rowan

⚠ Experimental. rowan is in development. Expect it to crash on large files and outputs to be at times poorly explained.

rowan is an abstract interpreter for Python: it runs your program on sets of possible values instead of real ones. Hover over any variable to see what it can hold - which literals, which dict keys, which are certain and which only maybe there.

def make_user(name, admin):
    user = {"name": name, "active": True}
    if admin:
        user["role"] = "admin"
    return user

ana = make_user("ana", True)    # hover: {"active": True, "name": "ana", "role": "admin"}
bo = make_user("bo", False)     # hover: {"active": True, "name": "bo"}

Besides python builtins, rowan also has models for pytorch, pandas, pydantic, and a few other libraries. Local imports are resolved but other external dependencies are not analysed.

Files are analysed when opened or saved, or with Rowan: Analyse current file.

Source, documentation and issues: github.com/kaidaniel/rowan.