|
} else {
|
|
|
|
for i := 0; i < m.key_values.len; i++ {
|
|
|
|
if !m.key_values.has_index(i) {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
unsafe {
|
2020-12-18 09:37:59 +03:00
|
|
|
pkey := m.key_values.key(i)
|
2020-12-27 16:18:46 +03:00
|
|
|
m.free_fn(pkey)
|
2020-12-06 02:24:24 +03:00
|
|
|
}
|
2020-07-15 22:56:50 +03:00
|
|
|
}
|
2020-12-23 21:13:42 +03:00
|
|
|
unsafe { free(m.key_values.all_deleted) }
|
2020-07-15 22:56:50 +03:00
|
|
|
}
|
2020-12-23 21:13:42 +03:00
|
|
|
unsafe { free(m.key_values.data) }
|
2019-07-23 23:57:06 +03:00
|
|
|
}
|