Commit Graph

6 Commits

Author SHA1 Message Date
Jun Wu
c4ec99ded4 minibytes: implement Text
Summary:
Text is a reference-counted shared String.
It's similar to Bytes but works for utf-8 strings.

The motivation is to replace configparser's use of Bytes to Text.

Reviewed By: markbt

Differential Revision: D20432940

fbshipit-source-id: ef990255d269e60d433c6520819f60ccdcbe488f
2020-03-13 10:51:41 -07:00
Jun Wu
7895e70dcf minibytes: make Bytes abstract
Summary: This makes it possible to implement "Text". See the next diff.

Reviewed By: markbt

Differential Revision: D20432943

fbshipit-source-id: 94b3810ab205c260d33f57bd637e4accc3ee871d
2020-03-13 10:51:40 -07:00
Jun Wu
e9b14b3608 minibytes: implement From<&'static {str,[u8]}>
Summary:
This makes the API easier to use.

Practically this makes it easier for configparser to migrate to minibytes.

Reviewed By: markbt

Differential Revision: D20432942

fbshipit-source-id: ad08eb118d2216054dc24c86b0b129ae82b9d17c
2020-03-13 10:51:40 -07:00
Jun Wu
abb8ccb346 minibytes: add serde support
Summary:
See also https://github.com/serde-rs/bytes/.

This will be used in the `dag` crate.

Reviewed By: DurhamG

Differential Revision: D19770858

fbshipit-source-id: 2a870a564e0ceecdc7a4667853b2b2a5ea4ce6e3
2020-02-07 14:21:39 -08:00
Jun Wu
8029cd3878 minibytes: port benchmark from tokio/bytes
Summary:
Performance looks okay comparing with tokio/bytes v0.5.4:

minibytes:

  test clone_arc_vec        ... bench:      16,542 ns/iter (+/- 1,524)
  test clone_shared         ... bench:      16,211 ns/iter (+/- 596)
  test clone_static         ... bench:       1,437 ns/iter (+/- 502)
  test deref_shared         ... bench:         367 ns/iter (+/- 101)
  test deref_static         ... bench:         366 ns/iter (+/- 1)
  test deref_unique         ... bench:         367 ns/iter (+/- 4)
  test from_long_slicd      ... bench:          91 ns/iter (+/- 18) = 1406 MB/s
  test slice_empty          ... bench:      10,382 ns/iter (+/- 104)
  test slice_short_from_arc ... bench:      23,823 ns/iter (+/- 1,411)

tokio/bytes:

  test clone_arc_vec        ... bench:      16,213 ns/iter (+/- 1,864)
  test clone_shared         ... bench:      18,685 ns/iter (+/- 634)
  test clone_static         ... bench:       3,983 ns/iter (+/- 163)
  test deref_shared         ... bench:         366 ns/iter (+/- 26)
  test deref_static         ... bench:         373 ns/iter (+/- 36)
  test deref_unique         ... bench:         391 ns/iter (+/- 33)
  test from_long_slice      ... bench:          67 ns/iter (+/- 7) = 1910 MB/s
  test slice_empty          ... bench:      15,149 ns/iter (+/- 1,708)
  test slice_short_from_arc ... bench:      36,541 ns/iter (+/- 3,485)

clone_static is faster because minibytes don't call into vtable's clone.
from_long_slice is slower because minibytes uses Arc unconditionally while bytes
can avoid Arc overhead if refcount is 1.

Reviewed By: DurhamG

Differential Revision: D19770857

fbshipit-source-id: 5bafcc57a38c68baccfcafd3906f1a47b2bf4530
2020-02-07 14:21:39 -08:00
Jun Wu
108f1c947a minibytes: minimalist zero-copy Bytes with mmap support
Summary:
This crate provides the core features of the commonly known `Bytes` crate:
zero-copy slicing and cloning, while also supports mmap-backed buffers.

The main motivation is to replace `Mmap` in `indexedlog`. That has multiple
benefits:
- Handles 0-sized mmap more cleanly.
- Handles clones more cleanly.
- Gain the flexibility to zero-copy data without lifetime / reference.
- Gain the flexibility to switch to non-mmap data.

The `bytes::Bytes` crate does not yet support mmap buffers as of its latest
release (0.5.4).

Implementation wise, `minibytes::Bytes` uses `Option<Arc<dyn Trait>>` for the
"trait object". This makes implementing the mmap storage just one line.
`bytes 0.5.4` re-invents the "trait object" manually using unsafe code. It requires
about 50 lines to implement the mmap storage (in D19756122).

Reviewed By: xavierd

Differential Revision: D19770856

fbshipit-source-id: 8cfa7052a18ac2e0cd6348b77d5e2a4acc61195c
2020-02-07 14:21:38 -08:00