v/vlib/builtin/array.v

645 lines
17 KiB
V
Raw Normal View History

// Copyright (c) 2019-2021 Alexander Medvednikov. All rights reserved.
2019-06-23 04:21:30 +02:00
// Use of this source code is governed by an MIT license
// that can be found in the LICENSE file.
2019-06-22 20:20:28 +02:00
module builtin
import strings
2019-08-10 23:02:48 +02:00
// array is a struct used for denoting array types in V
2020-01-12 19:59:57 +01:00
pub struct array {
pub:
element_size int // size in bytes of one element in the array.
pub mut:
data voidptr
offset int // in bytes (should be `size_t`)
len int // length of the array.
cap int // capacity of the array.
2019-06-22 20:20:28 +02:00
}
// array.data uses a void pointer, which allows implementing arrays without generics and without generating
// extra code for every type.
// Internal function, used by V (`nums := []int`)
2020-04-25 08:42:23 +02:00
fn __new_array(mylen int, cap int, elm_size int) array {
cap_ := if cap < mylen { mylen } else { cap }
2019-12-18 18:07:32 +01:00
arr := array{
element_size: elm_size
data: vcalloc(cap_ * elm_size)
2020-05-18 21:38:06 +02:00
len: mylen
cap: cap_
2019-06-22 20:20:28 +02:00
}
return arr
}
2020-04-15 20:12:06 +02:00
2020-05-16 15:21:37 +02:00
fn __new_array_with_default(mylen int, cap int, elm_size int, val voidptr) array {
cap_ := if cap < mylen { mylen } else { cap }
mut arr := array{
element_size: elm_size
data: vcalloc(cap_ * elm_size)
2020-05-18 21:38:06 +02:00
len: mylen
cap: cap_
2020-05-16 15:21:37 +02:00
}
if val != 0 {
for i in 0 .. arr.len {
unsafe { arr.set_unsafe(i, val) }
2020-05-16 15:21:37 +02:00
}
}
return arr
}
2020-06-12 01:24:25 +02:00
fn __new_array_with_array_default(mylen int, cap int, elm_size int, val array) array {
cap_ := if cap < mylen { mylen } else { cap }
mut arr := array{
2020-06-12 01:24:25 +02:00
element_size: elm_size
data: vcalloc(cap_ * elm_size)
len: mylen
cap: cap_
}
for i in 0 .. arr.len {
val_clone := unsafe { val.clone_to_depth(1) }
unsafe { arr.set_unsafe(i, &val_clone) }
2020-06-12 01:24:25 +02:00
}
return arr
}
2019-06-22 20:20:28 +02:00
// Private function, used by V (`nums := [1, 2, 3]`)
fn new_array_from_c_array(len int, cap int, elm_size int, c_array voidptr) array {
cap_ := if cap < len { len } else { cap }
2019-12-18 18:07:32 +01:00
arr := array{
element_size: elm_size
data: vcalloc(cap_ * elm_size)
2020-05-18 21:38:06 +02:00
len: len
cap: cap_
2019-06-22 20:20:28 +02:00
}
// TODO Write all memory functions (like memcpy) in V
unsafe { C.memcpy(arr.data, c_array, len * elm_size) }
2019-06-22 20:20:28 +02:00
return arr
}
// Private function, used by V (`nums := [1, 2, 3] !`)
fn new_array_from_c_array_no_alloc(len int, cap int, elm_size int, c_array voidptr) array {
2019-12-18 18:07:32 +01:00
arr := array{
element_size: elm_size
data: c_array
2020-05-18 21:38:06 +02:00
len: len
cap: cap
2019-06-22 20:20:28 +02:00
}
return arr
}
// Private function. Doubles array capacity if needed.
2020-05-17 13:51:18 +02:00
fn (mut a array) ensure_cap(required int) {
2020-01-19 13:11:58 +01:00
if required <= a.cap {
return
}
mut cap := if a.cap > 0 { a.cap } else { 2 }
2020-01-19 13:11:58 +01:00
for required > cap {
cap *= 2
}
new_size := cap * a.element_size
new_data := vcalloc(new_size)
if a.data != voidptr(0) {
unsafe { C.memcpy(new_data, a.data, a.len * a.element_size) }
// TODO: the old data may be leaked when no GC is used (ref-counting?)
