Original intention was to register Settings automatically for DI. This ran into problem when the container verifies the configuration is valid for a settings using the constructor for initialization. It tries to resolve the parameters and fails.
This removes the automatic registration and falls back ActivatorCreateInstance when no Setting is registered.
Probably something that only bothers me, but I'm in a habit of running dotnet test at the solution level. It was discovering this project as a test project even though it isn't. This produces an error complaining about not having a testhost.dll. I'm pretty sure this comes in via a props when xunit is referenced.
This attribute in the csproj marks the project explicitly as not a test project
Wanted a command to break down how Spectre views the command model. This outputs all the relevant settings in a tree. You can get a short explanation for all the commands, or a detailed explanation for all commands (there is a flag to go vice versa on the detailed view if you want)
app cli explain
app cli explain myappcommand
app cli explain -d
The RegEx runtime perf was never anything noticeable - it was the startup time that was eating over a third of time during initialization.
This shaves 200ms off the startup time.
These methods don't require a string.format call so we'll directly call the Render method without a call to string.format.
Added bonus of a a couple fewer allocations too.
Makes the Report method an explicit implementation to allow for better interoperability with standard .NET progress functionality while keeping backwards compatibility with existing ProgressTask functionality.
Closes#285
Previous behavior was that the only way to get a task to a finished state was to artificially set the Value to MaxValue.
With this change StopTask() will also complete the task with the change that a task cannot be restarted.
In both of these loops context is captured preventing caching of the lambda. this results in a pretty significant amount of allocations especially with progress bars that constantly are remeasuring
R# and Rider have quite a bit of noise related to documentation in the testing projects so this disables those warnings.
In the main projects, R# and Rider complain loudly about the namespaces not matching the file structure. The DotSettings file disables that warning.
Once you get rid of that noise there are quite a few opportunities for trimming out redundant code that R# points out especially with the nullable support enabled, plus there are some bugs related to multiple enumerations worth looking into I think.
