GitHub Copilot Unit Tests Low Coverage

Built a simple experiment using GitHub Copilot to generate unit tests for a basic Python function – it produced *some* tests, but the coverage was surprisingly low (around 60%) and required a ton of manual tweaking, making me wonder if it's truly helpful for anything beyond the most straightforward code.

▲ 12 upvotes 💬 3 replies ← Back to Community

3 Replies

Marcus Davis @marcus-d · 9 days ago ▲ 5

I found similar results with Pytest’s coverage tool; it seems Copilot’s suggestions often missed edge cases like handling empty lists, resulting in a coverage of 65% after I added some targeted tests.

Priya Rao @priya-r · 9 days ago ▲ 4

I completely agree – I’ve found that CodeClimate’s branch coverage analysis consistently reveals similar low test coverage when Copilot generates tests for complex functions, often missing critical edge cases.

Alex Johnson @alex-j · 9 days ago ▲ 3

I found similar results with Tabnine’s unit test generation – it yielded around 55% coverage in a comparable Python function, highlighting the need for careful human review and expansion even with these tools.

Join the discussion

Join Community →