Make Reader and Writer generic on Endianness #103
Conversation
src/serde/mod.rs
Outdated
| @@ -69,7 +70,7 @@ pub fn serialize<T>(value: &T, size_limit: SizeLimit) -> SerializeResult<Vec<u8> | |||
| SizeLimit::Infinite => Vec::new() | |||
| }; | |||
|
|
|||
| try!(serialize_into(&mut writer, value, SizeLimit::Infinite)); | |||
| try!(serialize_into::<_, _, ::byteorder::BigEndian>(&mut writer, value, SizeLimit::Infinite)); | |||
src/serde/mod.rs
Outdated
| @@ -107,7 +108,7 @@ pub fn deserialize_from<R, T>(reader: &mut R, size_limit: SizeLimit) -> Deserial | |||
| where R: Read, | |||
| T: serde::Deserialize, | |||
| { | |||
| let mut deserializer = Deserializer::new(reader, size_limit); | |||
| let mut deserializer = Deserializer::<_, BigEndian>::new(reader, size_limit); | |||
There was a problem hiding this comment.
deserialize_from not generic over byte order?
|
@aldanor: in its current state, this PR adds endian genericness to the backend, but not the public functions. That will change in the future, but I wanted to get the difficult part passing tests first. |
|
@TyOverby I see then, thanks; looks good to me given that this will be later exposed to the crate-level functions. // Should there be some explicit tests for big/little endian? |
TyOverby
force-pushed
the
endian-choice
branch
from
6434853
to
db0b3d8
Compare
TyOverby
force-pushed
the
endian-choice
branch
from
db0b3d8
to
dafeb04
Compare
|
@aldanor: I've gone with the decision to add an |
|
Oh, also, I've picked LittleEndian as the new default endianness (mainly for speed on the average computer) |
tests/test.rs
Outdated
| let x = 10u64; | ||
| let little = serialize::<_, byteorder::LittleEndian>(&x, Infinite).unwrap(); | ||
| let big = serialize::<_, byteorder::BigEndian>(&x, Infinite).unwrap(); | ||
| assert!(little != big); |
There was a problem hiding this comment.
Wouldn't it be better to use assert_ne!(little, big);? As I understand it you get better error messages that way.
There was a problem hiding this comment.
huh, I didn't actually know about that one
TyOverby
force-pushed
the
endian-choice
branch
from
eda2fb4
to
9df6a1e
Compare
|
That's one way to go about it I guess. Little endian as a default sounds grand; worth mentioning in the docstrings of the root module somewhere? |
While introducing selectable endianness in bincode-org#103 , the new type parameter has been hidden from the public `serialize()`, `deserialize()` etc. functions, and only made available through an alternate API entry point. The same kind of encapsulation also needs to be performed for the public `Serializer` and `Deserializer` types.
…128) While introducing selectable endianness in #103 , the new type parameter has been hidden from the public `serialize()`, `deserialize()` etc. functions, and only made available through an alternate API entry point. The same kind of encapsulation also needs to be performed for the public `Serializer` and `Deserializer` types.
No description provided.