Skip to content

Use case: test impact analysis and flaky test history #29

Closed

Description

@adityamukho

Summary

Add a use case: store which tests cover which code, plus every test result per commit.

The idea

Record coverage edges (test covers function or file) and test outcomes for each CI run. Large CI systems build this in-house. A small team could keep it in one file.

Questions it answers

  • Which tests must run for this change? (reverse reachability from the changed files)
  • Is this test flaky, or did it start failing at a real commit?
  • Which code areas have no covering test at a given release?

Why Minigraf fits

  • Test selection is a graph traversal from changed code to covering tests, including through call edges.
  • Telling flaky from broken needs result history across commits.

Suggested content

  • The idea and the two questions (selection and flakiness)
  • A fact schema (test, covers, calls, result, commit)
  • A rule for test selection from a set of changed files
  • A query that finds tests with mixed results on the same commit

Notes


Generated by Claude Code

Activity

  1. adityamukho commented on Oct 19, 2026

    @adityamukho
    ContributorAuthor

    Builds on the git sidecar idea in #34, which provides the commit and entity index this use case attaches facts to.


    Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentation

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions