Python: Better allow import * to work with API graphs

[RasmusWL]

I'm curious to see what people think of this solution :D

Performance impact

I did some testing on a random DB I had locally. On this particular DB the change in number of LocalSourceNodes and API graph edges was insignificant:

• local source nodes: 390,311 -> 390,411
• API graph edge: 364,762 -> 366,994

However, that's not to say it couldn't impact things heavily on other DBs. We'll see what DCA says 👍

[yoff]

Do I understand it correctly that all instances of an imported value will be considered local sources?

[tausbn]

Overall, this looks good to me (but I do have one specific comment).

I feel like this is a fairly principled solution (our implementation of flow in the global scope is a bit weird), and I don't think it would add a lot of extra nodes in general.

[RasmusWL]

Do I understand it correctly that all instances of an imported value will be considered local sources?

Yes, good point @yoff. In the example below, all three occurrences of Bar will be considered a LocalSourceNode 🤷

from foo import *

print(Bar)
print(Bar)
print(Bar)

[RasmusWL]

rebased to drop the commit to use API::Node in the predicates, to preserve old git blame info

[tausbn]

Do I understand it correctly that all instances of an imported value will be considered local sources?

Yes, good point @yoff. In the example below, all three occurrences of Bar will be considered a LocalSourceNode 🤷

from foo import *

print(Bar)
print(Bar)
print(Bar)

I think this is somewhat unavoidable due to how we do global flow. Right now, we have no way of knowing if the three instances of Bar are the same (imagine print(Bar); modify_Bar(); print(Bar), where the middle function changes the value of Bar), so this is the best we can do currently.

[yoff]

LGTM