zig/std/mem.zig

405 lines
12 KiB
Zig

const debug = @import("debug.zig");
const assert = debug.assert;
const math = @import("math/index.zig");
const os = @import("os/index.zig");
const io = @import("io.zig");
const builtin = @import("builtin");
const Os = builtin.Os;
pub const Cmp = math.Cmp;
error NoMem;
pub const Allocator = struct {
allocFn: fn (self: &Allocator, n: usize) -> %[]u8,
/// Note that old_mem may be a slice of length 0, in which case reallocFn
/// should simply call allocFn.
reallocFn: fn (self: &Allocator, old_mem: []u8, new_size: usize) -> %[]u8,
/// Note that mem may be a slice of length 0, in which case freeFn
/// should do nothing.
freeFn: fn (self: &Allocator, mem: []u8),
/// Aborts the program if an allocation fails.
fn checkedAlloc(self: &Allocator, comptime T: type, n: usize) -> []T {
alloc(self, T, n) %% |err| {
%%io.stderr.printf("allocation failure: {}\n", @errorName(err));
os.abort()
}
}
fn create(self: &Allocator, comptime T: type) -> %&T {
&(%return self.alloc(T, 1))[0]
}
fn destroy(self: &Allocator, ptr: var) {
self.free(ptr[0..1]);
}
fn alloc(self: &Allocator, comptime T: type, n: usize) -> %[]T {
const byte_count = %return math.mul(usize, @sizeOf(T), n);
([]T)(%return self.allocFn(self, byte_count))
}
fn realloc(self: &Allocator, comptime T: type, old_mem: []T, n: usize) -> %[]T {
const byte_count = %return math.mul(usize, @sizeOf(T), n);
([]T)(%return self.reallocFn(self, ([]u8)(old_mem), byte_count))
}
fn free(self: &Allocator, mem: var) {
self.freeFn(self, ([]u8)(mem));
}
};
pub const IncrementingAllocator = struct {
allocator: Allocator,
bytes: []u8,
end_index: usize,
fn init(capacity: usize) -> %IncrementingAllocator {
switch (builtin.os) {
Os.linux, Os.darwin, Os.macosx, Os.ios => {
const p = os.posix;
const addr = p.mmap(null, capacity, p.PROT_READ|p.PROT_WRITE,
p.MAP_PRIVATE|p.MAP_ANONYMOUS|p.MAP_NORESERVE, -1, 0);
if (addr == p.MAP_FAILED) {
return error.NoMem;
}
return IncrementingAllocator {
.allocator = Allocator {
.allocFn = alloc,
.reallocFn = realloc,
.freeFn = free,
},
.bytes = @intToPtr(&u8, addr)[0..capacity],
.end_index = 0,
};
},
else => @compileError("Unsupported OS"),
}
}
fn deinit(self: &IncrementingAllocator) {
_ = os.posix.munmap(self.bytes.ptr, self.bytes.len);
}
fn alloc(allocator: &Allocator, n: usize) -> %[]u8 {
const self = @fieldParentPtr(IncrementingAllocator, "allocator", allocator);
const new_end_index = self.end_index + n;
if (new_end_index > self.bytes.len) {
return error.NoMem;
}
const result = self.bytes[self.end_index..new_end_index];
self.end_index = new_end_index;
return result;
}
fn realloc(allocator: &Allocator, old_mem: []u8, new_size: usize) -> %[]u8 {
const result = %return alloc(allocator, new_size);
copy(u8, result, old_mem);
return result;
}
fn free(allocator: &Allocator, bytes: []u8) {
// Do nothing. That's the point of an incrementing allocator.
}
};
/// Copy all of source into dest at position 0.
