public object FfiConverterString: FfiConverter<String, RustBuffer.ByValue> {
    // Note: we don't inherit from FfiConverterRustBuffer, because we use a
    // special encoding when lowering/lifting.  We can use `RustBuffer.len` to
    // store our length and avoid writing it out to the buffer.
    override fun lift(value: RustBuffer.ByValue): String {
        try {
            val byteArr = ByteArray(value.len.toInt())
            value.asByteBuffer()!!.get(byteArr)
            return byteArr.toString(Charsets.UTF_8)
        } finally {
            RustBuffer.free(value)
        }
    }

    override fun read(buf: ByteBuffer): String {
        val len = buf.getInt()
        val byteArr = ByteArray(len)
        buf.get(byteArr)
        return byteArr.toString(Charsets.UTF_8)
    }

    fun toUtf8(value: String): ByteBuffer {
        // Make sure we don't have invalid UTF-16, check for lone surrogates.
        return Charsets.UTF_8.newEncoder().run {
            onMalformedInput(CodingErrorAction.REPORT)
            encode(CharBuffer.wrap(value))
        }
    }

    override fun lower(value: String): RustBuffer.ByValue {
        val byteBuf = toUtf8(value)
        // Ideally we'd pass these bytes to `ffi_bytebuffer_from_bytes`, but doing so would require us
        // to copy them into a JNA `Memory`. So we might as well directly copy them into a `RustBuffer`.
        val rbuf = RustBuffer.alloc(byteBuf.limit().toULong())
        rbuf.asByteBuffer()!!.put(byteBuf)
        return rbuf
    }

    // We aren't sure exactly how many bytes our string will be once it's UTF-8
    // encoded.  Allocate 3 bytes per UTF-16 code unit which will always be
    // enough.
    override fun allocationSize(value: String): ULong {
        val sizeForLength = 4UL
        val sizeForString = value.length.toULong() * 3UL
        return sizeForLength + sizeForString
    }

    override fun write(value: String, buf: ByteBuffer) {
        val byteBuf = toUtf8(value)
        buf.putInt(byteBuf.limit())
        buf.put(byteBuf)
    }
}