fix: BigQueryEngineAdapter.get_table_schema() mypy error

[newtonapple]

This follows the same logic from BigQueryEngineAdapter.columns() to ensure query_job is not None before using it.

[erindru]

Since asserts only work when running python in debug mode, it's generally better to do something like:

if query_job is None:
  raise ...

mypy will still pick up on that and it makes the check explicit

[georgesittas]

I think we have quite a few asserts in the codebase. Is this something we want to be consistent about? Not sure if we've been thinking about them the way you described, but you're right:

In the current implementation, the built-in variable __debug__ is True under normal circumstances, False when optimization is requested (command line option -O). The current code generator emits no code for an assert statement when optimization is requested at compile time

[erindru]

Yeah I realised that we have been using assert incorrectly outside of tests to make the type checker happy.

I learned that validating the assumptions using actual runtime checks:

• Still keeps the type checker happy in most cases
• Forces us to handle the case when something that theoretically shouldn't happen, happens anyway

[newtonapple]

Sure, I can change it to raise a RuntimeError or an explicit AssertionError. But like I said in the comment, I'm simply porting the original logic over. Do you prefer to change the BigQueryEngineAdapter.columns() implementation to raise an exception explicitly as well?

[georgesittas]

Perhaps we can one-shot this more generally in the codebase by asking Claude to do it. Like, "change assertions aiming to satisfy mypy with raise statements" or something.

Or... just do it for the BigQuery adapter and leave the bigger refactor for later :)

[georgesittas]

@newtonapple @erindru I'll go ahead and merge this PR for now to unblock CI in main / other PRs. Let's address the assert vs raise in a followup.