Closed mtdowling closed 1 month ago
We could add an excludeDirs
property to the extension, which by default is ${workspaceRoot}/build
, and pass that to the language server as an argument. That way if someone is using an output directory other than build/
, they can also ignore them. We will have to adjust the language server to accept this argument and ignore the directory when searching for model files.
This would be really useful for us. And we are just using the default build folder.
It looks like we're including the build directory when loading the semantic model. We should ignore the build directory since it will have duplicate shapes and could have conflicting shapes when using things like projections. For example, this shows up in the logs:
One negative effect of not filtering out the build directory is that you can get validation warnings that a previously built model conflicts with the model in your IDE. For example, build the model using the CLI or Gradle, then edit a member to target a different shape, and the IDE will tell you that the member in the IDE conflicts with the member in the build output.