PipReport back to experimental, add pre-generated PipReport parsing (#…

…1201)

* revert experiment graduation, bump threads, and enable fast deps

* put reqs back

* add ability for pip to detect pregenerated reports with a specific naming scheme

* better directory handling

* improve logging

docs/detectors/pip.md

@@ -19,6 +19,9 @@ Serialization specifications:
 - https://peps.python.org/pep-0508/
 - https://peps.python.org/pep-0301/
 
+The detector can also pick up installation reports that have already been generated in the same directory as the `setup.py` or `requirements.txt` files, 
+as long as the report adheres to the following naming scheme: `component-detection-pip-report.json` or `*.component-detection-pip-report.json`
+
 ### Legacy Detection (PipDetector, SimplePipDetector)
 
 Pip detection is performed by running the following code snippet on every *setup.py*:
@@ -60,3 +63,7 @@ The environment variable `PipReportOverrideBehavior` is used to override pip rep
 - `SourceCodeScan`: Scan `setup.py` and `requirements.txt` files, and record components explicitly from the package files without hitting a remote feed. Does not compile a dependency graph.
 
 The environment variable `PipReportSkipFallbackOnFailure` is used to skip the default fallback behavior if pip report fails. Default behavior scans `setup.py` and `requirements.txt` files, and record components explicitly from the package files without hitting a remote feed. Does not compile a dependency graph.
+
+The environment variable `PipReportFileLevelTimeoutSeconds` is used to control the timeout limit for generating the PipReport for individual files. This defaults to the overall timeout.
+
+The environment variable `PipReportDisableFastDeps` is used to disable the fast deps feature in PipReport.

src/Microsoft.ComponentDetection.Contracts/ScanRequest.cs

@@ -19,7 +19,7 @@ public class ScanRequest
     /// <param name="imagesToScan">Container images to scan.</param>
     /// <param name="componentRecorder">Detector component recorder.</param>
     /// <param name="maxThreads">Max number of threads to use for detection.</param>
-    public ScanRequest(DirectoryInfo sourceDirectory, ExcludeDirectoryPredicate directoryExclusionPredicate, ILogger logger, IDictionary<string, string> detectorArgs, IEnumerable<string> imagesToScan, IComponentRecorder componentRecorder, int maxThreads = 3)
+    public ScanRequest(DirectoryInfo sourceDirectory, ExcludeDirectoryPredicate directoryExclusionPredicate, ILogger logger, IDictionary<string, string> detectorArgs, IEnumerable<string> imagesToScan, IComponentRecorder componentRecorder, int maxThreads = 5)
     {
         this.SourceDirectory = sourceDirectory;
         this.DirectoryExclusionPredicate = directoryExclusionPredicate;

src/Microsoft.ComponentDetection.Detectors/pip/PipCommandService.cs

@@ -13,6 +13,8 @@ namespace Microsoft.ComponentDetection.Detectors.Pip;
 
 public class PipCommandService : IPipCommandService
 {
+    private const string PipReportDisableFastDepsEnvVar = "PipReportDisableFastDeps";
+
     private readonly ICommandLineInvocationService commandLineInvocationService;
     private readonly IPathUtilityService pathUtilityService;
     private readonly IFileUtilityService fileUtilityService;
@@ -127,6 +129,11 @@ private async Task<bool> CanCommandBeLocatedAsync(string pipPath)
             pipReportCommand += $" --index-url {this.environmentService.GetEnvironmentVariable("PIP_INDEX_URL")}";
         }
 
+        if (!this.environmentService.DoesEnvironmentVariableExist(PipReportDisableFastDepsEnvVar) || !this.environmentService.IsEnvironmentVariableValueTrue(PipReportDisableFastDepsEnvVar))
+        {
+            pipReportCommand += $" --use-feature=fast-deps";
+        }
+
         this.logger.LogDebug("PipReport: Generating pip installation report for {Path} with command: {Command}", formattedPath, pipReportCommand.RemoveSensitiveInformation());
         command = await this.commandLineInvocationService.ExecuteCommandAsync(
             pipExecutable,

src/Microsoft.ComponentDetection.Detectors/pip/PipComponentDetector.cs

@@ -11,7 +11,7 @@ namespace Microsoft.ComponentDetection.Detectors.Pip;
 using Microsoft.ComponentDetection.Contracts.TypedComponent;
 using Microsoft.Extensions.Logging;
 