2020-01-19 13:11:58 +01:00
}
a.data = new_data
a.offset = 0
2020-01-19 13:11:58 +01:00
a.cap = cap
}
// repeat returns a new array with the given array elements repeated given times.
// `cgen` will replace this with an apropriate call to `repeat_to_depth()`
// This is a dummy placeholder that will be overridden by `cgen` with an appropriate
// call to `repeat_to_depth()`. However the `checker` needs it here.
2020-02-29 15:25:49 +01:00
pub fn (a array) repeat(count int) array {
return unsafe { a.repeat_to_depth(count, 0) }
}
// version of `repeat()` that handles multi dimensional arrays
// `unsafe` to call directly because `depth` is not checked
[unsafe]
pub fn (a array) repeat_to_depth(count int, depth int) array {
2020-02-29 15:25:49 +01:00
if count < 0 {
panic('array.repeat: count is negative: $count')
}
2020-02-29 15:25:49 +01:00
mut size := count * a.len * a.element_size
2019-12-16 20:22:04 +01:00
if size == 0 {
size = a.element_size
}
2019-12-18 18:07:32 +01:00
arr := array{
element_size: a.element_size
data: vcalloc(size)
2020-05-18 21:38:06 +02:00
len: count * a.len
cap: count * a.len
2019-09-14 22:48:30 +02:00
}
if a.len > 0 {
for i in 0 .. count {
if depth > 0 {
ary_clone := unsafe { a.clone_to_depth(depth) }
unsafe { C.memcpy(arr.get_unsafe(i * a.len), &byte(ary_clone.data), a.len * a.element_size) }
} else {
unsafe { C.memcpy(arr.get_unsafe(i * a.len), &byte(a.data), a.len * a.element_size) }
}
}
2019-09-14 22:48:30 +02:00
}
return arr
}
// sort_with_compare sorts array in-place using given `compare` function as comparator.
2020-05-17 13:51:18 +02:00
pub fn (mut a array) sort_with_compare(compare voidptr) {
$if freestanding {
panic('sort does not work with -freestanding')
} $else {
C.qsort(mut a.data, a.len, a.element_size, compare)
}
2019-06-22 20:20:28 +02:00
}
// insert inserts a value in the array at index `i`
2020-05-17 13:51:18 +02:00
pub fn (mut a array) insert(i int, val voidptr) {
$if !no_bounds_checking ? {
if i < 0 || i > a.len {
panic('array.insert: index out of range (i == $i, a.len == $a.len)')
}
2019-06-22 20:20:28 +02:00
}
a.ensure_cap(a.len + 1)
unsafe {
C.memmove(a.get_unsafe(i + 1), a.get_unsafe(i), (a.len - i) * a.element_size)
a.set_unsafe(i, val)
}
a.len++
2019-06-22 20:20:28 +02:00
}
// insert_many inserts many values into the array from index `i`.
[unsafe]
pub fn (mut a array) insert_many(i int, val voidptr, size int) {
$if !no_bounds_checking ? {
if i < 0 || i > a.len {
panic('array.insert_many: index out of range (i == $i, a.len == $a.len)')
}
}
a.ensure_cap(a.len + size)
elem_size := a.element_size
unsafe {
iptr := a.get_unsafe(i)
C.memmove(a.get_unsafe(i + size), iptr, (a.len - i) * elem_size)
C.memcpy(iptr, val, size * elem_size)
}
a.len += size
}
// prepend prepends one value to the array.
2020-05-17 13:51:18 +02:00
pub fn (mut a array) prepend(val voidptr) {
2019-06-22 20:20:28 +02:00
a.insert(0, val)
}
// prepend_many prepends another array to this array.
[unsafe]
pub fn (mut a array) prepend_many(val voidptr, size int) {
unsafe { a.insert_many(0, val, size) }
}
// delete deletes array element at index `i`.
2020-05-17 13:51:18 +02:00
pub fn (mut a array) delete(i int) {
a.delete_many(i, 1)
}
// delete_many deletes `size` elements beginning with index `i`
pub fn (mut a array) delete_many(i int, size int) {
$if !no_bounds_checking ? {
if i < 0 || i + size > a.len {
endidx := if size > 1 { '..${i + size}' } else { '' }
panic('array.delete: index out of range (i == $i$endidx, a.len == $a.len)')
}
}
2020-06-12 01:24:25 +02:00
// NB: if a is [12,34], a.len = 2, a.delete(0)
2020-05-24 10:10:41 +02:00
// should move (2-0-1) elements = 1 element (the 34) forward
old_data := a.data
new_size := a.len - size
new_cap := if new_size == 0 { 1 } else { new_size }
a.data = vcalloc(new_cap * a.element_size)
unsafe { C.memcpy(a.data, old_data, i * a.element_size) }
unsafe {
C.memcpy(&byte(a.data) + i * a.element_size, &byte(old_data) + (i + size) * a.element_size,
(a.len - i - size) * a.element_size)
}
a.len = new_size
a.cap = new_cap
2019-06-22 20:20:28 +02:00
}
// clear clears the array without deallocating the allocated data.
2020-05-17 13:51:18 +02:00
pub fn (mut a array) clear() {
2020-01-27 22:31:48 +01:00
a.len = 0
}
// trim trims the array length to "index" without modifying the allocated data. If "index" is greater
// than len nothing will be changed.
2020-05-17 13:51:18 +02:00
pub fn (mut a array) trim(index int) {
2020-02-17 20:31:40 +01:00
if index < a.len {
a.len = index
}
}
// we manually inline this for single operations for performance without -prod
[inline; unsafe]
fn (a array) get_unsafe(i int) voidptr {
unsafe {
2021-04-05 20:39:32 +02:00
return &byte(a.data) + i * a.element_size
}
}
// Private function. Used to implement array[] operator.
fn (a array) get(i int) voidptr {
$if !no_bounds_checking ? {
if i < 0 || i >= a.len {
panic('array.get: index out of range (i == $i, a.len == $a.len)')
}
}
unsafe {
2021-04-05 20:39:32 +02:00
return &byte(a.data) + i * a.element_size
}
}
// Private function. Used to implement x = a[i] or { ... }
fn (a array) get_with_check(i int) voidptr {
if i < 0 || i >= a.len {
return 0
}
unsafe {
2021-04-05 20:39:32 +02:00
return &byte(a.data) + i * a.element_size
}
}
// first returns the first element of the array.
pub fn (a array) first() voidptr {
$if !no_bounds_checking ? {
if a.len == 0 {
panic('array.first: array is empty')
}
2019-06-22 20:20:28 +02:00
}
2020-04-02 15:31:44 +02:00
return a.data
2019-06-22 20:20:28 +02:00
}
// last returns the last element of the array.
pub fn (a array) last() voidptr {
$if !no_bounds_checking ? {
if a.len == 0 {
panic('array.last: array is empty')
}
2019-06-22 20:20:28 +02:00
}
unsafe {
2021-04-05 20:39:32 +02:00
return &byte(a.data) + (a.len - 1) * a.element_size
}
2019-06-22 20:20:28 +02:00
}
// pop returns the last element of the array, and removes it.
2020-07-14 18:55:44 +02:00
pub fn (mut a array) pop() voidptr {
// in a sense, this is the opposite of `a << x`
$if !no_bounds_checking ? {
2020-07-14 18:55:44 +02:00
if a.len == 0 {
panic('array.pop: array is empty')
}
}
new_len := a.len - 1
last_elem := unsafe { &byte(a.data) + new_len * a.element_size }
2020-07-14 18:55:44 +02:00
a.len = new_len
// NB: a.cap is not changed here *on purpose*, so that
// further << ops on that array will be more efficient.
return unsafe { memdup(last_elem, a.element_size) }
2020-07-14 18:55:44 +02:00
}
// delete_last efficiently deletes the last element of the array.
pub fn (mut a array) delete_last() {
// copy pasting code for performance
$if !no_bounds_checking ? {
if a.len == 0 {
panic('array.pop: array is empty')
}
}
a.len--
}
// slice returns an array using the same buffer as original array
// but starting from the `start` element and ending with the element before
// the `end` element of the original array with the length and capacity
// set to the number of the elements in the slice.
