Question

A developer is writing a `parseDuration` function that converts strings like "2h30m" into total seconds. Using Copilot Chat, they provide a few-shot prompt with three examples: "1h" → 3600, "45m" → 2700, "2h30m" → 9000. The generated implementation works for these but crashes on empty strings and inputs like "90m" (which should exceed an hour). What is the most effective refinement to the few-shot prompt to produce more robust code?

Accepted Answer

Add examples that specifically cover the edge and boundary cases, such as "" → 0 and "90m" → 5400, so Copilot infers handling for those situations.

Answer

Increase the number of typical examples from three to ten so Copilot has a larger pattern to generalize from.

Answer

Remove all examples and rely on a zero-shot instruction, since examples are biasing Copilot toward the happy path only.

Answer

Reorder the existing three examples so the most complex one appears first, giving it the strongest influence on the output.