All notable changes to this project will be documented in this file. This project adheres to Semantic Versioning.
#![no_std] in rmpvfeature="std" (on by default)RmpRead and RmpWrite traits.std::io::Read (and Write) are missing on #![no_std]Bytes and ByteBuf wrappers, that implement RmpRead/RmpWrite for no_std targets.rmp::decode::read_str_from_slice function for zero-copy reading strings from slices.rmp::decode::read_str_ref, because it's useless and inconvenient.Debug for MarkerReadError.write_sint, so it should mark positive values as unsigned integers. Before this change it marked signed integers larger than 4294967296 with I64 marker, which is not consistent with other MessagePack implementations. Now it should mark such integers with U64 marker.ValueWriteError into I/O error.From and Into traits for u8.read_int function, which allows to read integer values and automatically cast to the expected result type even if they aren't the same. An additional OutOfRange error will be returned in the case of failed numeric cast.NumValueReadError enum with additional OutOfRange variant to be able to detect whether integer decoding failed because of out of range.byteorder dependency to 1.0.write_sint now encodes 64-bit signed integers using the most compact representation.write_uint now encodes 64-bit unsigned integers using the most compact representation.read_array_size function to read_array_len for consistency.read_map_size function to read_map_len for consistency.FixedValueWriteError struct private. All functions, that previously returned such kind of error now return the Standard I/O error.Value and ValueRef enums and associated functions into the separate rmpv crate.byteorder crate errors, because since 0.5 there are no such errors.write_sint_eff function - its functionality can now be done using write_sint instead.write_uint_eff function - its functionality can now be done using write_uint instead.read_*_loosely and read_*_fit were removed in favor of generic read_int function, which allows to read integral values and cast them to the specified result type even if they aren't the same.read_bin_borrow function.is_* methods for Value for checking its underlying value without explicit matching.as_* methods for Value for borrowing its underlying value.Display trait for Value.to_owned instead of to_string while converting ValueRef into Value. This change improves ValueRef::to_owned() method performance by approximately 10-20%.s Also after this change it's cheaper to decode directly into ValueRef with further converting to owned value rather than decoding directly into Value.rmp, rmp-serialize and rmp-serde.ByteBuf.BigEndianRead trait to be implemented only for sized types.PositiveFixnum marker to FixPos.NegativeFixnum marker to FixNeg.FixedString marker to FixStr.FixedArray marker to FixArray.FixedMap to FixMap.rustc_serialize::Encoder should encode signed integers using the most effective underlying representation.rustc_serialize::Decoder should properly map integers to the result type if the decoded value fits in result type's range.ValueRef value struct represents MessagePack'ed value, but unlike an owning Value it owns nothing except its structure. It means that all strings and binaries it contains are borrowed from the byte array from which the value was created.BorrowRead trait, which looks like a standard BufRead but unlike the standard this has an explicit internal buffer lifetime, which allows to borrow from underlying buffer while mutating the type.ValueRef with its own error category.ValueRef with its own error category.ValueRef to Value.Copy trait for Integer and Float enums.write_str function allows to serialize the UTF-8 encoded strings the most efficient way.write_bin function allows to serialize the binary array the most efficient way.std::error::Error trait for error types.read_value function to the rmp::decode module.Value struct to the root crate namespace.Value algebraic data type, which represents an owning MessagePack object. It can be found in rmp::value module.