FileEncoder in rustc_serialize::opaque - Rust (original) (raw)
Struct FileEncoder
pub struct FileEncoder {
buf: Box<[u8; 65536]>,
buffered: usize,
flushed: usize,
file: File,
res: Result<(), Error>,
path: PathBuf,
finished: bool,
}Expand description
FileEncoder encodes data to file via fixed-size buffer.
There used to be a MemEncoder type that encoded all the data into aVec. FileEncoder is better because its memory use is determined by the size of the buffer, rather than the full length of the encoded data, and because it doesn’t need to reallocate memory along the way.
Write up to N bytes to this encoder.
This function can be used to avoid the overhead of calling memcpy for writes that have runtime-variable length, but are small and have a small fixed upper bound.
This can be used to do in-place encoding as is done for leb128 (without this function we would need to write to a temporary buffer then memcpy into the encoder), and it can also be used to implement the varint scheme we use for rmeta and dep graph encoding, where we only want to encode the first few bytes of an integer. Copying in the whole integer then only advancing the encoder state for the few bytes we care about is more efficient than calling FileEncoder::write_all, because variable-size copies are always lowered to memcpy, which has overhead and contains a lot of logic we can bypass with this function. Note that common architectures support fixed-size writes up to 8 bytes with one instruction, so while this does in some sense do wasted work, we come out ahead.
Note: Most layout information is completely unstable and may even differ between compilations. The only exception is types with certain repr(...) attributes. Please see the Rust Reference's “Type Layout” chapter for details on type layout guarantees.
Size: 64 bytes