Remove the duplicate get_context_modules

[Fokko]

While looking at the code, and figuring out how to inject custom modules, I've stumbled upon some duplicate code. Besides that, I also took the liberty of adding annotations.

[michael-erasmus]

Wow, that's crazy @Fokko, I was just preparing a PR to do the exact same thing as you (add the ability to inject custom modules), also came across this duplicated code and then I spotted your PR. Do you have a follow-up PR for injecting modules? If not, I'll merge your code first to do my PR :)

[Fokko]

Good to see that I'm not the only one :-) I'd like to do some counts, and send them to an external system.

I'm currently exploring two directions:

• Inject the function in some messy way:

def hello():
    return 'world'

dbt.context.base.CONTEXT_MODULES['custom'] = {
    'my_custom_function': hello
}

dbt.main()

• Do it using a Spark UDF. But I don't like this, because then we also have to distribute credentials to Spark.

[beckjake]

Thanks for the PR @Fokko - Funny story, I actually have a WIP branch that also fixes this, but it's not ready to go for unrelated reasons.

I'm happy to merge this standalone, some very minor feedback before I kick off the tests and we get this merged :)

[beckjake]

Can you keep this as a global function that returns a dict, instead of a global dict? I have a good reason and a bad/lazy reason:

• I don't want us to share the same dict into every jinja environment. I think jinja does a shallow copy of the context and I want to build a fresh one every time we generate a context, since dicts are mutable.
• I want to minimize merge conflicts with my other branch that makes the same change :)

[Fokko]

Sure, I've reverted the change.

[Fokko]

I've stopped pursuing implementing a custom Python function, and went with https://github.com/fishtown-analytics/dbt-event-logging/ instead. This does everything I need for now.

[beckjake]

I've kicked off the test suite, I think this is good to go. Thanks for your contribution to dbt!

[beckjake]

@Fokko sorry about this, but mypy is failing, I think before this function wasn't annotated so it wasn't inspected. Can you add a # type: ignore to the line that accesses pytz.__all__? That attribute does exist...

[Fokko]

@beckjake Sure, I've added the exception.