fn (a array) slice(start int, _end int) array {
2019-12-16 23:29:40 +01:00
mut end := _end
$if !no_bounds_checking ? {
if start > end {
panic('array.slice: invalid slice index ($start > $end)')
}
if end > a.len {
panic('array.slice: slice bounds out of range ($end >= $a.len)')
}
if start < 0 {
panic('array.slice: slice bounds out of range ($start < 0)')
}
2019-12-16 23:29:40 +01:00
}
offset := start * a.element_size
data := unsafe { &byte(a.data) + offset }
2019-12-16 23:29:40 +01:00
l := end - start
2019-12-18 18:07:32 +01:00
res := array{
2019-12-16 23:29:40 +01:00
element_size: a.element_size
data: data
offset: a.offset + offset
2019-12-16 23:29:40 +01:00
len: l
cap: l
}
return res
}
2020-01-08 10:19:12 +01:00
// used internally for [2..4]
fn (a array) slice2(start int, _end int, end_max bool) array {
2020-01-08 10:19:12 +01:00
end := if end_max { a.len } else { _end }
return a.slice(start, end)
}
// `clone_static_to_depth()` returns an independent copy of a given array.
// Unlike `clone_to_depth()` it has a value receiver and is used internally
// for slice-clone expressions like `a[2..4].clone()` and in -autofree generated code.
fn (a array) clone_static_to_depth(depth int) array {
return unsafe { a.clone_to_depth(depth) }
}
// clone returns an independent copy of a given array.
// this will be overwritten by `cgen` with an apropriate call to `.clone_to_depth()`
// However the `checker` needs it here.
2020-03-10 23:21:26 +01:00
pub fn (a &array) clone() array {
return unsafe { a.clone_to_depth(0) }
}
// recursively clone given array - `unsafe` when called directly because depth is not checked
[unsafe]
pub fn (a &array) clone_to_depth(depth int) array {
2020-01-08 10:19:12 +01:00
mut size := a.cap * a.element_size
if size == 0 {
size++
}
mut arr := array{
element_size: a.element_size
data: vcalloc(size)
2020-05-18 21:38:06 +02:00
len: a.len
cap: a.cap
2020-01-08 10:19:12 +01:00
}
2020-06-19 13:32:55 +02:00
// Recursively clone-generated elements if array element is array type
if depth > 0 && a.element_size == sizeof(array) && a.len >= 0 && a.cap >= a.len {
for i in 0 .. a.len {
2020-06-19 13:32:55 +02:00
ar := array{}
unsafe { C.memcpy(&ar, a.get_unsafe(i), int(sizeof(array))) }
ar_clone := unsafe { ar.clone_to_depth(depth - 1) }
unsafe { arr.set_unsafe(i, &ar_clone) }
2020-06-19 13:32:55 +02:00
}
return arr
} else {
if !isnil(a.data) {
unsafe { C.memcpy(&byte(arr.data), a.data, a.cap * a.element_size) }
}
return arr
2020-06-19 13:32:55 +02:00
}
2019-06-22 20:20:28 +02:00
}
// we manually inline this for single operations for performance without -prod
[inline; unsafe]
fn (mut a array) set_unsafe(i int, val voidptr) {
2021-04-05 20:39:32 +02:00
unsafe { C.memcpy(&byte(a.data) + a.element_size * i, val, a.element_size) }
}
// Private function. Used to implement assigment to the array element.
2020-05-17 13:51:18 +02:00
fn (mut a array) set(i int, val voidptr) {
$if !no_bounds_checking ? {
if i < 0 || i >= a.len {
panic('array.set: index out of range (i == $i, a.len == $a.len)')
}
2019-06-22 20:20:28 +02:00
}
2021-04-05 20:39:32 +02:00
unsafe { C.memcpy(&byte(a.data) + a.element_size * i, val, a.element_size) }
2019-06-22 20:20:28 +02:00
}
2020-05-17 13:51:18 +02:00
fn (mut a array) push(val voidptr) {
a.ensure_cap(a.len + 1)
2021-04-05 20:39:32 +02:00
unsafe { C.memmove(&byte(a.data) + a.element_size * a.len, val, a.element_size) }
a.len++
}
// push_many implements the functionality for pushing another array.