-public class PipComponentDetector : FileComponentDetector, IDefaultOffComponentDetector
+public class PipComponentDetector : FileComponentDetector
 {
     private readonly IPythonCommandService pythonCommandService;
     private readonly IPythonResolver pythonResolver;
@@ -38,7 +38,7 @@ public PipComponentDetector(
 
     public override IEnumerable<ComponentType> SupportedComponentTypes { get; } = new[] { ComponentType.Pip };
 
-    public override int Version { get; } = 11;
+    public override int Version { get; } = 12;
 
     protected override async Task<IObservable<ProcessRequest>> OnPrepareDetectionAsync(IObservable<ProcessRequest> processRequests, IDictionary<string, string> detectorArgs)
     {

src/Microsoft.ComponentDetection.Detectors/pip/PipReportComponentDetector.cs

@@ -13,11 +13,16 @@ namespace Microsoft.ComponentDetection.Detectors.Pip;
 using Microsoft.ComponentDetection.Contracts.Internal;
 using Microsoft.ComponentDetection.Contracts.TypedComponent;
 using Microsoft.Extensions.Logging;
+using Newtonsoft.Json;
 
-public class PipReportComponentDetector : FileComponentDetector
+public class PipReportComponentDetector : FileComponentDetector, IExperimentalDetector
 {
+    // environment variables
     private const string PipReportOverrideBehaviorEnvVar = "PipReportOverrideBehavior";
     private const string PipReportSkipFallbackOnFailureEnvVar = "PipReportSkipFallbackOnFailure";
+    private const string PipReportFileLevelTimeoutSecondsEnvVar = "PipReportFileLevelTimeoutSeconds";
+
+    private static readonly IList<string> PipReportPreGeneratedFilePatterns = new List<string> { "*.component-detection-pip-report.json", "component-detection-pip-report.json" };
 
     /// <summary>
     /// The maximum version of the report specification that this detector can handle.
@@ -33,6 +38,7 @@ public class PipReportComponentDetector : FileComponentDetector
     private readonly IEnvironmentVariableService envVarService;
     private readonly IPythonCommandService pythonCommandService;
     private readonly IPythonResolver pythonResolver;
+    private readonly IFileUtilityService fileUtilityService;
 
     public PipReportComponentDetector(
         IComponentStreamEnumerableFactory componentStreamEnumerableFactory,
@@ -41,6 +47,7 @@ public PipReportComponentDetector(
         IEnvironmentVariableService envVarService,
         IPythonCommandService pythonCommandService,
         IPythonResolver pythonResolver,
+        IFileUtilityService fileUtilityService,
         ILogger<PipReportComponentDetector> logger)
     {
         this.ComponentStreamEnumerableFactory = componentStreamEnumerableFactory;
@@ -49,6 +56,7 @@ public PipReportComponentDetector(
         this.envVarService = envVarService;
         this.pythonCommandService = pythonCommandService;
         this.pythonResolver = pythonResolver;
+        this.fileUtilityService = fileUtilityService;
         this.Logger = logger;
     }
 
@@ -67,7 +75,7 @@ private enum PipReportOverrideBehavior
 
     public override IEnumerable<ComponentType> SupportedComponentTypes { get; } = new[] { ComponentType.Pip };
 
-    public override int Version { get; } = 4;
+    public override int Version { get; } = 5;
 
     protected override bool EnableParallelism { get; set; } = true;
 
@@ -117,7 +125,7 @@ protected override async Task OnFileFoundAsync(ProcessRequest processRequest, ID
         var singleFileComponentRecorder = processRequest.SingleFileComponentRecorder;
         var file = processRequest.ComponentStream;
 
-        FileInfo reportFile = null;
+        List<FileInfo> reportFiles = new();
         try
         {
             var pipOverride = this.GetPipReportOverrideBehavior();
@@ -150,44 +158,103 @@ protected override async Task OnFileFoundAsync(ProcessRequest processRequest, ID
             }
 
             var stopwatch = Stopwatch.StartNew();
-            this.Logger.LogInformation("PipReport: Generating pip installation report for {File}", file.Location);
 
