class UnbufferedPseudoFile

Unbuffered pseudo-file.

This variant is optimized for atomically reading and writing small properties. Unlike buffered

pseudo-files, it is not necessary to re-open the pseudo-file to observe side-effects; the client

can simply seek back to the zero offset and read or write again.

Because reads and writes are not buffered, the maximum size of the property is limited to what

will fit in a single I/O transaction. Unbuffered pseudo-files generally work best for properties

which are likely to be polled or repeatedly modified and which are no larger than the nominal I/O

buffer size used by the intended clients.

As a conservative guideline, we recommend using |BufferedPseudoFile| instead for content larger

than the system page size.

In read mode, the pseudo-file invokes its read handler each time |Read()| is called with a seek

offset of 0, returning at most as many bytes as the client requested and discarding the remainder

(if any).

Reading with a non-zero seek offset returns empty data, indicating end of file.

In write mode, the pseudo-file invokes its write handler each time |Write()| with a seek offset

of 0 is called, passing all of the bytes written by the client as the input string. Likewise,

|Append()| invokes the write handler each time it is called and returns a new end of file offset

of 0.

Writing with a non-zero seek offset returns |ZX_ERR_NO_SPACE|, indicating an attempt to write

data beyond what was accepted by the write handler.

Truncating the length to zero and then closing it without an intervening write is equivalent to

writing 0 bytes. Truncating to a non-zero length returns |ZX_ERR_INVALID_ARGS|.

This class is thread-safe.