// `val` is array.data and user facing usage is `a << [1,2,3]`
[unsafe]
2020-05-17 13:51:18 +02:00
pub fn (mut a3 array) push_many(val voidptr, size int) {
if a3.data == val && !isnil(a3.data) {
2020-03-08 22:11:56 +01:00
// handle `arr << arr`
copy := a3.clone()
a3.ensure_cap(a3.len + size)
unsafe {
// C.memcpy(a.data, copy.data, copy.element_size * copy.len)
C.memcpy(a3.get_unsafe(a3.len), copy.data, a3.element_size * size)
}
2020-01-19 13:11:58 +01:00
} else {
2020-03-08 22:11:56 +01:00
a3.ensure_cap(a3.len + size)
if !isnil(a3.data) && !isnil(val) {
unsafe { C.memcpy(a3.get_unsafe(a3.len), val, a3.element_size * size) }
}
2020-01-19 13:11:58 +01:00
}
2020-03-08 22:11:56 +01:00
a3.len += size
}
// reverse_in_place reverses existing array data, modifying original array.
pub fn (mut a array) reverse_in_place() {
if a.len < 2 {
return
}
unsafe {
mut tmp_value := malloc(a.element_size)
for i in 0 .. a.len / 2 {
2021-04-05 20:39:32 +02:00
C.memcpy(tmp_value, &byte(a.data) + i * a.element_size, a.element_size)
C.memcpy(&byte(a.data) + i * a.element_size, &byte(a.data) +
(a.len - 1 - i) * a.element_size, a.element_size)
2021-04-05 20:39:32 +02:00
C.memcpy(&byte(a.data) + (a.len - 1 - i) * a.element_size, tmp_value, a.element_size)
}
free(tmp_value)
}
}
// reverse returns a new array with the elements of the original array in reverse order.
2019-07-17 18:17:07 +02:00
pub fn (a array) reverse() array {
2020-02-29 20:44:02 +01:00
if a.len < 2 {
return a
}
mut arr := array{
element_size: a.element_size
data: vcalloc(a.cap * a.element_size)
2020-05-18 21:38:06 +02:00
len: a.len
cap: a.cap
2019-07-17 18:17:07 +02:00
}
for i in 0 .. a.len {
unsafe { arr.set_unsafe(i, a.get_unsafe(a.len - 1 - i)) }
2019-07-17 18:17:07 +02:00
}
return arr
}
2019-12-18 18:07:32 +01:00
// pub fn (a []int) free() {
// free frees all memory occupied by the array.
2020-08-09 11:22:11 +02:00
[unsafe]
pub fn (a &array) free() {
2020-07-11 13:22:16 +02:00
$if prealloc {
return
}
2019-12-18 18:07:32 +01:00
// if a.is_slice {
// return
// }
unsafe { free(&byte(a.data) - a.offset) }
2019-06-22 20:20:28 +02:00
}
// filter creates a new array with all elements that pass the test implemented by the provided function
pub fn (a array) filter(predicate fn (voidptr) bool) array
struct ZZZTmp1 {}
// any tests whether at least one element in the array passes the test implemented by the
// provided function. It returns true if, in the array, it finds an element for which the provided
// function returns true; otherwise it returns false. It doesn't modify the array
pub fn (a array) any(predicate fn (voidptr) bool) bool
struct ZZZTmp2 {}
// all tests whether all elements in the array pass the test implemented by the provided function
pub fn (a array) all(predicate fn (voidptr) bool) bool
struct ZZZTmp3 {}
// map creates a new array populated with the results of calling a provided function
// on every element in the calling array
pub fn (a array) map(callback fn (voidptr) voidptr) array
struct ZZZTmp4 {}
// sort sorts an array in place in ascending order.
pub fn (mut a array) sort(callback fn (voidptr, voidptr) int)
struct ZZZTmp5 {}
// contains determines whether an array includes a certain value among its entries
pub fn (a array) contains(val voidptr) bool
struct ZZZTmp6 {}
// index returns the first index at which a given element can be found in the array
// or -1 if the value is not found.
pub fn (a array) index(value voidptr) int
[unsafe]
pub fn (mut a []string) free() {
$if prealloc {
return
}
for s in a {
unsafe { s.free() }
}
unsafe { free(a.data) }
}
// str returns a string representation of the array of strings
// => '["a", "b", "c"]'.