/// dest.len must be >= source.len.
pub fn copy(comptime T: type, dest: []T, source: []const T) {
// TODO instead of manually doing this check for the whole array
// and turning off debug safety, the compiler should detect loops like
// this and automatically omit safety checks for loops
@setDebugSafety(this, false);
assert(dest.len >= source.len);
for (source) |s, i| dest[i] = s;
}
pub fn set(comptime T: type, dest: []T, value: T) {
for (dest) |*d| *d = value;
}
/// Return < 0, == 0, or > 0 if memory a is less than, equal to, or greater than,
/// memory b, respectively.
pub fn cmp(comptime T: type, a: []const T, b: []const T) -> Cmp {
const n = math.min(a.len, b.len);
var i: usize = 0;
while (i < n) : (i += 1) {
if (a[i] == b[i]) continue;
return if (a[i] > b[i]) Cmp.Greater else if (a[i] < b[i]) Cmp.Less else Cmp.Equal;
}
return if (a.len > b.len) Cmp.Greater else if (a.len < b.len) Cmp.Less else Cmp.Equal;
}
/// Compares two slices and returns whether they are equal.
pub fn eql(comptime T: type, a: []const T, b: []const T) -> bool {
if (a.len != b.len) return false;
for (a) |item, index| {
if (b[index] != item) return false;
}
return true;
}
/// Copies ::m to newly allocated memory. Caller is responsible to free it.
pub fn dupe(allocator: &Allocator, comptime T: type, m: []const T) -> %[]T {
const new_buf = %return allocator.alloc(T, m.len);
copy(T, new_buf, m);
return new_buf;
}
/// Linear search for the index of a scalar value inside a slice.
pub fn indexOfScalar(comptime T: type, slice: []const T, value: T) -> ?usize {
for (slice) |item, i| {
if (item == value) {
return i;
}
}
return null;
}
// TODO boyer-moore algorithm
pub fn indexOf(comptime T: type, haystack: []const T, needle: []const T) -> ?usize {
if (needle.len > haystack.len)
return null;
var i: usize = 0;
const end = haystack.len - needle.len;
while (i <= end) : (i += 1) {
if (eql(T, haystack[i .. i + needle.len], needle))
return i;
}
return null;
}
test "mem.indexOf" {
assert(??indexOf(u8, "one two three four", "four") == 14);
assert(indexOf(u8, "one two three four", "gour") == null);
assert(??indexOf(u8, "foo", "foo") == 0);
assert(indexOf(u8, "foo", "fool") == null);
}
/// Reads an integer from memory with size equal to bytes.len.
/// T specifies the return type, which must be large enough to store
/// the result.
pub fn readInt(bytes: []const u8, comptime T: type, big_endian: bool) -> T {
if (T.bit_count == 8) {
return bytes[0];
}
var result: T = 0;
if (big_endian) {
for (bytes) |b| {
result = (result << 8) | b;
}
} else {
const ShiftType = math.Log2Int(T);
for (bytes) |b, index| {
result = result | (T(b) << ShiftType(index * 8));
}
}
return result;
}
/// Writes an integer to memory with size equal to bytes.len. Pads with zeroes
/// to fill the entire buffer provided.
/// value must be an integer.
pub fn writeInt(buf: []u8, value: var, big_endian: bool) {
const uint = @IntType(false, @typeOf(value).bit_count);
var bits = @truncate(uint, value);
if (big_endian) {
var index: usize = buf.len;
while (index != 0) {
index -= 1;
buf[index] = @truncate(u8, bits);
bits >>= 8;
}
} else {
for (buf) |*b| {
*b = @truncate(u8, bits);
bits >>= 8;
}
}
assert(bits == 0);
}
pub fn hash_slice_u8(k: []const u8) -> u32 {
// FNV 32-bit hash
var h: u32 = 2166136261;
for (k) |b| {
h = (h ^ b) *% 16777619;
}
return h;
}
pub fn eql_slice_u8(a: []const u8, b: []const u8) -> bool {
return eql(u8, a, b);
}
/// Returns an iterator that iterates over the slices of ::s that are not
/// the byte ::c.
/// split(" abc def ghi ")
/// Will return slices for "abc", "def", "ghi", null, in that order.
pub fn split(s: []const u8, c: u8) -> SplitIterator {
SplitIterator {
.index = 0,
.s = s,
.c = c,
}
}
test "mem.split" {
var it = split(" abc def ghi ", ' ');
assert(eql(u8, ??it.next(), "abc"));
assert(eql(u8, ??it.next(), "def"));
assert(eql(u8, ??it.next(), "ghi"));
assert(it.next() == null);
}
pub fn startsWith(comptime T: type, haystack: []const T, needle: []const T) -> bool {
return if (needle.len > haystack.len) false else eql(T, haystack[0 .. needle.len], needle);
}
const SplitIterator = struct {
s: []const u8,
c: u8,
index: usize,
pub fn next(self: &SplitIterator) -> ?[]const u8 {
// move to beginning of token
while (self.index < self.s.len and self.s[self.index] == self.c) : (self.index += 1) {}
const start = self.index;
if (start == self.s.len) {
return null;
}
// move to end of token
while (self.index < self.s.len and self.s[self.index] != self.c) : (self.index += 1) {}
const end = self.index;
return self.s[start..end];
}
/// Returns a slice of the remaining bytes. Does not affect iterator state.
pub fn rest(self: &const SplitIterator) -> []const u8 {
// move to beginning of token
var index: usize = self.index;
while (index < self.s.len and self.s[index] == self.c) : (index += 1) {}
return self.s[index..];
}
};
/// Naively combines a series of strings with a separator.
/// Allocates memory for the result, which must be freed by the caller.
pub fn join(allocator: &Allocator, sep: u8, strings: ...) -> %[]u8 {
comptime assert(strings.len >= 1);
var total_strings_len: usize = strings.len; // 1 sep per string
{
comptime var string_i = 0;
inline while (string_i < strings.len) : (string_i += 1) {
const arg = ([]const u8)(strings[string_i]);
total_strings_len += arg.len;
}
}
const buf = %return allocator.alloc(u8, total_strings_len);
%defer allocator.free(buf);
var buf_index: usize = 0;
comptime var string_i = 0;
inline while (true) {
const arg = ([]const u8)(strings[string_i]);
string_i += 1;
copy(u8, buf[buf_index..], arg);
buf_index += arg.len;
if (string_i >= strings.len) break;
if (buf[buf_index - 1] != sep) {
buf[buf_index] = sep;
buf_index += 1;
}
}
return buf[0..buf_index];
}
test "mem.join" {
assert(eql(u8, %%join(&debug.global_allocator, ',', "a", "b", "c"), "a,b,c"));
assert(eql(u8, %%join(&debug.global_allocator, ',', "a"), "a"));
}
test "testStringEquality" {
assert(eql(u8, "abcd", "abcd"));
assert(!eql(u8, "abcdef", "abZdef"));
assert(!eql(u8, "abcdefg", "abcdef"));
}
test "testReadInt" {
testReadIntImpl();
comptime testReadIntImpl();
}
fn testReadIntImpl() {
{
const bytes = []u8{ 0x12, 0x34, 0x56, 0x78 };
assert(readInt(bytes, u32, true) == 0x12345678);
assert(readInt(bytes, u32, false) == 0x78563412);
}
{
const buf = []u8{0x00, 0x00, 0x12, 0x34};
const answer = readInt(buf, u64, true);
assert(answer == 0x00001234);
}
{
const buf = []u8{0x12, 0x34, 0x00, 0x00};
const answer = readInt(buf, u64, false);
assert(answer == 0x00003412);
}
}
test "testWriteInt" {
testWriteIntImpl();
comptime testWriteIntImpl();
}
fn testWriteIntImpl() {
var bytes: [4]u8 = undefined;
writeInt(bytes[0..], u32(0x12345678), true);
assert(eql(u8, bytes, []u8{ 0x12, 0x34, 0x56, 0x78 }));
writeInt(bytes[0..], u32(0x78563412), false);
assert(eql(u8, bytes, []u8{ 0x12, 0x34, 0x56, 0x78 }));
writeInt(bytes[0..], u16(0x1234), true);
assert(eql(u8, bytes, []u8{ 0x00, 0x00, 0x12, 0x34 }));
writeInt(bytes[0..], u16(0x1234), false);
assert(eql(u8, bytes, []u8{ 0x34, 0x12, 0x00, 0x00 }));
}
pub fn min(comptime T: type, slice: []const T) -> T {
var best = slice[0];
var i: usize = 1;
while (i < slice.len) : (i += 1) {
best = math.min(best, slice[i]);
}
return best;
}
test "mem.min" {
assert(min(u8, "abcdefg") == 'a');
}
pub fn max(comptime T: type, slice: []const T) -> T {
var best = slice[0];
var i: usize = 1;
while (i < slice.len) : (i += 1) {
best = math.max(best, slice[i]);
}
return best;
}
test "mem.max" {
assert(max(u8, "abcdefg") == 'g');
}