#[derive(Default, Ord, Hash, Eq, PartialEq, PartialOrd, Clone)]

    fn deref(&self) -> &[u8] {

        &self.0

    }

        if let Ok(string) = std::str::from_utf8(self.as_ref()) {

            if string.bytes().all(|b| (32..=127).contains(&b)) {

                return f.write_str(string);

            }

        }

        if let Some((first_nonzero_byte, _)) = self

            .as_ref()

            .iter()

            .copied()

            .enumerate()

            .find(|(_index, b)| *b != 0)

            if first_nonzero_byte > 0 {

                write!(f, "{first_nonzero_byte:x}$")?;

                f.write_char('$')?;

            for (index, byte) in self[first_nonzero_byte..].iter().enumerate() {

                if index > 0 {

                    write!(f, "{byte:02x}")?;

                    write!(f, "{byte:x}")?;

            Ok(())

            write!(f, "{:x}$", self.len())

    }

    fn from(id: DocumentId) -> Self {

        Identifier::from(id.to_vec())

    }

    fn from(id: &'a DocumentId) -> Self {

        Identifier::from(&**id)

    }

#[test]

fn document_id_identifier_tests() {

    assert_eq!(

        Identifier::from(DocumentId::new("hello").unwrap()),

        Identifier::from("hello")

    );

    assert_eq!(

        Identifier::from(DocumentId::from_u64(1)),

        Identifier::from(1)

    );

}

const fn decode_hex_nibble(byte: u8) -> Result<u8, InvalidHexadecimal> {

    match byte {

        b'0'..=b'9' => Ok(byte - b'0'),

        b'A'..=b'F' => Ok(byte - b'A' + 10),

        b'a'..=b'f' => Ok(byte - b'a' + 10),

}

    fn from_str(s: &str) -> Result<Self, Self::Err> {

        if s.is_empty() {

        }

        let bytes = s.as_bytes();

        if let Some((pound_offset, _)) = s.bytes().enumerate().find(|(_index, b)| *b == b'$') {

            if pound_offset > 5 {

            }

            let preceding_zeroes = if pound_offset > 0 {

                let mut length = TinyVec::<[u8; 1]>::new();

                decode_big_endian_hex(&bytes[0..pound_offset], &mut length)?;

                let mut zeroes = [0_u8; size_of::<usize>()];

                let offset = zeroes.len() - length.len();

                zeroes[offset..].copy_from_slice(&length);

                usize::from_be_bytes(zeroes)

                0

            let mut id = TinyVec::new();

            decode_big_endian_hex(&bytes[pound_offset + 1..], &mut id)?;

            if preceding_zeroes > 0 {

                let total_length = preceding_zeroes + id.len();

                if total_length > Self::MAX_LENGTH {

                }

                // The full length indicated a longer ID, so we need to prefix some null bytes.

                id.splice(0..0, std::iter::repeat(0).take(preceding_zeroes));

            }

            Ok(Self(id))

        } else if bytes.len() > Self::MAX_LENGTH {

            Self::try_from(bytes)

    }

fn decode_big_endian_hex<A: Array<Item = u8>>(

    bytes: &[u8],

    output: &mut TinyVec<A>,

) -> Result<(), crate::Error> {

    let mut chunks = if bytes.len() & 1 == 0 {

        bytes.chunks_exact(2)

        output.push(decode_hex_nibble(bytes[0])?);

        bytes[1..].chunks_exact(2)

    for chunk in &mut chunks {

        let upper = decode_hex_nibble(chunk[0])?;

        let lower = decode_hex_nibble(chunk[1])?;

        output.push(upper << 4 | lower);

    if !chunks.remainder().is_empty() {

    }

    Ok(())

}

#[test]

fn document_id_parsing() {

    fn test_id(bytes: &[u8], display: &str) {

        let id = DocumentId::try_from(bytes).unwrap();

        let as_string = id.to_string();

        assert_eq!(as_string, display);

        let parsed = DocumentId::from_str(&as_string).unwrap();

        assert_eq!(&*parsed, bytes);

    }

    test_id(b"hello", "hello");

    test_id(b"\x00\x0a\xaf\xfa", "1$aaffa");

    test_id(&1_u128.to_be_bytes(), "f$1");

    test_id(&17_u8.to_be_bytes(), "$11");

    test_id(&[0_u8; 63], "3f$");

    // The above test is the same as this one, at the time of writing, but in

    // case we update MAX_LENGTH in the future, this extra test will ensure the

    // max-length formatting is always tested.

    test_id(

        &vec![0_u8; DocumentId::MAX_LENGTH],

        &format!("{:x}$", DocumentId::MAX_LENGTH),

    );

}

    fn try_from(bytes: &'a [u8]) -> Result<Self, Self::Error> {

        if bytes.len() <= Self::MAX_LENGTH {

            Ok(Self(TinyVec::from(bytes)))

    }

    pub fn new<PrimaryKey: for<'k> Key<'k>, PrimaryKeyRef: KeyEncoding<PrimaryKey> + ?Sized>(

        value: &PrimaryKeyRef,

    ) -> Result<Self, crate::Error> {

        let bytes = value

            .as_ord_bytes()

            .map_err(|err| crate::Error::other("key serialization", err))?;

        Self::try_from(&bytes[..])

    }

    pub fn from_u64(id: u64) -> Self {

        Self::try_from(&id.to_be_bytes()[..]).unwrap()

    }

    pub fn deserialize<'k, PrimaryKey: Key<'k>>(&'k self) -> Result<PrimaryKey, crate::Error> {

        PrimaryKey::from_ord_bytes(ByteSource::Borrowed(self.as_ref()))

            .map_err(|err| crate::Error::other("key serialization", err))

    }

    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>

    where

        S: serde::Serializer,

    {

        serializer.serialize_bytes(self.as_ref())

    }

    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>

    where

        D: serde::Deserializer<'de>,

    {

        deserializer.deserialize_byte_buf(DocumentIdVisitor)

    }

    fn visit_bytes<E>(self, v: &[u8]) -> Result<Self::Value, E>

    where

        E: serde::de::Error,

    {

        Ok(DocumentId(TinyVec::from(v)))

    }

    fn as_ord_bytes(&self) -> Result<Cow<'_, [u8]>, Self::Error> {

        Ok(Cow::Borrowed(self))

    }