API to get type-annotated AST

Is there an existing API, that would let me programmatically analyze my Python project?

If there’s not one yet, I could work on one if only team can point which classes could provide such a thing, and, if possible, what to do to have a minimal setup to get them up and running on a specific Python module.

To give some background: I was already able to get PythonInterpreterFactoryWithDatabase.GenerateDatabase complete successfully from a console app (exitCode == 0) on my Python 3.6 installation without passing path to the module I want to analyze yet.

[zooba]

Right now it exists, but we are in the process of reworking it significantly and the existing API is likely to become unsupported in favour of running our analyzer and using the JSON protocol (which will eventually be the language server protocol).

I can’t stop right now to do a full write-up of the current API, unfortunately (and I can’t find any of our existing ones), but if you look through the AnalysisTests project you’ll see plenty of examples where we analyze small snippets of code and examine the results. You should be able to do the same thing from your own application requiring only Microsoft.PythonTools.Analysis.dll and Microsoft.PythonTools.Common.dll (and if your application is a VS extension, then add a prerequisite for Python support and you won’t even need to install those DLLs).

[zooba]

Ah, I see. We rely on SourceLocation.Index to be set correctly when using it to resolve the lookup scope, but we don’t actually make the index available anywhere - we have private methods on the AST to figure it out.

I think the problem is leaking out the LocationInfo class, which seems to be meant to be internal only. SourceLocation should be the only public one (the difference is that LocationInfo knows which file the location is within, and has internal mechanisms for lazy resolution).

We do have a likely big breaking change coming up (#2846) when we can also make LocationInfo private and update the other APIs to return and accept SourceLocation (since in most cases the file they belong to is known). Looking at this use case, we probably also need to add a better public API for requesting specific locations (e.g. location of name, location within scope, etc.).

The best way I can see to get the information you need is to use an AST walker (PythonWalker subclass and ast.Walk(walker)) to find the function, rather than the ModuleAnalysis APIs. This will give you access to the FunctionDefinition object, which has the SourceLocation values you need in order to query analysis in a particular scope. Our QualifiedFunctionNameWalker is a fairly simple example of locating a specific function and generating its full name (though in most cases you’ll want to override Walk rather than PostWalk). If you provide the ModuleAnalysis object when constructing your walker, you can call into it as you walk each function to get whatever info you need.

Hope that helps!