|
| 1 | +# NuGet Scanning Architecture |
| 2 | + |
| 3 | +## Overview |
| 4 | + |
| 5 | +The NuGet scanning subsystem is responsible for discovering, parsing, and analyzing NuGet package dependencies across .NET projects in a repository. It handles both direct package references and transitive (indirect) dependencies, with special support for central package management. |
| 6 | + |
| 7 | +## Architecture |
| 8 | + |
| 9 | +### Components |
| 10 | + |
| 11 | +``` |
| 12 | +RepositoryScanner |
| 13 | + │ |
| 14 | + ├── NuGetInspector |
| 15 | + │ │ |
| 16 | + │ ├── RestorePackagesAsync() # Restore packages for lock/assets files |
| 17 | + │ ├── RefreshCentralPackageVersionFiles() |
| 18 | + │ ├── ReadPackageReferences() # Parse direct PackageReference entries |
| 19 | + │ │ |
| 20 | + │ └── ExtractTransitiveDependencies() # Extract indirect dependencies |
| 21 | + │ │ |
| 22 | + │ ├── ExtractFromLockFile() # From packages.lock.json |
| 23 | + │ └── ExtractFromAssetsFile() # From project.assets.json |
| 24 | + │ |
| 25 | + └── PackageVersionMismatchDetection # Cross-project version alignment |
| 26 | +``` |
| 27 | + |
| 28 | +### Key Classes |
| 29 | + |
| 30 | +#### **RepositoryScanner** |
| 31 | +The main scanning orchestrator that: |
| 32 | +- Discovers all `.csproj` files in the repository |
| 33 | +- Delegates NuGet operations to `NuGetInspector` |
| 34 | +- Aggregates results from all projects |
| 35 | +- Generates comprehensive health reports |
| 36 | +- Detects package version mismatches across projects |
| 37 | + |
| 38 | +**Key Methods:** |
| 39 | +- `ScanAsync()` - Main scanning entry point |
| 40 | +- `GenerateReport()` - Creates structured report document |
| 41 | +- `FindPackageVersionMismatches()` - Identifies inconsistent package versions |
| 42 | + |
| 43 | +#### **NuGetInspector** |
| 44 | +Specialized handler for all NuGet-related operations: |
| 45 | +- Package restore and lock file generation |
| 46 | +- Direct package reference resolution |
| 47 | +- Central package management (Directory.Packages.props) support |
| 48 | +- Transitive dependency extraction |
| 49 | + |
| 50 | +**Key Methods:** |
| 51 | +- `RestorePackagesAsync()` - Executes `dotnet restore` to generate lock/assets files |
| 52 | +- `RefreshCentralPackageVersionFiles()` - Discovers Directory.Packages.props files |
| 53 | +- `ReadPackageReferences()` - Parses and resolves direct dependencies |
| 54 | +- `ExtractTransitiveDependencies()` - Extracts indirect dependencies |
| 55 | + |
| 56 | +### Data Flow |
| 57 | + |
| 58 | +``` |
| 59 | +1. Repository Scan |
| 60 | + └─> Discover all .csproj files |
| 61 | + |
| 62 | +2. Package Restore |
| 63 | + └─> NuGetInspector.RestorePackagesAsync() |
| 64 | + └─> Generates: packages.lock.json or obj/project.assets.json |
| 65 | + |
| 66 | +3. Central Package Discovery |
| 67 | + └─> NuGetInspector.RefreshCentralPackageVersionFiles() |
| 68 | + └─> Finds all Directory.Packages.props files |
| 69 | + └─> Caches version definitions |
| 70 | + |
| 71 | +4. Parse Project Files |
| 72 | + └─> For each .csproj: |
| 73 | + |
| 74 | + a. Read PropertyGroup settings |
| 75 | + └─> Target framework, output type, nullable, etc. |
| 76 | + |
| 77 | + b. Read PackageReference entries |
| 78 | + └─> NuGetInspector.ReadPackageReferences() |
| 79 | + └─> Resolves missing versions from Directory.Packages.props |
| 80 | + └─> Builds Package(Name, Version) list |
| 81 | + |
| 82 | + c. Read ProjectReference entries |
| 83 | + └─> Internal project-to-project references |
| 84 | + |
| 85 | + d. Extract transitive dependencies |
| 86 | + └─> NuGetInspector.ExtractTransitiveDependencies() |
| 87 | + └─> Reads packages.lock.json (preferred) |
| 88 | + OR obj/project.assets.json (fallback) |
| 89 | + └─> Filters out direct deps and project refs |
| 90 | + └─> Builds TransitiveDependency list |
| 91 | + |
| 92 | +5. Aggregate and Detect Mismatches |
| 93 | + └─> FindPackageVersionMismatches() |
| 94 | + └─> Groups packages by name across projects |
| 95 | + └─> Identifies versions with conflicts |
| 96 | + |
| 97 | +6. Generate Report |
| 98 | + └─> ReportDocument with sections: |
| 99 | + - Summary statistics |
| 100 | + - Package version mismatches (if any) |
| 101 | + - Project listing and details |
| 102 | + - Parse errors (if any) |
| 103 | +``` |
| 104 | + |
| 105 | +## Central Package Management Support |
| 106 | + |
