2019-06-14 20:37:10 +02:00

38 lines
2.5 KiB
Markdown

# capnproto-dotnetcore
A Cap'n Proto implementation for .NET Core
["Cap'n Proto is an insanely fast data interchange format and capability-based RPC system."](https://capnproto.org/) Whilst the original implementation is written in C++ there are several ports to other languages. This is a C# implementation for .NET Core.
Disclaimer: Neither this project nor its author are affiliated with Cap'n Proto. This is just yet another independent implementation of the specification. The following sections assume that you are familiar with [Cap'n Proto](https://capnproto.org/) and probably its [GitHub project](https://github.com/capnproto/capnproto).
## Getting started
For building from scratch you will need Visual Studio >= 2019 (e.g. Community Edition) with suitable workloads for C# / .NET Core (currently .NET Core 2.1) development. For the test suite, you will also need the C++ native workload, [vcpkg](https://github.com/microsoft/vcpkg) and Cap'n Proto release 0.7.0:
```
vcpkg install capnproto
```
Solution/project structure is as follows:
- Capnp.Net.sln contains three projects:
* Capnp.Net.Runtime is the runtime implementation, a .NET assembly.
* capnpc-csharp is the compiler backend for C# language
* Capnp.Net.Runtime.Tests is an MS Unit Testing assembly, containing - you guessed it - the test suite
- CapnpCompatTest.sln compiles to a native x86 executable which depends on the original Cap'n Proto C++ implementation. It is (partially) required by the test suite for interoperability testing.
## Features
The following Cap'n Proto features are currently implemented:
- Serialization/deserialization of all kinds ofdata (structs, groups, unions, lists, capabilities, data, text, enums, even primitives)
- Generics
- Level 1 RPC, including promise pipelining, embargos, and automatic tail calls
- Security (pointer validation, protection against amplification and stack overflow DoS attacks)
- Compiler backend generates reader/writer classes, interfaces, proxies, skeletons (as you know it from the C++ implementation), and additionally so-called "domain classes" for all struct types. A domain class is like a "plain old C# class" for representing a schema-defined struct, but it is decoupled from any underlying message. It provides serialize/deserialize methods for assembling/disassembling the actual message. This provides more convenience, but comes at the price of non-zero serialization overhead (not "infinitely" faster anymore).
These features are not yet implemented:
- Level N RPC with N >= 2
- Packing
- Compression
- Canonicalization
- Dynamic Reflection
- mmap