The .NET Compiler Platform ("Roslyn") provides open-source C# and Visual Basic compilers with rich code analysis APIs.
Guys, I'm working on an analyzer which bans some APIs. So it basically goes through syntaxTree.GetRoot().DescendantNodes().OfType<IdentifierNameSyntax>() and checks INamespaceSymbol namespaceSymbol = semanticModel.GetSymbolInfo(node).Symbol?.ContainingNamespace;.
The function that does this is public IEnumerable<(IdentifierNameSyntax, INamespaceSymbol )> FindForbiddenIdentifiers(SemanticModel semanticModel). My implementation of the method does "yield return (node, @namespace);" so that FindForbiddenIdentifiers returns as soon as possible and a diagnostic can be reported to a Visual Studio user. My colleague would prefer FindForbiddenIdentifiers to return a list (as it is computed faster 240ms as opposed to 300ms with yield return- the intervals are taken from unit tests). Is there some recommendation what is better? I would expect that yield return should be used by default as I understand I don't really need an allocation for the returning list but still my implementation is "slower" in total.
@MartyIX first consider: use of yield return will create a compiler-generated state machine, and you'll allocate that at runtime
the difference likely comes from the fact that the List can utilize a struct enumerator and all the items in the list are pre-computed, as opposed to yield return which lazily evaluates the next item in the sequence for every MoveNext
Hi, I wonder, is it better to have as a rule of thumb:
#1 Dictionary<Key1, Dictionary<Key, Value>>
or
#2 Dictionary <string, Value> where key is string myKey = $"{enumValue1}-{enumValue2}"
?
First variant requires typically more code, the second is typically more comfortable to work with. That's one view. However, I wonder if I keep allocating myKeys again and again or if some string interning is helping me here...?
43 replies
var colRefLocation = AppDomain.CurrentDomain.GetAssemblies().Single(a => a.GetName().Name == "System.Collections").Location;
var stdRefLocation = AppDomain.CurrentDomain.GetAssemblies().Single(a => a.GetName().Name == "netstandard").Location;
var sysrLocation = AppDomain.CurrentDomain.GetAssemblies().Single(a => a.GetName().Name == "System.Runtime").Location;
Hi there, I have a solution that contains 3 projects:
- a library (netstandard 2.0) that is using ISourceGenerator
- a WebAPI project (net5.0) that has some controllers with actions
- a Test (net5.0 with xunit) that has unit tests
Basically the flow I want to achieve is that the SourceGenerator class will analyse all action methods and generate a sample unit test [fact] for each action method.
However what I am realising is that SyntaxTrees only accesses one executing assembly at a time - i.e. if the source generator is being executed on the Test proj (even with WebAPI proj referenced), syntaxtrees for WebAPI are not accessible.
Is there a way around this?
Hi All
I am trying to disable the "SA1101" stylecop warning in .net core
added below line the .ediitorconfig, but still showing the warining
Please help me on this
dotnet_diagnostic.SA1101.severity = none
a
Guys, I have NLog as dependency in my analyzer: https://github.com/dotnet/roslyn/discussions/56643#discussioncomment-1378880.
Now I wonder about this line <PackageReference Include="NLog" Version="4.7.5" PrivateAssets="all" GeneratePathProperty="true" /> in my analyzer project. Is it recommended to add PrivateAssets="all" or not? My fear is that NLog dependency ends up in CUSTOMER's project by transitivity rule. Is that reasonable fear or is it a nonsense?