-            // Call pip executable to generate the installation report of a given project file.
-            (var report, reportFile) = await this.pipCommandService.GenerateInstallationReportAsync(file.Location, pipExePath, cancellationToken);
+            // Search for a pre-generated pip report file in the same directory as the file being scanned.
+            var fileParentDirectory = Path.GetDirectoryName(file.Location);
+            if (fileParentDirectory is null)
+            {
+                this.Logger.LogWarning("PipReport: Unable to determine parent directory for {File}.", file.Location);
+                return;
+            }
+
+            var fileParentDirectoryInfo = Directory.Exists(fileParentDirectory)
+                ? new DirectoryInfo(fileParentDirectory)
+                : null;
 
-            // The report version is used to determine how to parse the report. If it is greater
-            // than the maximum supported version, there may be new fields and the parsing will fail.
-            if (!int.TryParse(report.Version, out var reportVersion) || reportVersion > MaxReportVersion.Major)
+            List<FileInfo> preGeneratedReportFiles = null;
+            if (fileParentDirectoryInfo is not null)
             {
-                this.Logger.LogWarning(
-                    "PipReport: The pip installation report version {ReportVersion} is not supported. The maximum supported version is {MaxVersion}.",
-                    report.Version,
-                    MaxReportVersion);
+                preGeneratedReportFiles = PipReportPreGeneratedFilePatterns
+                    .SelectMany(pattern => fileParentDirectoryInfo.GetFiles(pattern))
+                    .Where(file => File.Exists(file.FullName))
+                    .ToList();
+            }
+
+            List<PipInstallationReport> reports = new();
+            if (preGeneratedReportFiles is not null && preGeneratedReportFiles.Any())
+            {
+                this.Logger.LogInformation("PipReport: Found pre-generated pip report(s) for {File}.", file.Location);
 
-                using var versionRecord = new InvalidParseVersionTelemetryRecord
+                foreach (var existingReport in preGeneratedReportFiles)
                 {
-                    DetectorId = this.Id,
-                    FilePath = file.Location,
-                    Version = report.Version,
-                    MaxVersion = MaxReportVersion.ToString(),
-                };
+                    this.Logger.LogInformation("PipReport: Using pre-generated pip report '{ReportFile}' for package file '{File}'.", existingReport.FullName, file.Location);
+                    var reportOutput = await this.fileUtilityService.ReadAllTextAsync(existingReport);
+                    var report = JsonConvert.DeserializeObject<PipInstallationReport>(reportOutput);
+                    reports.Add(report);
+                }
+            }
+            else
+            {
+                this.Logger.LogInformation("PipReport: Generating pip installation report for {File}", file.Location);
 
-                return;
+                // create linked cancellation token that will cancel if the file level timeout is reached, or if the parent token is cancelled.
+                // default to only using parent token if the env var is not set or is invalid
+                var childCts = CancellationTokenSource.CreateLinkedTokenSource(cancellationToken);
+                if (this.envVarService.DoesEnvironmentVariableExist(PipReportFileLevelTimeoutSecondsEnvVar)
+                    && int.TryParse(this.envVarService.GetEnvironmentVariable(PipReportFileLevelTimeoutSecondsEnvVar), out var timeoutSeconds))
+                {
+                    childCts.CancelAfter(TimeSpan.FromSeconds(timeoutSeconds));
+                }
+
+                // Call pip executable to generate the installation report of a given project file.
+                (var report, var reportFile) = await this.pipCommandService.GenerateInstallationReportAsync(file.Location, pipExePath, childCts.Token);
+                reports.Add(report);
+                reportFiles.Add(reportFile);
             }
 
-            stopwatch.Stop();
-            this.Logger.LogInformation(
-                "PipReport: Generating pip installation report for {File} completed in {TotalSeconds} seconds with {PkgCount} detected packages.",
-                file.Location,
-                stopwatch.ElapsedMilliseconds / 1000.0,
-                report.InstallItems?.Length ?? 0);
+            if (!reports.Any())
+            {
+                this.Logger.LogWarning("PipReport: Failed to generate or find pip installation report for {File}.", file.Location);
+                return;
+            }
 
