experiment.yml

name: Mutation Testing Experiment

on:
  workflow_dispatch:
    inputs:
      packages:
        description: "JSON file that specifies packages to generate mutants for"
        default: "benchmarks.json"
      template:
        description: "Template to use for generating mutants"
        default: "template-full"
      systemPrompt:
        description: "System prompt to use when obtaining completions"
        default: "SystemPrompt-MutationTestingExpert"
      temperature:
        type: number
        description: "Sampling temperature to try when obtaining completions"
        default: 0.0
      model:
        description: "Which LLM API to use"
        type: choice
        options:
          - "gpt-4o-mini"
          - "gpt-3.5-turbo"
          - "gpt-4-turbo"
          - "meta-llama/llama-3.3-70b-instruct"
        default: "gpt-4o-mini"
      # rateLimit:
      #     description: "number of milliseconds between requests sent to the LLM"
      #     type: number
      #     default: 0
      benchmarkMode:
          description: "use custom rate limiting for benchmarking"
          type: "boolean"
          default: false
      maxNrPrompts:
          type: number
          description: "maximum number of prompts to produce"
          default: 2000
      maxTokensInCompletion:
          type: number
          description: "maximum number of tokens in a completion"
          default: 250
      strykerOptions:
        description: "stryker options (e.g., --concurrency 1) to pass to stryker"
        default: "--concurrency 1"
      debug_enabled:
        type: boolean
        description: "Run the build with tmate debugging enabled (https://github.com/marketplace/actions/debugging-with-tmate)"
        default: false

jobs:
  setup:
    runs-on: ubuntu-latest
    outputs:
      packages: "${{ steps.parse_packages.outputs.packages }}"
      temperatures: "${{ github.event.inputs.temperature || '0.0' }}"
      maxNrPrompts: "${{ github.event.inputs.maxNrPrompts || '2000' }}"
      strykerOptions: "${{ github.event.inputs.strykerOptions }}"
    steps:
      - uses: actions/checkout@v3

      - uses: actions/setup-node@v3
        with:
          node-version: 20

      - id: parse_packages
        run: |
          packages=$(node \
                     .github/find_benchmarks.js " \
                     .github/${{ github.event.inputs.packages || 'benchmarks.json' }}")
          packages=$(echo $packages | tr '\n' ' ')
          echo "packages=$packages" >> $GITHUB_OUTPUT

  benchmark:
    needs:
      - setup
    runs-on: ubuntu-latest
          
    continue-on-error: true
    strategy:
      fail-fast: false
      matrix:
        package: ${{ fromJson(needs.setup.outputs.packages) }}
    steps:
      - name: checkout master branch of StrykerJS from https://github.com/neu-se/stryker-js into directory stryker-js
        uses: actions/checkout@v3
        with:
          repository: neu-se/stryker-js
          ref: master
          path: stryker-js

      - name: build StrykerJS
        run: |
          cd stryker-js
          npm install
          npm run build

      - name: print package info
        run: |
          echo "package.host=${{ matrix.package.host }}" 
          echo "package.name=${{ matrix.package.name }}"
          echo "package.owner=${{ matrix.package.owner }}"
          echo "package.repo=${{ matrix.package.repo }}"
          echo "package.sha=${{ matrix.package.sha }}"
          echo "package.edits=${{ matrix.package.edits }}"
          echo "package.files=${{ matrix.package.files }}"
          echo "package.ignore=${{ matrix.package.ignore }}"

          # print the various options
          echo "temperature=${{ needs.setup.outputs.temperatures }}"
          echo "template=${{ github.event.inputs.template }}"
          echo "systemPrompt=${{ github.event.inputs.systemPrompt }}"
          echo "model=${{ github.event.inputs.model }}"
          echo "benchmarkMode=${{ github.event.inputs.benchmarkMode }}"
          echo "maxNrPrompts=${{ needs.setup.outputs.maxNrPrompts }}"
          echo "maxTokensInCompletion=${{ github.event.inputs.maxTokensInCompletion }}"
          echo "strykerOptions=${{ needs.setup.outputs.strykerOptions }}"

