opentia 1.2.5

Test Impact Analysis for C# / .NET, Java, Android (Kotlin), and Node.js — selects only tests affected by a git diff. Polyglot and monorepo aware.

opentia — Test Impact Analysis

Analyses a git diff and selects only the tests whose execution path could have been affected by the change. Skips the full suite on every push.

No external dependencies — Python 3.8+ stdlib only.

Language support: C# / .NET, Java (Maven / Gradle), Android (Kotlin, nested Gradle modules, instrumented tests), and Node.js (Jest / Vitest / npm test scripts — single packages, npm/pnpm/lerna workspaces). Mixed-language repos are handled in a single run: changes are routed to the right ecosystem automatically.

---

Requirements

Tool	Version	Purpose
Python	3.8+	Run opentia
Git	any	Diff source

---

Installation

pip install opentia

This installs the opentia command on your PATH.

---

Quick start

# Analyse the last commit
opentia --base HEAD~1 --root <path-to-your-project>

# Analyse uncommitted (staged + unstaged) changes — no commit needed
opentia --unstaged --root <path-to-your-project>

# Analyse and immediately run the selected tests
opentia --base HEAD~1 --root <path-to-your-project> --run

--root is where your .sln / .csproj / pom.xml / build.gradle / package.json lives. It does not need to be the git root — opentia locates the actual git root automatically.

---

Usage reference

opentia [OPTIONS] [-- TEST_ARGS]

  --base REF      Git ref to diff against (e.g. HEAD~1, main, origin/main)
  --head REF      Head ref to diff from (default: HEAD)
  --root DIR      Directory containing project files (default: .)
  --lang LANG     Force one adapter: dotnet | java | node (default: auto-detect all)
  --strategy      project | convention | symbol | hybrid (default: hybrid)
  --output, -o    human | json | github-actions | azure-devops (default: human)
  --run           Execute the test command after analysis
  --unstaged      Analyse working-tree changes (staged + unstaged)
  --staged        Analyse only staged changes — useful before committing
  --              Everything after this is forwarded to the test runner

Output formats

# Human-readable (default)
opentia --base HEAD~1 --root .

# JSON — pipe into scripts or CI steps
opentia --base HEAD~1 --root . --output json

# GitHub Actions
opentia --base HEAD~1 --root . --output github-actions

# Azure DevOps
opentia --base HEAD~1 --root . --output azure-devops

JSON output fields

{
  "run_all": false, // true = targeted selection was abandoned
  "language": "dotnet", // dotnet | java | node
  "test_filter": "FullyQualifiedName~PricingServiceTests",
  "test_project_paths": ["...SampleApp.Services.Tests.csproj"],
  "affected_test_projects": ["SampleApp.Services.Tests"],
  "affected_test_classes": ["PricingServiceTests"],
  "test_command": "dotnet test \"...\" --filter \"...\" --logger trx",
  "test_result_paths": [".../TestResults/*.trx"], // result globs for CI
  "reason": "Analysis complete",
  "strategy_notes": [], // warnings / fallback explanations
}

When changes span multiple ecosystems in one run, the same fields are emitted merged at the top level (language: "java,node", test_command joined with &&) plus a results array containing one full per-language object each.

Test results & artifacts

test_result_paths lists posix-style globs where the runner writes result files, so CI can upload/publish exactly the tests that ran:

Runner	Emits results by default?	Glob
Gradle (incl. Android)	yes	<module>/build/test-results/<task>/TEST-*.xml — variant-aware (testDebugUnitTest for Android unit, test for plain JVM); build/outputs/androidTest-results/ for instrumented
Maven	yes	<module>/target/surefire-reports/TEST-*.xml
.NET	with --logger	opentia defaults the command to --logger trx (built-in), so <project>/TestResults/*.trx is produced; pass your own -- --logger ... to override
Node (Jest/Vitest)	no	best-effort common locations only — JUnit XML requires a reporter you configure (jest-junit, or vitest --reporter=junit --outputFile=)

---

Node.js projects

opentia detects Jest and Vitest automatically. For workspaces (npm workspaces, pnpm-workspace.yaml, or lerna.json), each sub-package is analysed independently and the dependency graph is resolved across internal references — workspace:*, file:, and plain version ranges that match a sibling package name. Changing a shared package triggers tests in every package that depends on it.

The test command is a single npx jest (or npx vitest run) invocation run from the workspace root with a test-path-pattern filter. Packages without jest/vitest but with a test script (karma, ng test, mocha-via-script) fall back to npm test --prefix <package> — selection stays package-accurate, but those packages run their full suite.

