diff --git a/evaluate/metrics/assessment.go b/evaluate/metrics/assessment.go index 5b788ec81..a8b67974c 100644 --- a/evaluate/metrics/assessment.go +++ b/evaluate/metrics/assessment.go @@ -152,3 +152,19 @@ func (a Assessments) StringCSV() (row []string) { return row } + +// CombineWithSymflowerFixAssessments combines the model assessments with the ones from "symflower fix". +func CombineWithSymflowerFixAssessments(model Assessments, fixed Assessments) (combined Assessments) { + combined = NewAssessments() + + combined[AssessmentKeyCoverage] = fixed[AssessmentKeyCoverage] + combined[AssessmentKeyFilesExecuted] = fixed[AssessmentKeyFilesExecuted] + combined[AssessmentKeyGenerateTestsForFileCharacterCount] = model[AssessmentKeyGenerateTestsForFileCharacterCount] + combined[AssessmentKeyProcessingTime] = model[AssessmentKeyProcessingTime] + fixed[AssessmentKeyProcessingTime] + combined[AssessmentKeyResponseCharacterCount] = model[AssessmentKeyResponseCharacterCount] + combined[AssessmentKeyResponseNoError] = model[AssessmentKeyResponseNoError] + combined[AssessmentKeyResponseNoExcess] = model[AssessmentKeyResponseNoExcess] + combined[AssessmentKeyResponseWithCode] = model[AssessmentKeyResponseWithCode] + + return combined +} diff --git a/evaluate/metrics/assessment_test.go b/evaluate/metrics/assessment_test.go index 1ae4fe327..c39e002d0 100644 --- a/evaluate/metrics/assessment_test.go +++ b/evaluate/metrics/assessment_test.go @@ -272,3 +272,54 @@ func TestAssessmentsScore(t *testing.T) { ExpectedScore: uint64(9), }) } + +func TestCombineModelAndSymflowerFixAssessments(t *testing.T) { + type testCase struct { + Name string + + ModelAssessment Assessments + SymflowerFixAssessments Assessments + + ExpectedAssessments Assessments + } + + validate := func(t *testing.T, tc *testCase) { + t.Run(tc.Name, func(t *testing.T) { + actualAssessments := CombineWithSymflowerFixAssessments(tc.ModelAssessment, tc.SymflowerFixAssessments) + + assert.Equal(t, tc.ExpectedAssessments, actualAssessments) + }) + } + + validate(t, &testCase{ + Name: "Simple", + + ModelAssessment: Assessments{ + AssessmentKeyFilesExecuted: 1, + AssessmentKeyProcessingTime: uint64(200), + AssessmentKeyCoverage: 0, + AssessmentKeyResponseCharacterCount: 100, + AssessmentKeyGenerateTestsForFileCharacterCount: 50, + AssessmentKeyResponseNoError: 0, + AssessmentKeyResponseWithCode: 1, + AssessmentKeyResponseNoExcess: 1, + }, + SymflowerFixAssessments: Assessments{ + AssessmentKeyFilesExecuted: 1, + AssessmentKeyProcessingTime: uint64(100), + AssessmentKeyCoverage: 10, + AssessmentKeyResponseNoError: 1, + }, + + ExpectedAssessments: Assessments{ + AssessmentKeyFilesExecuted: 1, + AssessmentKeyProcessingTime: uint64(300), + AssessmentKeyCoverage: 10, + AssessmentKeyResponseCharacterCount: 100, + AssessmentKeyGenerateTestsForFileCharacterCount: 50, + AssessmentKeyResponseNoError: 0, + AssessmentKeyResponseWithCode: 1, + AssessmentKeyResponseNoExcess: 1, + }, + }) +} diff --git a/evaluate/task/symflower-fix.go b/evaluate/task/symflower-fix.go new file mode 100644 index 000000000..a441f4cbd --- /dev/null +++ b/evaluate/task/symflower-fix.go @@ -0,0 +1,32 @@ +package task + +import ( + "context" + "time" + + pkgerrors "github.com/pkg/errors" + "github.com/symflower/eval-dev-quality/evaluate/metrics" + "github.com/symflower/eval-dev-quality/language" + "github.com/symflower/eval-dev-quality/log" + "github.com/symflower/eval-dev-quality/tools" + "github.com/symflower/eval-dev-quality/util" +) + +// symflowerFix runs the "symflower fix" command and returns its execution time in milliseconds. +func symflowerFix(logger *log.Logger, modelAssessment metrics.Assessments, repositoryPath string, language language.Language) (duration uint64, err error) { + start := time.Now() + _, err = util.CommandWithResult(context.Background(), logger, &util.Command{ + Command: []string{ + tools.SymflowerPath, "fix", + "--language", language.ID(), + "--workspace", repositoryPath, + }, + + Directory: repositoryPath, + }) + if err != nil { + return 0, pkgerrors.WithStack(err) + } + + return uint64(time.Since(start).Milliseconds()), nil +} diff --git a/evaluate/task/task-write-test.go b/evaluate/task/task-write-test.go index 2b991dead..afe0a09d3 100644 --- a/evaluate/task/task-write-test.go +++ b/evaluate/task/task-write-test.go @@ -1,6 +1,7 @@ package task import ( + "maps" "path/filepath" pkgerrors "github.com/pkg/errors" @@ -63,6 +64,7 @@ func (t *TaskWriteTests) Run(repository evaltask.Repository) (repositoryAssessme } modelAssessment := metrics.NewAssessments() + var withSymflowerAssessment metrics.Assessments for _, filePath := range filePaths { if err := repository.Reset(t.Logger); err != nil { t.Logger.Panicf("ERROR: unable to reset temporary repository path: %s", err) @@ -93,15 +95,49 @@ func (t *TaskWriteTests) Run(repository evaltask.Repository) (repositoryAssessme if err != nil { problems = append(problems, pkgerrors.WithMessage(err, filePath)) - continue + // Run "symflower fix" if the model response fails to execute. + if t.Language.ID() == "golang" { // Currently we only support Go for "symflower fix". + log.Print("model response alone failed execution, attempting to fix with \"symflower fix \"") + + symflowerAssessment := metrics.NewAssessments() + duration, err := symflowerFix(log, modelAssessment, dataPath, t.Language) + if err != nil { + problems = append(problems, err) + + continue + } + symflowerAssessment[metrics.AssessmentKeyProcessingTime] = duration + + coverage, ps, err := t.Language.Execute(log, dataPath) + problems = append(problems, ps...) + if err != nil { + problems = append(problems, pkgerrors.WithMessage(err, "symflower fix")) + + continue + } + log.Printf("with symflower repair: Executes tests with %d coverage objects", coverage) + + symflowerAssessment.Award(metrics.AssessmentKeyFilesExecuted) + symflowerAssessment.AwardPoints(metrics.AssessmentKeyCoverage, coverage) + + withSymflowerAssessment = metrics.CombineWithSymflowerFixAssessments(modelAssessment, symflowerAssessment) + } else { + continue + } + } else { + log.Printf("Executes tests with %d coverage objects", coverage) + modelAssessment.Award(metrics.AssessmentKeyFilesExecuted) + modelAssessment.AwardPoints(metrics.AssessmentKeyCoverage, coverage) } - log.Printf("Executes tests with %d coverage objects", coverage) - modelAssessment.Award(metrics.AssessmentKeyFilesExecuted) - modelAssessment.AwardPoints(metrics.AssessmentKeyCoverage, coverage) } + // The symflower fix assessment should show how symflower can improve the result, so in case we did not need symflower use the original model assessment. + if withSymflowerAssessment == nil { + withSymflowerAssessment = maps.Clone(modelAssessment) + } repositoryAssessment = map[evaltask.Identifier]metrics.Assessments{ - IdentifierWriteTests: modelAssessment, + IdentifierWriteTests: modelAssessment, + IdentifierWriteTestsSymflowerFix: withSymflowerAssessment, } return repositoryAssessment, problems, nil diff --git a/evaluate/task/task-write-test_test.go b/evaluate/task/task-write-test_test.go index 5b5b2cdc6..8db51495c 100644 --- a/evaluate/task/task-write-test_test.go +++ b/evaluate/task/task-write-test_test.go @@ -14,6 +14,8 @@ import ( "github.com/symflower/eval-dev-quality/log" modeltesting "github.com/symflower/eval-dev-quality/model/testing" "github.com/symflower/eval-dev-quality/task" + "github.com/zimmski/osutil" + "github.com/zimmski/osutil/bytesutil" ) func TestTaskWriteTestsRun(t *testing.T) { @@ -77,4 +79,85 @@ func TestTaskWriteTestsRun(t *testing.T) { }, }) }) + + t.Run("Symflower Fix", func(t *testing.T) { + t.Run("Go", func(t *testing.T) { + validateGo := func(t *testing.T, testName string, testFileContent string, expectedAssessments map[task.Identifier]metrics.Assessments, expectedProblems []string) { + temporaryDirectoryPath := t.TempDir() + repositoryPath := filepath.Join(temporaryDirectoryPath, "golang", "plain") + require.NoError(t, osutil.CopyTree(filepath.Join("..", "..", "testdata", "golang", "plain"), repositoryPath)) + + modelMock := modeltesting.NewMockModelNamed(t, "mocked-model") + modelMock.RegisterGenerateSuccess(t, IdentifierWriteTests, "plain_test.go", testFileContent, metricstesting.AssessmentsWithProcessingTime).Once() + + validate(t, &tasktesting.TestCaseTask{ + Name: testName, + + Model: modelMock, + Language: &golang.Language{}, + TestDataPath: temporaryDirectoryPath, + RepositoryPath: filepath.Join("golang", "plain"), + + ExpectedRepositoryAssessment: expectedAssessments, + ExpectedProblemContains: expectedProblems, + ExpectedResultFiles: map[string]func(t *testing.T, filePath string, data string){ + filepath.Join(string(IdentifierWriteTests), "mocked-model", "golang", "golang", "plain.log"): func(t *testing.T, filePath, data string) { + assert.Contains(t, data, "Evaluating model \"mocked-model\"") + assert.Contains(t, data, "PASS: TestPlain") + }, + }, + }) + } + { + expectedAssessments := map[task.Identifier]metrics.Assessments{ + IdentifierWriteTests: metrics.Assessments{ + metrics.AssessmentKeyFilesExecuted: 1, + metrics.AssessmentKeyResponseNoError: 1, + metrics.AssessmentKeyCoverage: 10, + }, + IdentifierWriteTestsSymflowerFix: metrics.Assessments{ + metrics.AssessmentKeyFilesExecuted: 1, + metrics.AssessmentKeyResponseNoError: 1, + metrics.AssessmentKeyCoverage: 10, + }, + } + validateGo(t, "Model generated correct test", bytesutil.StringTrimIndentations(` + package plain + + import "testing" + + func TestPlain(t *testing.T) { + plain() + } + `), expectedAssessments, nil) + } + { + expectedAssessments := map[task.Identifier]metrics.Assessments{ + IdentifierWriteTests: metrics.Assessments{ + metrics.AssessmentKeyResponseNoError: 1, + }, + IdentifierWriteTestsSymflowerFix: metrics.Assessments{ + metrics.AssessmentKeyFilesExecuted: 1, + metrics.AssessmentKeyResponseNoError: 1, + metrics.AssessmentKeyCoverage: 10, + }, + } + expectedProblems := []string{ + "imported and not used", + } + validateGo(t, "Model generated test with unused import", bytesutil.StringTrimIndentations(` + package plain + + import ( + "testing" + "strings" + ) + + func TestPlain(t *testing.T) { + plain() + } + `), expectedAssessments, expectedProblems) + } + }) + }) } diff --git a/evaluate/task/task.go b/evaluate/task/task.go index 5ddda6ef5..cb422e10c 100644 --- a/evaluate/task/task.go +++ b/evaluate/task/task.go @@ -33,6 +33,8 @@ func registerIdentifier(name string) (identifier evaltask.Identifier) { var ( // IdentifierWriteTests holds the identifier for the "write test" task. IdentifierWriteTests = registerIdentifier("write-tests") + // IdentifierWriteTestsSymflowerFix holds the identifier for the "write test" task with the "symflower fix" applied. + IdentifierWriteTestsSymflowerFix = registerIdentifier("write-tests-symflower-fix") // IdentifierCodeRepair holds the identifier for the "code repair" task. IdentifierCodeRepair = registerIdentifier("code-repair") ) diff --git a/tools/symflower.go b/tools/symflower.go index 8e08b2123..c841a91df 100644 --- a/tools/symflower.go +++ b/tools/symflower.go @@ -91,7 +91,7 @@ func (*symflower) CheckVersion(logger *log.Logger, binaryPath string) (err error } // SymflowerVersionRequired holds the version of Symflower required for this revision of the evaluation. -const SymflowerVersionRequired = "37153" +const SymflowerVersionRequired = "38036" // RequiredVersion returns the required version of the tool. func (*symflower) RequiredVersion() string {