      - name: check out benchmark
        if: ${{ matrix.package.host == 'github.com' }}
        uses: actions/checkout@v3
        with:
          repository: ${{ format('{0}/{1}', matrix.package.owner, matrix.package.repo) }}
          ref: ${{ matrix.package.sha }}
          path: ${{ matrix.package.name }}

      - name: Checkout gitlab package repo
        if: ${{ matrix.package.host == 'gitlab.com' }}
        run: |
          git clone ${{ format('https://gitlab.com/{0}/{1}', matrix.package.owner, matrix.package.repo) }} ${{ matrix.package.name }}
          cd ${{ matrix.package.name }}
          git checkout ${{ matrix.package.sha }}

      - name: Setup tmate session
        uses: mxschmitt/action-tmate@v3
        if: ${{ github.event_name == 'workflow_dispatch' && inputs.debug_enabled }}

      - name: build project
        run: |
          cd ${{ matrix.package.name }}
          # if an edit command for editing package.json is specified (e.g., to disable linting), run it
          if [ -n "$MATRIX_PACKAGE_EDITS" ]; then
            ${{ matrix.package.edits }}
          fi
          npm install
          # if a build script exists, run it
          npm run build || echo "No build script found"
        env:
          MATRIX_PACKAGE_EDITS: ${{ matrix.package.edits }}

      - name: Check out llm-mutation-testing
        uses: actions/checkout@v3
        with:
          path: llm-mutation-testing 

      - name: Set up llm-mutation-testing
        run: |
          cd llm-mutation-testing
          npm run build

      - name: Generate mutants
        env:
          LLMORPHEUS_LLM_API_ENDPOINT: '${{ secrets.OPENROUTER_LLM_API_ENDPOINT }}'
          LLMORPHEUS_LLM_AUTH_HEADERS: '${{ secrets.OPENROUTER_LLM_AUTH_HEADERS }}'
        run: |
          cd ${{ matrix.package.name }}
          BENCHMARK_DIR=`pwd`
          cd ../llm-mutation-testing
          echo "Generating mutants for ${{ matrix.package.name }}"           
          LLMORPHEUS_OPTIONS="--temperature ${{ needs.setup.outputs.temperatures }} --maxNrPrompts ${{ needs.setup.outputs.maxNrPrompts }} --systemPrompt ${{ github.event.inputs.systemPrompt }}.txt --model ${{ github.event.inputs.model }} --benchmark ${{ github.event.inputs.benchmarkMode }} --maxTokens ${{ github.event.inputs.maxTokensInCompletion }}"
          (time node --max-old-space-size=6144 benchmark/createMutants.js --path $BENCHMARK_DIR --mutate "${{ matrix.package.files }}" --ignore "${{ matrix.package.ignore }}"  --template templates/${{ github.event.inputs.template }}.hb  --caching false  $LLMORPHEUS_OPTIONS)  2>&1 | tee -a ../${{ matrix.package.name }}/LLMorpheusOutput.txt

      - name: extract mutant-related info
        run: |
          cd ${{ matrix.package.name }}
          mkdir ${{ matrix.package.name }}
          mkdir ${{ matrix.package.name }}/${{ matrix.package.name }}
          cp -r MUTATION_TESTING/*/* ${{ matrix.package.name }}/${{ matrix.package.name }}
          cp LLMorpheusOutput.txt ${{ matrix.package.name }}/${{ matrix.package.name }}/LLMorpheusOutput.txt

      - name: Upload mutants info
        uses: actions/upload-artifact@v4
        with:
          name: mutants-${{ matrix.package.name }}
          path: ${{ matrix.package.name }}/${{ matrix.package.name }}

      - name: install stryker-js into benchmark and run stryker (precomputed mutators)
        run: |
          cd ${{ matrix.package.name }}
          npm install install-local # install-local is needed to install our custom version of stryker-js  
          npx install-local ../stryker-js/packages/{core,util,api,instrumenter,*-runner}  --legacy-peer-deps
          export MUTANTS_FILE=MUTATION_TESTING/${{ github.event.inputs.template }}_${{ github.event.inputs.model }}_${{ github.event.inputs.temperature }}/mutants.json 
          STRYKER_FILES=$(node ../llm-mutation-testing/.github/expandGlob.js $(pwd) "${{ matrix.package.files }}" "${{ matrix.package.ignore }}")
          STRYKER_OPTIONS="${{ needs.setup.outputs.strykerOptions }}"
          (time npx stryker run $STRYKER_OPTIONS --usePrecomputed  --mutate $STRYKER_FILES)  2>&1 | tee -a StrykerOutput.txt