| 107 | +CodeMedic supports the MSBuild [central package management](https://learn.microsoft.com/en-us/nuget/consume-packages/central-package-management) feature. |
| 108 | + |
| 109 | +### How It Works |
| 110 | + |
| 111 | +When `Directory.Packages.props` exists at the repository root or in project directories: |
| 112 | + |
| 113 | +1. **Discovery Phase** |
| 114 | + - `RefreshCentralPackageVersionFiles()` recursively finds all `Directory.Packages.props` files |
| 115 | + - Paths are cached for fast lookups |
| 116 | + |
| 117 | +2. **Resolution Phase** |
| 118 | + - When a project references a package without a version attribute: |
| 119 | + ```xml |
| 120 | + <PackageReference Include="Spectre.Console" /> |
| 121 | + ``` |
| 122 | + - `ReadPackageReferences()` walks up from the project directory to the repo root |
| 123 | + - Looks for `Directory.Packages.props` in each directory |
| 124 | + - Parses `<PackageVersion>` entries: |
| 125 | + ```xml |
| 126 | + <PackageVersion Include="Spectre.Console" Version="0.49.0" /> |
| 127 | + ``` |
| 128 | + |
| 129 | +3. **Version Resolution** |
| 130 | + - Version resolved from central file replaces missing inline version |
| 131 | + - Supports `Update` attribute for version overrides |
| 132 | + - Falls back to "unknown" if version cannot be resolved |
| 133 | + |
| 134 | +### Example Structure |
| 135 | + |
| 136 | +``` |
| 137 | +Repository Root/ |
| 138 | +├── Directory.Packages.props # Central definitions |
| 139 | +│ └── <PackageVersion Include="Serilog" Version="3.0.0" /> |
| 140 | +│ |
| 141 | +└── src/ |
| 142 | + ├── Project1/ |
| 143 | + │ └── Project1.csproj |
| 144 | + │ └── <PackageReference Include="Serilog" /> ← Resolved to 3.0.0 |
| 145 | + │ |
| 146 | + └── Project2/ |
| 147 | + └── Project2.csproj |
| 148 | + └── <PackageReference Include="Serilog" /> ← Resolved to 3.0.0 |
| 149 | +``` |
| 150 | +
|
| 151 | +## Package Version Mismatch Detection |
| 152 | +
|
| 153 | +CodeMedic identifies when different projects use different versions of the same package and recommends alignment. |
| 154 | +
|
| 155 | +### Algorithm |
| 156 | +
|
| 157 | +1. Aggregate all packages from all projects |
| 158 | +2. Group by package name (case-insensitive) |
| 159 | +3. For each package: |
| 160 | + - Collect all distinct versions used |
| 161 | + - If more than one distinct version exists: |
| 162 | + - Add to mismatches list |
| 163 | + - Record which projects use which versions |
| 164 | +
|
| 165 | +### Report Output |
| 166 | +
|
| 167 | +When mismatches are detected, the report includes a dedicated section: |
| 168 | +
|
| 169 | +``` |
| 170 | +Package Version Mismatches |
| 171 | +────────────────────────── |
| 172 | +Align package versions across projects to avoid restore/runtime drift. |
| 173 | + |
| 174 | +Packages with differing versions: |
| 175 | + • Newtonsoft.Json: Project1=12.0.3, Project2=13.0.1 |
| 176 | + • Serilog: Project1=2.12.0, Project2=3.0.0 |
| 177 | +``` |
| 178 | +
|
| 179 | +## Transitive Dependencies |
| 180 | +
|
| 181 | +CodeMedic distinguishes between direct and transitive dependencies: |
| 182 | +
|
| 183 | +- **Direct dependencies** come from `<PackageReference>` in project files |
| 184 | +- **Transitive dependencies** are pulled in by direct dependencies (from lock/assets files) |
| 185 | +
|
| 186 | +### Source Tracking |
| 187 | +
|
| 188 | +For transitive dependencies, the system attempts to track which direct dependency introduced them: |
| 189 | +- Walks the NuGet graph from lock/assets file |
| 190 | +- Links transitive package back to originating direct dependency |
| 191 | +- Stored in `TransitiveDependency.SourcePackage` property |
| 192 | +
|
| 193 | +### Private Assets Handling |
| 194 | +
|
| 195 | +When a dependency is marked as private (`PrivateAssets="All"`): |
| 196 | +- Not exposed to projects that reference the current project |
| 197 | +- Still tracked and reported |
| 198 | +- Marked with `IsPrivate` flag |
| 199 | +
|
| 200 | +## File System Abstraction |
| 201 | +
|
| 202 | +To improve testability, `NuGetInspector` uses an abstraction layer for file operations: |
| 203 | +
|
| 204 | +### INuGetFileSystem Interface |
| 205 | +
|
| 206 | +```csharp |
| 207 | +public interface INuGetFileSystem |
| 208 | +{ |
| 209 | + IEnumerable<string> EnumerateFiles(string path, string searchPattern, SearchOption searchOption); |
