AI for Pruning Bloated Snapshot Test Suites

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-ai-coding-AI-snapshot-test-pruning-creators

1. When using an LLM to prune a snapshot test suite, what is the primary input the model needs?

   1. The compiled JavaScript bundle and build logs
   2. The entire project's node_modules directory
   3. The snapshot files along with the components they cover
   4. Only the test file names and their line counts

2. A snapshot test captures a timestamp that changes every time the test runs. How should this be classified?

   1. keep
   2. delete-noise
   3. mark-as-flaky
   4. replace-with-behavior-test

3. What is a key limitation of using AI to prune snapshot tests?

   1. AI cannot parse JavaScript syntax
   2. AI cannot identify duplicate snapshots
   3. AI cannot read JSON files
   4. AI cannot determine which UI states actually matter to users

4. An LLM finds two different snapshot files that contain identical UI output for the same component. What should it recommend?

   1. Keep both snapshots since redundancy improves coverage
   2. Replace both with a single behavior test
   3. Mark both for manual review only
   4. Delete one as a duplicate and keep the other

5. What type of test would an LLM likely suggest as a replacement for a snapshot that captures only dynamic IDs?

   1. An integration test
   2. A visual regression test
   3. A behavior-driven unit test
   4. A performance benchmark

6. Why is it important to avoid mixing snapshot deletions with behavior changes in a single pull request?

   1. Version control systems cannot handle mixed changes
   2. Behavior tests require different test runners
   3. The review becomes too complex to understand
   4. Mixed changes trigger too many CI pipeline failures

7. What information should accompany each snapshot classification verdict from an LLM?

   1. The test run duration in milliseconds
   2. The Git blame author information
   3. The exact file size in bytes
   4. A one-line reason explaining the classification

8. After an LLM proposes snapshot deletions, what must a human developer still do?

   1. Rewrite the entire test suite from scratch
   2. Delete the snapshot files without review
   3. Run the tests to confirm coverage is maintained
   4. Compile the TypeScript types

9. In the pruning workflow, what does the LLM use to group its recommendations?

   1. By component coverage
   2. By file creation date
   3. By Git branch name
   4. By test execution time

10. What characteristic makes a snapshot a candidate for deletion rather than replacement?

    1. It is located in a deeply nested directory
    2. It was created more than six months ago
    3. It uses the old Jest snapshot format
    4. It captures mostly noise like dates, IDs, or hashes

11. Which of these is explicitly listed as something AI cannot do when pruning snapshot tests?

    1. Parse JSON snapshot files
    2. Replace human design intent
    3. Generate new test code
    4. Identify syntactically invalid snapshots

12. What tool is specifically mentioned as an example of a snapshot testing framework?

    1. Jest
    2. Mocha
    3. Cypress
    4. Chai

13. When should a snapshot be classified as 'replace-with-behavior-test' rather than deleted?

    1. When the test has never failed
    2. When the snapshot file is larger than 1MB
    3. When the component being tested is user-facing and important
    4. When the snapshot was created in the current sprint

14. What risk exists if you rely solely on AI to determine which snapshots to keep?

    1. Important user-facing component snapshots might be incorrectly deleted
    2. The AI might use too much memory
    3. The test files might become encrypted
    4. Build times might increase

15. What does the 'verdict' in the LLM pruning workflow represent?

    1. A vote from the development team
    2. A security audit result
    3. A legal ruling about test ownership
    4. The final classification decision: keep, delete-noise, or replace-with-behavior-test