make schemas available to on-run-end hooks (#908) #1028
Conversation
| @@ -117,6 +120,7 @@ def compile_node(self, node, manifest): | |||
|
|
|||
| context = dbt.context.runtime.generate( | |||
There was a problem hiding this comment.
couldn't the runtime context generate the list of schemas from the list of nodes? is there a reason to do this in on-run-end hooks based on the list of results, rather than here from the list of nodes?
There was a problem hiding this comment.
If you run with models selected, this way we only get the actually executed models.
beckjake
force-pushed
the
feature/schemas-on-on-run-end
branch
from
8333a88
to
30b6868
Compare
| r.node.schema for r in results | ||
| if not any((r.errored, r.failed, r.skipped)) | ||
| )) | ||
| cls.safe_run_hooks(config, adapter, manifest, RunHookType.End, |
There was a problem hiding this comment.
would it be crazy to just add the results into extra_content here too? I imagine it would be useful for users to iterate over the results and do something intelligent with errors, for example. Any drawbacks to that idea?
There was a problem hiding this comment.
I don't think it would be crazy at all, I wrote it as a dict for pretty much that reason. I'll add it in.
There was a problem hiding this comment.
haha, cool, let's do it :)
Implements #908. In on-run-end hooks (and only in on-run-end hooks!) the value
schemasis now injected into the context. It's a list of unique schemas that dbt successfully executed models/seeds into.The general model here is adding an argument to
compile_node(): a dict of "extra things" to add to the context, and plumbing it through from the ModelRunner when running on-run-end hooks. This injects the value into both the runtime context and the compilation context, though it's the latter that the test case actually cares about.