Question 1

How do you choose the right exception and deal with the exception hierarchy?

Accepted Answer

Raise an exception that precisely expresses the cause of the failure (such as ValueError, KeyError or a domain-specific exception), and catch the narrowest type you can actually handle instead of except Exception everywhere.

Question 2

What is the difference between except, else and finally in a try statement?

Accepted Answer

except handles the matching exception, else runs only when the try block succeeds without an exception, and finally always runs whether the operation succeeded or failed.

Question 3

When should you create a custom exception?

Accepted Answer

When the caller needs to distinguish a specific failure in the domain and make a different decision based on it, or when you need to attach structured data to the error.

Question 4

What is the difference between the EAFP and LBYL styles?

Accepted Answer

EAFP (easier to ask forgiveness than permission) tries the operation directly and handles the exception, while LBYL (look before you leap) checks conditions before acting.

Question 5

What is raise ... from for?

Accepted Answer

It links the new exception to the original cause through __cause__, so the traceback shows the full chain with the phrase "The above exception was the direct cause...", which preserves diagnostic context.

Question 6

How does contextlib help with resource management?

Accepted Answer

It provides tools to build and combine context managers: contextmanager to write a manager as a generator, ExitStack to manage a variable number of resources, suppress to ignore a specific exception, plus closing and nullcontext.

Question 7

How do you log an exception without losing the traceback?

Accepted Answer

Use logger.exception() inside an except block (it logs at ERROR level with the traceback), or pass exc_info=True at any other level, instead of logging only str(err).

Question 8

How do you design timeouts and retries without multiplying the damage?

Accepted Answer

Set a timeout on every I/O operation and retry only on transient errors, with exponential backoff, jitter and a maximum number of attempts, and only for operations that are safe to repeat (idempotent).

Question 9

When do you need ExceptionGroup and except*?

Accepted Answer

When several concurrent operations fail at the same time (such as TaskGroup tasks) and you need to show every error and handle each type, instead of settling for the first error.

Question 10

Where should validation of data coming from outside the system be placed?

Accepted Answer

At trust boundaries: HTTP requests, CLI arguments, files, environment variables and queue messages. Convert the data there into a trusted model (parse, don't validate) and then pass it inward.