High-level overview of changes in the .NET Core tools
This document describes the changes associated with moving from project.json to MSBuild and the csproj project system with information on the changes to the layering of the .NET Core tooling and the implementation of the CLI commands. These changes occurred with the release of .NET Core SDK 1.0 and Visual Studio 2017 on March 7, 2017 (see the announcement) but were initially implemented with the release of the .NET Core SDK Preview 3.
Moving away from project.json
The biggest change in the tooling for .NET Core is certainly the move away from project.json to csproj as the project system. The latest versions of the command-line tools don't support project.json files. That means that it cannot be used to build, run or publish project.json based applications and libraries. In order to use this version of the tools, you will need to migrate your existing projects or start new ones.
As part of this move, the custom build engine that was developed to build project.json projects was replaced with a mature and fully capable build engine called MSBuild. MSBuild is a well-known engine in the .NET community, since it has been a key technology since the platform's first release. Of course, because it needs to build .NET Core applications, MSBuild has been ported to .NET Core and can be used on any platform that .NET Core runs on. One of the main promises of .NET Core is that of a cross-platform development stack, and we have made sure that this move does not break that promise.
If you are new to MSBuild and would like to learn more about it, you can start by reading the MSBuild Concepts article.
The tooling layers
With the move away from the existing project system as well as with building engine switches, the question that naturally follows is do any of these changes change the overall "layering" of the whole .NET Core tooling ecosystem? Are there new bits and components?
Let's start with a quick refresher on Preview 2 layering as shown in the following picture:
The layering of the tools is quite simple. At the bottom we have the .NET Core Command Line tools as a foundation. All other, higher-level tools such as Visual Studio or Visual Studio Code, depend and rely on the CLI to build projects, restore dependencies and so on. This meant that, for example, if Visual Studio wanted to perform a restore operation, it would call into
dotnet restore (see note) command in the CLI.
With the move to the new project system, the previous diagram changes:
The main difference is that the CLI is not the foundational layer anymore; this role is now filled by the "shared SDK component". This shared SDK component is a set of targets and associated tasks that are responsible for compiling your code, publishing it, packing NuGet packages etc. The SDK itself is open-source and is available on GitHub on the SDK repo.
A "target" is a MSBuild term that indicates a named operation that MSBuild can invoke. It is usually coupled with one or more tasks that execute some logic that the target is supposed to do. MSBuild supports many ready-made targets such as
Execute; it also allows users to write their own tasks using managed code and define targets to execute those tasks. For more information, see MSBuild tasks.
All the toolsets now consume the shared SDK component and its targets, CLI included. For example, the next version of Visual Studio will not call into
dotnet restore (see note) command to restore dependencies for .NET Core projects, it will use the "Restore" target directly. Since these are MSBuild targets, you can also use raw MSBuild to execute them using the dotnet msbuild command.
The shared SDK component means that the majority of existing CLI commands have been re-implemented as MSBuild tasks and targets. What does this mean for the CLI commands and your usage of the toolset?
From an usage perspective, it doesn't change the way you use the CLI. The CLI still has the core commands that exist in Preview 2 release:
These commands still do what you expect them to do (new up a project, build it, publish it, pack it and so on). Majority of the options are not changed, and are still there, and you can consult either the commands' help screens (using
dotnet <command> --help) or documentation on this site to get familiar with any changes.
From an execution perspective, the CLI commands will take their parameters and construct a call to "raw" MSBuild that will set the needed properties and run the desired target. To better illustrate this, consider the following command:
dotnet publish -o pub -c Release
This command is publishing an application into a
pub folder using the "Release" configuration. Internally, this command gets translated into the following MSBuild invocation:
dotnet msbuild -t:Publish -p:OutputPath=pub -p:Configuration=Release
The notable exception to this rule are
run commands, as they have not been implemented as MSBuild targets.
Starting with .NET Core 2.0 SDK, you don't have to run
dotnet restore because it's run implicitly by all commands that require a restore to occur, such as
dotnet build and
It's still a valid command in certain scenarios where doing an explicit restore makes sense, such as continuous integration builds in Azure DevOps Services or in build systems that need to explicitly control the time at which the restore occurs.