mirror of
https://github.com/oven-sh/bun
synced 2026-02-11 19:38:58 +00:00
Add a fast path to TextDecoder UTF-16 for when you pass a Uint16Array
This commit is contained in:
Jarred Sumner
@@ -415,6 +415,125 @@ pub const TextDecoder = struct {
|
||||
}
|
||||
const Vector16 = std.meta.Vector(16, u16);
|
||||
const max_16_ascii: Vector16 = @splat(16, @as(u16, 127));
|
||||
|
||||
fn decodeUTF16WithAlignment(
|
||||
_: *TextDecoder,
|
||||
comptime Slice: type,
|
||||
slice: Slice,
|
||||
ctx: js.JSContextRef,
|
||||
) JSC.C.JSValueRef {
|
||||
var i: usize = 0;
|
||||
|
||||
while (i < slice.len) {
|
||||
while (i + strings.ascii_u16_vector_size <= slice.len) {
|
||||
const vec: strings.AsciiU16Vector = slice[i..][0..strings.ascii_u16_vector_size].*;
|
||||
if ((@reduce(
|
||||
.Or,
|
||||
@bitCast(
|
||||
strings.AsciiVectorU16U1,
|
||||
vec > strings.max_u16_ascii,
|
||||
) | @bitCast(
|
||||
strings.AsciiVectorU16U1,
|
||||
vec < strings.min_u16_ascii,
|
||||
),
|
||||
) == 0)) {
|
||||
break;
|
||||
}
|
||||
i += strings.ascii_u16_vector_size;
|
||||
}
|
||||
while (i < slice.len and slice[i] <= 127) {
|
||||
i += 1;
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
||||
// is this actually a UTF-16 string that is just ascii?
|
||||
// we can still allocate as UTF-16 and just copy the bytes
|
||||
if (i == slice.len) {
|
||||
if (comptime Slice == []u16) {
|
||||
return JSC.C.JSValueMakeString(ctx, JSC.C.JSStringCreateWithCharacters(slice.ptr, slice.len));
|
||||
} else {
|
||||
var str = ZigString.init("");
|
||||
str.ptr = @ptrCast([*]u8, slice.ptr);
|
||||
str.len = slice.len;
|
||||
str.markUTF16();
|
||||
return str.toValueGC(ctx.ptr()).asObjectRef();
|
||||
}
|
||||
}
|
||||
|
||||
var buffer = std.ArrayListAlignedUnmanaged(u16, @alignOf(@TypeOf(slice.ptr))){};
|
||||
buffer.ensureTotalCapacity(default_allocator, slice.len) catch unreachable;
|
||||
buffer.items.len = i;
|
||||
defer buffer.deinit(
|
||||
default_allocator,
|
||||
);
|
||||
|
||||
if (comptime Slice == []u16) {
|
||||
@memcpy(
|
||||
std.mem.sliceAsBytes(buffer.items).ptr,
|
||||
std.mem.sliceAsBytes(slice).ptr,
|
||||
std.mem.sliceAsBytes(slice[0..i]).len,
|
||||
);
|
||||
} else {
|
||||
for (slice[0..i]) |ch, j| {
|
||||
buffer.items[j] = ch;
|
||||
}
|
||||
}
|
||||
|
||||
const first_high_surrogate = 0xD800;
|
||||
const last_high_surrogate = 0xDBFF;
|
||||
const first_low_surrogate = 0xDC00;
|
||||
const last_low_surrogate = 0xDFFF;
|
||||
|
||||
var remainder = slice[i..];
|
||||
while (remainder.len > 0) {
|
||||
switch (remainder[0]) {
|
||||
0...127 => {
|
||||
var count: usize = 1;
|
||||
while (remainder.len > count and remainder[count] <= 127) : (count += 1) {}
|
||||
buffer.ensureUnusedCapacity(default_allocator, count) catch unreachable;
|
||||
const prev = buffer.items.len;
|
||||
buffer.items.len += count;
|
||||
for (remainder[0..count]) |char, j| {
|
||||
buffer.items[prev + j] = char;
|
||||
}
|
||||
remainder = remainder[count..];
|
||||
},
|
||||
first_high_surrogate...last_high_surrogate => |first| {
|
||||
if (remainder.len > 1) {
|
||||
if (remainder[1] >= first_low_surrogate and remainder[1] <= last_low_surrogate) {
|
||||
buffer.ensureUnusedCapacity(default_allocator, 2) catch unreachable;
|
||||
buffer.items.ptr[buffer.items.len] = first;
|
||||
buffer.items.ptr[buffer.items.len + 1] = remainder[1];
|
||||
buffer.items.len += 2;
|
||||
remainder = remainder[2..];
|
||||
continue;
|
||||
}
|
||||
}
|
||||
buffer.ensureUnusedCapacity(default_allocator, 1) catch unreachable;
|
||||
buffer.items.ptr[buffer.items.len] = strings.unicode_replacement;
|
||||
buffer.items.len += 1;
|
||||
remainder = remainder[1..];
|
||||
continue;
|
||||
},
|
||||
|
||||
// Is this an unpaired low surrogate or four-digit hex escape?
