check for node name uniqueness across refable resource types

[drewbanin]

fixes: #736

This came up because a view model and a seed file had the same name. Previously, dbt's check for name uniqueness only pertained to models.

[cmcarthur]

do we have logic like this anywhere else?

[drewbanin]

This code previously existed in the loader, but I ripped it out and put it in the compiler class. I liked the idea of the Loader just returning a flat graph and not validating uniqueness or other business logic constraints. It feels like that will be important if we expose the loader through our API

[drewbanin]

We do something similar here: https://github.com/fishtown-analytics/dbt/blob/development/dbt/parser.py#L290

[drewbanin]

@cmcarthur I accidentally deleted your comment 🙃

You asked if these resources were full paths or just node names.

I think the answer is that it's both, depending where we call this. Will update the code to make it consistent

[drewbanin]

@cmcarthur I cleaned up the code and improved the error message. Now the exception function just takes a couple of nodes. Seems like a better approach. New error:

Compilation Error
  dbt found two resources with the name "idk". Since these resources have the same name,
  dbt will be unable to find the correct resource when ref("idk") is used. To fix this,
  change the name of one of these resources:
  - model.my_project.idk (models/idk.sql)
  - model.my_project.idk (models/tmp/idk.sql)

We do need to check this twice, since the flat map is a dict of unique_id --> node. If we don't check this in the parser, then nodes will be overwritten by a secondary node with a duplicated name. Crucially, two models with the same name in the same package will have equivalent unique_ids.

We need to check a second time to assert uniqueness across projects and resources types.