Android projects

Android repos are handled by the Gradle adapter with Kotlin-aware analysis (Kotlin types and functions are public by default — no modifier needed for symbol matching). Per module:

• Unit tests (src/test) run via ./gradlew :path:to:module:test --tests=...; instrumented tests (src/androidTest) are selected separately and routed to :path:to:module:connectedAndroidTest (device/emulator required).
• Nested module references (project(":core:model")) resolve by path, and each module's command carries only its own test classes — a --tests pattern matching nothing would fail the task.
• gradle/libs.versions.toml is workspace-level INFRA (full run); local.properties, keystores (.jks/.keystore), build/ output, and hidden tooling dirs (.github/, .claude/, …) are ignored; proguard-rules.pro scopes to its owning module.
• Method-level narrowing only applies when every changed method in a test class is @Test-annotated; a changed helper widens to the whole class.

Mixed-language monorepos

A single run covers every ecosystem under --root. Each changed file is routed to the adapter owning its nearest build file (.csproj/.sln, pom.xml/build.gradle, package.json), so a fullstack repo — say a Maven backend with an Angular frontend — selects backend tests for .java changes and frontend specs for .ts changes in one invocation:

opentia --base HEAD~1 --root .   # all ecosystems, one combined result

The combined test_command chains each runner with &&. To restrict analysis to one ecosystem, pass --lang dotnet|java|node.

Two more monorepo behaviours worth knowing:

• Changes outside --root are ignored (reported in strategy_notes) rather than triggering a full run of the app you pointed at.
• Module-level build files (a leaf .csproj, a module pom.xml, a workspace package's package.json) scope through the dependency graph like any other change to that project. Only workspace-level files (.sln, root/parent pom.xml, settings.gradle, root package.json, lockfile-style global config) force a full run.

---

CI integration

GitHub Actions — pull request

on:
  pull_request:
    branches: [main, staging]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: Test Impact Analysis
        id: tia
        run: opentia --base ${{ github.event.pull_request.base.sha }} --root . --output github-actions

      - name: Run affected tests
        if: steps.tia.outputs.has_tests == 'true'
        env:
          TEST_COMMAND: ${{ steps.tia.outputs.test_command }}
        run: bash -c "$TEST_COMMAND"

      # Publish only the results of the tests that ran (test_result_paths is a
      # newline-separated list of globs — drops straight into upload-artifact).
      - name: Upload test results
        if: always() && steps.tia.outputs.has_tests == 'true'
        uses: actions/upload-artifact@v4
        with:
          name: test-results
          path: ${{ steps.tia.outputs.test_result_paths }}

      - name: Publish test report
        if: always() && steps.tia.outputs.has_tests == 'true'
        uses: EnricoMi/publish-unit-test-result-action@v2
        with:
          files: ${{ steps.tia.outputs.test_result_paths }}

Available outputs: test_filter, run_all, has_tests, test_project_paths, test_command, test_result_paths.

Security: pass test_command through an env: variable and run bash -c "$TEST_COMMAND", rather than interpolating ${{ … }} directly into run:. opentia shell-quotes the command components (so the string is safe to evaluate in a POSIX shell), but textual ${{ }} interpolation pastes the value into the script before the shell parses it — the env-indirection keeps a repo-controlled file or directory name from ever being re-parsed as workflow/shell syntax. The emitted command targets a POSIX shell (bash); on Windows runners, set shell: bash.

GitHub Actions — push to branch

- name: Test Impact Analysis
  id: tia
  run: opentia --base ${{ github.event.before }} --root . --output github-actions

- name: Run affected tests
  if: steps.tia.outputs.has_tests == 'true'
  env:
    TEST_COMMAND: ${{ steps.tia.outputs.test_command }}
  run: bash -c "$TEST_COMMAND"

Azure DevOps

- script: opentia --base $(System.PullRequest.TargetBranchName) --root . --output azure-devops
  displayName: Test Impact Analysis

- script: $(testCommand)
  condition: eq(variables['hasTests'], 'true')
  displayName: Run affected tests

- task: PublishTestResults@2
  condition: and(always(), eq(variables['hasTests'], 'true'))
  inputs:
    testResultsFormat: JUnit # use 'VSTest' for .NET (.trx)
    testResultsFiles: $(testResultPaths) # comma-separated globs

Available variables: testFilter, runAllTests, hasTests, testProjectPaths, testCommand, testResultPaths. .