|
||||
else => {
|
||||
buffer.ensureUnusedCapacity(default_allocator, 1) catch unreachable;
|
||||
buffer.items.ptr[buffer.items.len] = strings.unicode_replacement;
|
||||
buffer.items.len += 1;
|
||||
remainder = remainder[1..];
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
var out = ZigString.init("");
|
||||
out.ptr = @ptrCast([*]u8, buffer.items.ptr);
|
||||
out.len = buffer.items.len;
|
||||
out.markUTF16();
|
||||
return out.toValueGC(ctx.ptr()).asObjectRef();
|
||||
}
|
||||
|
||||
pub fn decode(
|
||||
this: *TextDecoder,
|
||||
ctx: js.JSContextRef,
|
||||
@@ -438,6 +557,9 @@ pub const TextDecoder = struct {
|
||||
return ZigString.Empty.toValue(ctx.ptr()).asObjectRef();
|
||||
}
|
||||
|
||||
JSC.C.JSValueProtect(ctx, args[0]);
|
||||
defer JSC.C.JSValueUnprotect(ctx, args[0]);
|
||||
|
||||
switch (this.encoding) {
|
||||
EncodingLabel.@"latin1" => {
|
||||
var str = ZigString.init(array_buffer.slice());
|
||||
@@ -486,96 +608,16 @@ pub const TextDecoder = struct {
|
||||
}
|
||||
}
|
||||
}
|
||||
return str.toValue(ctx.ptr()).asObjectRef();
|
||||
|
||||
return str.toValueGC(ctx.ptr()).asObjectRef();
|
||||
},
|
||||
|
||||
EncodingLabel.@"UTF-16LE" => {
|
||||
var slice = array_buffer.asU16();
|
||||
var i: usize = 0;
|
||||
|
||||
while (i < slice.len) {
|
||||
while (i + 16 <= slice.len) {
|
||||
const vec: Vector16 = slice[i..][0..16].*;
|
||||
if ((@reduce(.Or, vec > max_16_ascii))) {
|
||||
break;
|
||||
}
|
||||
i += 16;
|
||||
}
|
||||
while (i < slice.len and slice[i] <= 127) {
|
||||
i += 1;
|
||||
}
|
||||
break;
|
||||
if (std.mem.isAligned(@ptrToInt(array_buffer.ptr) + @as(usize, array_buffer.offset), @alignOf([*]u16))) {
|
||||
return this.decodeUTF16WithAlignment([]u16, array_buffer.asU16(), ctx);
|
||||
}
|
||||
|
||||
// is this actually a UTF-16 string that is just ascii?
|
||||
// we can still allocate as UTF-16 and just copy the bytes
|
||||
if (i == slice.len) {
|
||||
return JSC.C.JSValueMakeString(ctx, JSC.C.JSStringCreateWithCharacters(slice.ptr, slice.len));
|
||||
}
|
||||
|
||||
var buffer = std.ArrayListAlignedUnmanaged(u16, 1){};
|
||||
buffer.ensureTotalCapacity(default_allocator, slice.len) catch unreachable;
|
||||
buffer.items.len = i;
|
||||
defer buffer.deinit(
|
||||
default_allocator,
|
||||
);
|
||||
|
||||
for (slice[0..i]) |char, j| {
|
||||
buffer.items[j] = char;
|
||||
}
|
||||
|
||||
const first_high_surrogate = 0xD800;
|
||||
const last_high_surrogate = 0xDBFF;
|
||||
const first_low_surrogate = 0xDC00;
|
||||
const last_low_surrogate = 0xDFFF;
|
||||
|
||||
var remainder = slice[i..];
|
||||
while (remainder.len > 0) {
|
||||
switch (remainder[0]) {
|
||||
0...127 => {
|
||||
var count: usize = 1;
|
||||
while (remainder.len > count and remainder[count] <= 127) : (count += 1) {}
|
||||
buffer.ensureUnusedCapacity(default_allocator, count) catch unreachable;
|
||||
const prev = buffer.items.len;
|
||||
buffer.items.len += count;
|
||||
for (remainder[0..count]) |char, j| {
|
||||
buffer.items[prev + j] = char;
|
||||
}
|
||||
remainder = remainder[count..];
|
||||
},
|
||||
first_high_surrogate...last_high_surrogate => |first| {
|
||||
if (remainder.len > 1) {
|
||||
if (remainder[1] >= first_low_surrogate and remainder[1] <= last_low_surrogate) {
|
||||
buffer.ensureUnusedCapacity(default_allocator, 2) catch unreachable;
|
||||
buffer.items.ptr[buffer.items.len] = first;
|
||||
buffer.items.ptr[buffer.items.len + 1] = remainder[1];
|
||||
buffer.items.len += 2;
|
||||
remainder = remainder[2..];
|
||||
continue;
|
||||
}
|
||||
}
|
||||
buffer.ensureUnusedCapacity(default_allocator, 1) catch unreachable;
|
||||
buffer.items.ptr[buffer.items.len] = strings.unicode_replacement;
|
||||
buffer.items.len += 1;
|
||||
remainder = remainder[1..];
|
||||
continue;
|
||||
},
|
||||
|
||||
// Is this an unpaired low surrogate or four-digit hex escape?
|
||||
else => {
|
||||
buffer.ensureUnusedCapacity(default_allocator, 1) catch unreachable;
|
||||
buffer.items.ptr[buffer.items.len] = strings.unicode_replacement;
|
||||
buffer.items.len += 1;
|
||||
remainder = remainder[1..];
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
var out = ZigString.init("");
|
||||
out.ptr = @ptrCast([*]u8, buffer.items.ptr);
|
||||
out.len = buffer.items.len;
|
||||
out.markUTF16();
|
||||
return out.toValueGC(ctx.ptr()).asObjectRef();
|
||||
return this.decodeUTF16WithAlignment([]align(1) u16, array_buffer.asU16Unaligned(), ctx);
|
||||
},
|
||||
else => {
|
||||
JSC.throwInvalidArguments("TextDecoder.decode set to unsupported encoding", .{}, ctx, exception);
|
||||
|
||||
Reference in New Issue
Block a user