      - name: extract summary from StrykerOutput.txt
        run: |
          cd ${{ matrix.package.name }}
          node ../llm-mutation-testing/.github/parseStrykerReport.js StrykerOutput.txt    

      - name: gather reports/mutation/mutation.*, StrykerOutput.txt, and StrykerInfo.json into results.zip
        run: |
          cd ${{ matrix.package.name }}
          mkdir results
          mkdir results/results-${{ matrix.package.name }}
          cp reports/mutation/mutation.* results/results-${{ matrix.package.name }}/
          cp StrykerOutput.txt results/results-${{ matrix.package.name }}/StrykerOutput.txt
          cp StrykerInfo.json results/results-${{ matrix.package.name }}/StrykerInfo.json

      - name: Upload artifacts
        uses: actions/upload-artifact@v4
        with:
          name: results-${{ matrix.package.name }}
          path: ${{ matrix.package.name }}/results

  combine_output:
    name: Combine output from all benchmarks
    needs:
      - setup
      - benchmark
    runs-on: ubuntu-latest
    steps:
      - name: Download output zips
        uses: actions/download-artifact@v4 
      - name: Combine outputs
        run: |
          mkdir results
          mkdir results/results
          cd results/results

          # move all benchmark info into a single directory
          for benchmark in ../../results-*
          do
             # extract benchmark name
             name=$(echo $benchmark | sed 's/..\/..\/results-//')
             mv $benchmark/results-$name $name
          done
          cd ../..

      - name: Upload combined output files
        uses: actions/upload-artifact@v4
        with:
          name: results
          path: results/results

  generate_report:
    name: Generate report
    needs: 
      - setup
      - combine_output
    runs-on: ubuntu-latest

    steps:
       

       # download all files mutants-*   
       - name: Download LLMorpheus results
         uses: actions/download-artifact@v4
         with:
            pattern: "mutants-*"

       # download all files results-*   
       - name: Download Stryker results
         uses: actions/download-artifact@v4
         with:
           pattern: "results-*"

       # move all mutants-* directories into a single directory
       - name: Gather mutants into single directory
         run: |
           mkdir mutants
           cd mutants
           for mutants in ../mutants-*
           do
             # remove "../mutants-" prefix, leaving only the benchmark name
             name=$(echo $mutants | sed 's/\.\.\/mutants-//')
             echo "moving $mutants to $name"
             mv $mutants/* $name 
           done
           cd ..
           
       # move all results-* directories into a single directory
       - name: Gather Stryker results into single directory
         run: |
           mkdir results
           cd results
           for results in ../results-*
           do
             # extract benchmark name
             name=$(echo $results | sed 's/\.\.\/results-//')
             echo "moving $results to $name"
             mv $results/* $name 
           done
           cd .. 

       - name: Check out llm-mutation-testing
         uses: actions/checkout@v3
         with:
           path: llm-mutation-testing
           
       - name: Generate report
         run: |
           cd llm-mutation-testing
           npm run build
           # if stryker options are specified, include them in the title
           STRYKER_OPTIONS="${{ needs.setup.outputs.strykerOptions }}" 
           if [ -n "$STRYKER_OPTIONS" ]; then
             title="Report (Precomputed mutators $STRYKER_OPTIONS)"
           else
             title="Report (Precomputed mutators)"
           fi
           node .github/generateReport.js "$title" ../results ../mutants > report.md
           more report.md > $GITHUB_STEP_SUMMARY

       - name: Upload report
         uses: actions/upload-artifact@v4
         with:
           name: report.md
           path: llm-mutation-testing/report.md   

       - name: generate Latex table
         run: |
           cd llm-mutation-testing
           node .github/generateLatexTable.js "$title" ../results ../mutants > table.tex        

       - name: Upload table
         uses: actions/upload-artifact@v4
         with:
           name: table.tex
           path: llm-mutation-testing/table.tex