Desse
Ultra fast binary serialization and deserialization for types with a constant size (known at compile time). This
crate cannot be used to serialize or deserialize dynamically allocated types, such as,
String
, Vec
, HashMap
, etc., and types
with unknown size at compile time such as slices
, &str
, etc.
Binary Encoding Scheme
This crate uses a minimal binary encoding scheme such that the size of encoded object will be smaller than (in cases
where Rust adds padding bytes for alignment) or equal to it's size in a running Rust program. For example, consider
the following struct
:
struct MyStruct {
a: u8,
b: u16,
}
DesseStatic::serialize
will serialize this struct in [u8; 3]
where 3
is the sum of sizes of u8
and u16
.
Usage
Add desse
in your Cargo.toml
's dependencies
section.
[dependencies]
desse = "0.2"
DesseStatic
trait can be implemented for any struct or enum (whose size is known at compile time) using derive
macro. This crate also provides a derive
macro for implementing DesseSized
trait which is necessary for implementing
DesseStatic
trait.
use desse::{DesseStatic, DesseSized};
#[derive(Debug, PartialEq, DesseStatic, DesseSized)]
struct MyStruct {
a: u8,
b: u16,
}
Now, you can use DesseStatic::serialize
and DesseStatic::deserialize_from
for serialization and deserialization of
this struct.
let my_struct = MyStruct { a: 5, b: 1005 };
let serialized: [u8; 3] = my_struct.serialize();
let new_struct = MyStruct::deserialize_from(&serialized);
assert_eq!(my_struct, new_struct);
Note that DesseStatic::serialize
returns an array of fixed length (3
in above case) and Desse::deserialize
takes
reference to an array of fixed length as argument.
Performance
This crate values performance more than anything. We don't shy away from using tested and verified unsafe code if it improves performance.
Benchmarks
Below are the benchmark results of comparison between desse
and bincode
serializing and deserializing same struct
:
struct::serialize/desse::serialize
time: [1.6228 ns 1.6326 ns 1.6434 ns]
change: [-1.1985% +0.0554% +1.2769%] (p = 0.94 > 0.05)
No change in performance detected.
Found 8 outliers among 100 measurements (8.00%)
2 (2.00%) high mild
6 (6.00%) high severe
struct::serialize/bincode::serialize
time: [19.991 ns 20.081 ns 20.201 ns]
change: [-1.0739% +0.3569% +1.7361%] (p = 0.63 > 0.05)
No change in performance detected.
Found 12 outliers among 100 measurements (12.00%)
3 (3.00%) high mild
9 (9.00%) high severe
struct::deserialize/desse::deserialize
time: [1.6063 ns 1.6101 ns 1.6144 ns]
change: [-1.3079% -0.1278% +1.0394%] (p = 0.84 > 0.05)
No change in performance detected.
Found 7 outliers among 100 measurements (7.00%)
1 (1.00%) high mild
6 (6.00%) high severe
struct::deserialize/bincode::deserialize
time: [22.004 ns 22.094 ns 22.209 ns]
change: [-1.1573% +0.0698% +1.3631%] (p = 0.92 > 0.05)
No change in performance detected.
Found 9 outliers among 100 measurements (9.00%)
3 (3.00%) high mild
6 (6.00%) high severe
It is clear from above benchmarks that bincode
takes 20.081 ns
on an average for serialization whereas desse
takes
1.6326 ns
. The results are also similar for deserialization where bincode
takes 22.094 ns
and desse
takes
1.6101 ns
.
You can run benchmarks by running following command:
cargo bench
Future Improvements
Once const_generics
is implemented
in Rust, we can provide default implementations for many types such as, impl DesseStatic for [T; n] where T: DesseStatic
,
and other variable size statically allocated types in Rust.
License
Licensed under either of
- Apache License, Version 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
- MIT license (http://opensource.org/licenses/MIT)
at your option.
Contribution
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.