utf16LeToUtf8Impl

fn utf16LeToUtf8Impl(utf8: []u8, utf16le: []const u16, comptime surrogates: Surrogates) (switch (surrogates)

File

Code

fn utf16LeToUtf8Impl(utf8: []u8, utf16le: []const u16, comptime surrogates: Surrogates) (switch (surrogates) {
    .cannot_encode_surrogate_half => Utf16LeToUtf8Error,
    .can_encode_surrogate_half => error{},
})!usize {
    var dest_index: usize = 0;

    var remaining = utf16le;
    vectorized: {
        const chunk_len = std.simd.suggestVectorLength(u16) orelse break :vectorized;
        const Chunk = @Vector(chunk_len, u16);

        // Fast path. Check for and encode ASCII characters at the start of the input.
        while (remaining.len >= chunk_len) {
            const chunk: Chunk = remaining[0..chunk_len].*;
            const mask: Chunk = @splat(mem.nativeToLittle(u16, 0x7F));
            if (@reduce(.Or, chunk | mask != mask)) {
                // found a non ASCII code unit
                break;
            }
            const ascii_chunk: @Vector(chunk_len, u8) = @truncate(mem.nativeToLittle(Chunk, chunk));
            utf8[dest_index..][0..chunk_len].* = ascii_chunk;
            dest_index += chunk_len;
            remaining = remaining[chunk_len..];
        }
    }

    switch (surrogates) {
        .cannot_encode_surrogate_half => {
            var it = Utf16LeIterator.init(remaining);
            while (try it.nextCodepoint()) |codepoint| {
                dest_index += utf8Encode(codepoint, utf8[dest_index..]) catch |err| switch (err) {
                    // The maximum possible codepoint encoded by UTF-16 is U+10FFFF,
                    // which is within the valid codepoint range.
                    error.CodepointTooLarge => unreachable,
                    // We know the codepoint was valid in UTF-16, meaning it is not
                    // an unpaired surrogate codepoint.
                    error.Utf8CannotEncodeSurrogateHalf => unreachable,
                };
            }
        },
        .can_encode_surrogate_half => {
            var it = Wtf16LeIterator.init(remaining);
            while (it.nextCodepoint()) |codepoint| {
                dest_index += wtf8Encode(codepoint, utf8[dest_index..]) catch |err| switch (err) {
                    // The maximum possible codepoint encoded by UTF-16 is U+10FFFF,
                    // which is within the valid codepoint range.
                    error.CodepointTooLarge => unreachable,
                };
            }
        },
    }
    return dest_index;
}