Deprecated Adamant Temporary Compiler
The Adamant compiler is being bootstrapped. To serve this process, a compiler must first be written in another language. In this case C#. This is that temporary compiler. It will be thrown away when the compiler is re-written in Adamant.
Project Status: Alpha Deprecated
This project should not be used. The current Adamant compiler is the "adamant.tools.compiler.bootstrap" project.
Download and Use
Clone this git repo and compile using Visual Studio 2015.
Reasons for Deprecation
Because this project was attempting to form the basis of a future port of the compiler to the Adamant language, it took rather involved approaches (i.e. creating a lexer generator). This approach was simply not getting to a working prototype compiler fast enough. Since there is no comparable language available at this time it is important to rapidly test the design of the language and possibly iterate on that design. This project wasn't meeting those needs.
Explanation of this Project
The Adamant compiler is being bootstrapped. To serve this process, a compiler must first be written in another language. In this case C#. This is that temporary compiler. It will be thrown away when the compiler is re-written in Adamant. To facilitate the re-writing in Adamant, this project is structured along lines that make sense for the Adamant compiler. C# was chosen for this because:
- It is a high-level language
- It supports many of the same language features as Adamant including:
- Async Methods
- Covariance and Contravariance (at least for interfaces)
- C# being a garbage collected language means lifetimes will not have to be managed
- I (Jeff Walker) am very familiar with C#
This approach has been adopted after trying to write an Adamant to C# direct translator in C#. That is a translator without type-checking, borrow checking or significant code transformations. Indeed it didn't even build a symbol table. However, this approach was found to be inadequate because there were important language features it just wasn't possible to translate this way (like covariant and contravariant types, type inference etc.). After considering a number of alternatives, the current approach was settled on to minimize wasted work and the time to get to the working front-end of an Adamant compiler. That is a compiler capable of fully parsing and validating Adamant code, though not necessarily being able to compile into any target language/machine code.
Architecting for re-write Guidelines
In order to facilitate the later re-write of the compiler into Adamant the following guidelines should be followed.
- Code as if you were in the Adamant language. For example, make use of the ability to pass
voidas a type parameter and async IO
- Don't attempt to write most Adamant system libraries. Rather, use the closest equivalent in the .NET framework. This includes:
- Tasks (for async)
- Character classes
- Create classes and methods that can be translated to Adamant one-to-one
- Interfaces are written as needed (rather than for every class as one might think is needed to represent Adamant)
- Projects are named after the package they will be re-written to
- Projects that would be under "System." in Adamant are named "Sys." instead to avoid conflicts with the C# standard library
- The "Adamant.Core" project contains classes to bridge the gap between C# and Adamant language features. For example,