2021-03-23 21:11:32 +01:00
[manualfree]
2019-06-30 13:06:46 +02:00
pub fn (a []string) str() string {
2019-08-31 01:35:05 +02:00
mut sb := strings.new_builder(a.len * 3)
sb.write_string('[')
for i in 0 .. a.len {
2019-06-22 20:20:28 +02:00
val := a[i]
sb.write_string("'")
sb.write_string(val)
sb.write_string("'")
2019-06-22 20:20:28 +02:00
if i < a.len - 1 {
sb.write_string(', ')
2019-06-22 20:20:28 +02:00
}
}
sb.write_string(']')
2021-03-23 21:11:32 +01:00
res := sb.str()
unsafe { sb.free() }
return res
2019-06-22 20:20:28 +02:00
}
// hex returns a string with the hexadecimal representation
// of the byte elements of the array.
2019-07-15 17:49:01 +02:00
pub fn (b []byte) hex() string {
mut hex := unsafe { malloc(b.len * 2 + 1) }
mut dst_i := 0
for i in b {
2020-03-18 16:47:37 +01:00
n0 := i >> 4
unsafe {
hex[dst_i] = if n0 < 10 { n0 + `0` } else { n0 + byte(87) }
dst_i++
}
n1 := i & 0xF
unsafe {
hex[dst_i] = if n1 < 10 { n1 + `0` } else { n1 + byte(87) }
dst_i++
}
}
unsafe {
hex[dst_i] = 0
return tos(hex, dst_i)
}
2019-07-15 17:49:01 +02:00
}
2019-07-28 17:19:59 +02:00
// copy copies the `src` byte array elements to the `dst` byte array.
// The number of the elements copied is the minimum of the length of both arrays.
// Returns the number of elements copied.
2019-07-28 17:19:59 +02:00
// TODO: implement for all types
pub fn copy(dst []byte, src []byte) int {
2021-03-02 17:14:42 +01:00
min := if dst.len < src.len { dst.len } else { src.len }
if min > 0 {
2021-04-05 20:39:32 +02:00
unsafe { C.memcpy(&byte(dst.data), src.data, min) }
2019-07-28 17:19:59 +02:00
}
2021-03-02 17:14:42 +01:00
return min
2019-07-28 17:19:59 +02:00
}
2019-09-01 19:55:34 +02:00
// reduce executes a given reducer function on each element of the array,
2019-10-11 03:12:40 +02:00
// resulting in a single output value.
pub fn (a []int) reduce(iter fn (int, int) int, accum_start int) int {
mut accum_ := accum_start
for i in a {
accum_ = iter(accum_, i)
2019-10-11 03:12:40 +02:00
}
return accum_
2019-10-11 03:12:40 +02:00
}
// grow_cap grows the array's capacity by `amount` elements.
pub fn (mut a array) grow_cap(amount int) {
a.ensure_cap(a.cap + amount)
}
// grow_len ensures that an array has a.len + amount of length
[unsafe]
pub fn (mut a array) grow_len(amount int) {
a.ensure_cap(a.len + amount)
a.len += amount
}
// pointers returns a new array, where each element
// is the address of the corresponding element in the array.
[unsafe]
pub fn (a array) pointers() []voidptr {
mut res := []voidptr{}
for i in 0 .. a.len {
unsafe { res << a.get_unsafe(i) }
}
return res
}
// voidptr.vbytes() - makes a V []byte structure from a C style memory buffer. NB: the data is reused, NOT copied!
[unsafe]
pub fn (data voidptr) vbytes(len int) []byte {
res := array{
element_size: 1
data: data
len: len
cap: len
}
return res
}
// byteptr.vbytes() - makes a V []byte structure from a C style memory buffer. NB: the data is reused, NOT copied!
[unsafe]
2021-04-04 19:14:51 +02:00
pub fn (data &byte) vbytes(len int) []byte {
return unsafe { voidptr(data).vbytes(len) }
}