-            // Now that all installed packages are known, we can build a graph of the dependencies.
-            if (report.InstallItems is not null)
+            foreach (var report in reports)
             {
-                var graph = this.BuildGraphFromInstallationReport(report);
-                this.RecordComponents(singleFileComponentRecorder, graph);
+                // The report version is used to determine how to parse the report. If it is greater
+                // than the maximum supported version, there may be new fields and the parsing will fail.
+                if (!int.TryParse(report.Version, out var reportVersion) || reportVersion > MaxReportVersion.Major)
+                {
+                    this.Logger.LogWarning(
+                        "PipReport: The pip installation report version {ReportVersion} is not supported. The maximum supported version is {MaxVersion}.",
+                        report.Version,
+                        MaxReportVersion);
+
+                    using var versionRecord = new InvalidParseVersionTelemetryRecord
+                    {
+                        DetectorId = this.Id,
+                        FilePath = file.Location,
+                        Version = report.Version,
+                        MaxVersion = MaxReportVersion.ToString(),
+                    };
+
+                    return;
+                }
+
+                this.Logger.LogInformation(
+                    "PipReport: Pip installation report for {File} completed in {TotalSeconds} seconds with {PkgCount} detected packages.",
+                    file.Location,
+                    stopwatch.ElapsedMilliseconds / 1000.0,
+                    report.InstallItems?.Length ?? 0);
+
+                // Now that all installed packages are known, we can build a graph of the dependencies.
+                if (report.InstallItems is not null)
+                {
+                    var graph = this.BuildGraphFromInstallationReport(report);
+                    this.RecordComponents(singleFileComponentRecorder, graph);
+                }
             }
+
+            stopwatch.Stop();
         }
         catch (Exception e)
         {
@@ -211,9 +278,12 @@ protected override async Task OnFileFoundAsync(ProcessRequest processRequest, ID
         finally
         {
             // Clean up the report output JSON file so it isn't left on the machine.
-            if (reportFile is not null && reportFile.Exists)
+            foreach (var reportFile in reportFiles)
             {
-                reportFile.Delete();
+                if (reportFile is not null && reportFile.Exists)
+                {
+                    reportFile.Delete();
+                }
             }
         }
     }

src/Microsoft.ComponentDetection.Orchestrator/Experiments/Configs/PipReportExperiment.cs

@@ -0,0 +1,18 @@
+namespace Microsoft.ComponentDetection.Orchestrator.Experiments.Configs;
+
+using Microsoft.ComponentDetection.Contracts;
+using Microsoft.ComponentDetection.Detectors.Pip;
+
+/// <summary>
+/// Validating the <see cref="PipReportComponentDetector"/>.
+/// </summary>
+public class PipReportExperiment : IExperimentConfiguration
+{
+    public string Name => "PipReport";
+
+    public bool IsInControlGroup(IComponentDetector componentDetector) => componentDetector is PipComponentDetector;
+
+    public bool IsInExperimentGroup(IComponentDetector componentDetector) => componentDetector is PipReportComponentDetector;
+
+    public bool ShouldRecord(IComponentDetector componentDetector, int numComponents) => true;
+}

src/Microsoft.ComponentDetection.Orchestrator/Extensions/ServiceCollectionExtensions.cs

@@ -64,6 +64,7 @@ public static IServiceCollection AddComponentDetection(this IServiceCollection s
         services.AddSingleton<IExperimentConfiguration, RustCliDetectorExperiment>();
         services.AddSingleton<IExperimentConfiguration, VcpkgExperiment>();
         services.AddSingleton<IExperimentConfiguration, GoDetectorReplaceExperiment>();
+        services.AddSingleton<IExperimentConfiguration, PipReportExperiment>();
 
         // Detectors
         // CocoaPods

src/Microsoft.ComponentDetection.Orchestrator/Services/DetectorProcessingService.cs

@@ -23,7 +23,7 @@ namespace Microsoft.ComponentDetection.Orchestrator.Services;
 
 public class DetectorProcessingService : IDetectorProcessingService
 {
-    private const int DefaultMaxDetectionThreads = 3;
+    private const int DefaultMaxDetectionThreads = 5;
     private const int ExperimentalTimeoutSeconds = 240; // 4 minutes
     private const int ProcessTimeoutBufferSeconds = 5;
 

test/Microsoft.ComponentDetection.Detectors.Tests/Microsoft.ComponentDetection.Detectors.Tests.csproj

@@ -46,6 +46,9 @@
         <None Update="Mocks\pip_report_single_pkg_bad_version.json">
             <CopyToOutputDirectory>Always</CopyToOutputDirectory>
         </None>
+        <None Update="Mocks\test.component-detection-pip-report.json">
+          <CopyToOutputDirectory>Always</CopyToOutputDirectory>
+        </None>
     </ItemGroup>
 
 </Project>