fix #263

[bsbds]

fix #263
Please briefly answer these questions:

• what problem are you trying to solve? (or if there's no problem, what's the motivation for this change?)

  improve the read and write logic for the RocksSnapshot

• what changes does this pull request make?

  change of SnapshotApi in read_exact and write_all funtions and all associated calls on these funtions

• are there any non-obvious implications of these changes? (does it break compatibility with previous versions, etc)

  read_exact and write_all funtions now use bytes::{Bytes, BytesMut} as parameter types

[codecov]

Codecov Report

Merging #264 (8626b05) into master (74fee6d) will increase coverage by 0.11%.
The diff coverage is 60.36%.

@@            Coverage Diff             @@
##           master     #264      +/-   ##
==========================================
+ Coverage   51.80%   51.92%   +0.11%     
==========================================
  Files          72       73       +1     
  Lines       12265    12264       -1     
==========================================
+ Hits         6354     6368      +14     
+ Misses       5911     5896      -15     

Impacted Files	Coverage Δ
curp/src/server/gc.rs	90.60% <0.00%> (-3.81%)	⬇️
xline/src/client/restore.rs	0.00% <0.00%> (ø)
xline/src/server/watch_server.rs	60.49% <ø> (ø)
xline/src/server/xline_server.rs	0.00% <0.00%> (ø)
xline/src/storage/lease_store/mod.rs	65.83% <ø> (ø)
curp/src/server/curp_node.rs	26.49% <7.01%> (-0.82%)	⬇️
curp/src/server/raw_curp/mod.rs	65.64% <50.00%> (ø)
curp/src/rpc/connect.rs	25.56% <60.00%> (ø)
engine/src/api/snapshot_api.rs	60.00% <60.00%> (ø)
curp/src/server/raw_curp/log.rs	92.47% <66.66%> (-0.24%)	⬇️
... and 13 more

[bsbds]

Current implementation in chunk_mut for BytesMut only expand the capacity of the BytesMut when it already full. So it won't automatically expand the capacity because either the read_buf function only called once(unless the buf's capacity is zero), or the condition that the length is smaller than the capacity have already been checked. @themanforfree

    #[inline]
    fn chunk_mut(&mut self) -> &mut UninitSlice {
        if self.capacity() == self.len() {
            self.reserve(64);
        }
        UninitSlice::from_slice(self.spare_capacity_mut())
    }

So we do rely on this specific implementation of BytesMut, if we use impl BufMut in the parameter, the chunk_mut may grow capacity before it gets full. @Phoenix500526

[Phoenix500526]

The chunk_mut method implemented for Vec<u8> doesn't look very different from the one that BytesMut implemented. FYI.

unsafe impl BufMut for Vec<u8> {
    #[inline]
    fn chunk_mut(&mut self) -> &mut UninitSlice {
        if self.capacity() == self.len() {
            self.reserve(64); // Grow the vec
        }

        let cap = self.capacity();
        let len = self.len();

        let ptr = self.as_mut_ptr();
        unsafe { &mut UninitSlice::from_raw_parts_mut(ptr, cap)[len..] }
    }
}

A Vec<u8> will also grow its capacity when necessary. And a &mut [u8] also implements the BufMut trait, but we don't need to bother this issue.

[bsbds]

Since we refactored the code to use bytes crate, the BytesMut objects are typically used uninitialized. Maybe we should consider change the method name from read_exact to read_buf_exact to match with std::io::Read trait method. https://doc.rust-lang.org/std/io/trait.Read.html#method.read_buf_exact