| 210 | + bool FileExists(string path); |
| 211 | + Stream OpenRead(string path); |
| 212 | +} |
| 213 | +``` |
| 214 | + |
| 215 | +### Implementation Options |
| 216 | + |
| 217 | +- **PhysicalNuGetFileSystem** - Default implementation using actual file system |
| 218 | +- **Mock implementations** - For unit testing (inject custom implementations) |
| 219 | + |
| 220 | +### Usage Example |
| 221 | + |
| 222 | +```csharp |
| 223 | +// Production usage (physical file system) |
| 224 | +var inspector = new NuGetInspector(rootPath); |
| 225 | + |
| 226 | +// Testing usage (mock file system) |
| 227 | +var mockFs = new MockNuGetFileSystem(); |
| 228 | +var inspector = new NuGetInspector(rootPath, mockFs); |
| 229 | +``` |
| 230 | + |
| 231 | +## Error Handling |
| 232 | + |
| 233 | +The NuGet scanning subsystem is designed to be resilient: |
| 234 | + |
| 235 | +- **Package restore failures** are logged but don't stop scanning |
| 236 | +- **Missing lock/assets files** result in no transitive dependencies (acceptable) |
| 237 | +- **Central package file parse errors** are caught and logged |
| 238 | +- **Individual project failures** are recorded in `ProjectInfo.ParseErrors` |
| 239 | +- **Partial results** are preferred over complete failure |
| 240 | + |
| 241 | +## Performance Considerations |
| 242 | + |
| 243 | +### Optimizations |
| 244 | + |
| 245 | +1. **Parallel Line Counting** |
| 246 | + - C# file counting uses `Parallel.ForEach` for multi-core utilization |
| 247 | + - Respects system processor count for optimal performance |
| 248 | + |
| 249 | +2. **Caching** |
| 250 | + - Central package versions cached after first read |
| 251 | + - Prevents redundant XML parsing |
| 252 | + |
| 253 | +3. **Early Returns** |
| 254 | + - Transitive extraction checks `packages.lock.json` first (faster) |
| 255 | + - Falls back to `project.assets.json` only if needed |
| 256 | + |
| 257 | +4. **Streaming** |
| 258 | + - Uses streams for XML and JSON parsing (lower memory overhead) |
| 259 | + - Replaces string-based parsing |
| 260 | + |
| 261 | +### Time Complexity |
| 262 | + |
| 263 | +- **Per-project scanning** - O(n) where n = number of C# files |
| 264 | +- **Lock file parsing** - O(m) where m = number of dependencies |
| 265 | +- **Central package lookup** - O(log d) where d = directory depth |
| 266 | +- **Overall** - Linear with repository size |
| 267 | + |
| 268 | +## Testing Strategy |
| 269 | + |
| 270 | +The file system abstraction enables comprehensive unit testing: |
| 271 | + |
| 272 | +```csharp |
| 273 | +[Fact] |
| 274 | +public void ReadPackageReferences_ResolvesFromCentralPackageFile() |
| 275 | +{ |
| 276 | + var mockFs = new MockNuGetFileSystem(); |
| 277 | + mockFs.AddFile("Directory.Packages.props", |
| 278 | + @"<PackageVersion Include=""TestPkg"" Version=""1.0.0"" />"); |
| 279 | + |
| 280 | + var inspector = new NuGetInspector(rootPath, mockFs); |
| 281 | + var packages = inspector.ReadPackageReferences(projectRoot, ns, projectDir); |
| 282 | + |
| 283 | + Assert.Contains(packages, p => p.Name == "TestPkg" && p.Version == "1.0.0"); |
| 284 | +} |
| 285 | +``` |
| 286 | + |
| 287 | +## Future Enhancements |
| 288 | + |
| 289 | +1. **Transitive Version Conflict Detection** |
| 290 | + - Warn when transitive dependencies have conflicting versions |
| 291 | + - Suggest explicit pinning via PackageVersion entries |
| 292 | + |
| 293 | +2. **Dependency Graph Visualization** |
| 294 | + - Generate mermaid diagrams showing dependency chains |
| 295 | + - Export as JSON for tooling integration |
| 296 | + |
| 297 | +3. **Vulnerability Scanning** |
| 298 | + - Integrate with NuGet security advisories |
| 299 | + - Flag known vulnerabilities in dependencies |
| 300 | + |
| 301 | +4. **License Analysis** |
| 302 | + - Extract and report package licenses |
| 303 | + - Detect license compliance issues |
| 304 | + |
| 305 | +5. **Package Age and Maintenance** |
| 306 | + - Check when packages were last updated |
| 307 | + - Flag abandoned or dormant packages |
| 308 | + |
| 309 | +## Related Documentation |
| 310 | + |
| 311 | +- [Repository Health Dashboard](feature_repository-health-dashboard.md) - High-level overview |
| 312 | +- [Bill of Materials](feature_bill-of-materials.md) - Comprehensive dependency inventory |
| 313 | +- [CLI Architecture](cli_architecture.md) - Command structure and extensibility |
0 commit comments