dealing with empty context names

[techtonik]

Just run into the same issue as #796.

I'm always frustrated when I don't understand what's happening. Although my guess was the same - that (empty) context is an import, I thought I open another issue to clarify how to deal with that.

Describe the solution you'd like

1. The most basic thing that could help is to name the context as (import) or {import}, so that people won't have to guess.

2. Then it could be the top level file where import is coming from as {import}: filename

3. Then it could be the import stack {import}: filename1 -> filename2

Describe alternatives you've considered

The alternative is to display "import stack" in Python flat "import space" rather than "filesystem space". But it is more practical to know the filename rather than {level1}.somethings.{lamdba15}.something internal address.

Perhaps it is possible to store both.

Additional context

Avoid using word "context" as it is too abstract as the title above says. Once you get the details, it is easy, but the path through abstractions is not. Give more examples and database queries.

Also, add foreign keys to allow diagram tools automatically link tables?

---

Forgot to add - that stuff is super awesome. =)

[nedbat]

The empty context might be more than just "imports". It's any code run before an explicit context has been set. When using pytest test contexts, it's all the code run by pytest before a test begins.

I understand that "(empty)" isn't very helpful, but I'm not sure what else I can use as a label in the absence of any information.

[nedbat]

Good idea about the foreign keys. I've added those on master.

[techtonik]

In case of pytest - does that include the code before coverage is initialized? I am right that the process is the following.

1. pytest runs and discovers tests - no coverage
2. pytest initializes coverage
3. run each test

So, any pytest code that is run after coverage initialization is also a part of report? I thought that there is mechanism to exclude those files from coverage.

[nedbat]

It depends how you are running coverage. First, you might not be using pytest at all. Second, you might be running pytest under coverage ("coverage run -m pytest"). In this case, all of pytest is being measured, even the test discovery. If you run coverage from within the pytest plugin ("--cov"), I'm not sure when coverage measurement starts.