Clean up, allow for graph dumping, fix the graph

This change adds:

* the ability to write a DOT file representing the
dependency graph of the project with new CLI option
* fixes the graph by correcting a mistake where edges could be linked
  multiple times, and removes original transitive depependency finding
  since that's now in the graph
* adds support for 'embed frameworks' copy files build phase

PBXProjParser/Sources/PBXProjParser/XcodeProject.swift

@@ -42,12 +42,9 @@ public struct XcodeProject {
 
 		packages = model.objects(of: .swiftPackageProductDependency, as: XCSwiftPackageProductDependency.self)
 
-		// First pass - get all the direct dependencies
+		// get all the direct dependencies
 		targets.forEach { determineDirectDependencies($0) }
 
-		// Second pass - get all the transitive dependencies
-		targets.forEach { determineTransitiveDependencies($0) }
-
 		targets.forEach { target in
 			logger.debug("target: \(target.name). Dependencies: \(target.targetDependencies.map { $0.1.name })")
 		}
@@ -105,8 +102,13 @@ public struct XcodeProject {
 			.compactMap { model.object(forKey: $0, as: XCSwiftPackageProductDependency.self) }
 			.forEach { target.add(dependency: .package($0)) }
 
+		// Calculate dependencies from "Embed Frameworks" copy files build phase
+		let embeddedFrameworks = determineEmbeddedFrameworksDependencies(target, with: model)
+
 		// Calculate the dependencies from "Link Binary with Library" build phase
-		let buildFiles = determineBuildPhaseFrameworkDependencies(target, with: model)
+		let linkLibraries = determineBuildPhaseFrameworkDependencies(target, with: model)
+
+		let buildFiles = embeddedFrameworks + linkLibraries
 
 		// Now, we have two potential targets - file & package dependencies.
 		// File dependencies will likely have a reference in another Xcode Project. We might not have seen said project yet, so we need to offload discovery until after we've parsed all projects...
@@ -205,6 +207,25 @@ private func determineBuildPhaseFrameworkDependencies(_ target: PBXNativeTarget,
 		return []
 	}
 
-	return buildPhase.files
+	return buildPhase
+		.files
 		.compactMap { model.object(forKey: $0, as: PBXBuildFile.self) }
 }
+
+private func determineEmbeddedFrameworksDependencies(_ target: PBXNativeTarget, with model: PBXProj) -> [PBXBuildFile] {
+	// Find the "Embed Frameworks" build phase (copy files build phase)
+	let buildPhases = target
+		.buildPhases
+		.compactMap { model.object(forKey: $0, as: PBXCopyFilesBuildPhase.self) }
+
+	return buildPhases
+		.flatMap { $0.files }
+		.compactMap { model.object(forKey: $0, as: PBXBuildFile.self) }
+		.filter { file in
+			guard let ref = file.fileRef else { return false }
+			guard let object = model.object(forKey: ref, as: PBXFileReference.self) else { return false }
+			guard object.explicitFileType == "wrapper.framework" else { return false }
+
+			return true
+		}
+}

Sources/GenIR/Dependency Graph/DependencyGraph.swift

@@ -5,6 +5,8 @@
 //  Created by Thomas Hedderwick on 28/08/2023.
 //
 
+import Foundation
+
 /// A directed graph that maps dependencies between targets (nodes) via edges (directions between nodes)
 class DependencyGraph {
 	/// All the nodes in the graph
@@ -42,6 +44,19 @@ class DependencyGraph {
 		return depthFirstSearch(startingAt: targetNode)
 	}
 
+	func toDot(_ path: String) throws {
+		var contents = "digraph DependencyGraph {\n"
+
+		for node in nodes {
+			for edge in node.edges.filter({ $0.relationship == .dependency }) {
+				contents.append("\(node.name.replacingOccurrences(of: "-", with: "_")) -> \(edge.to.name.replacingOccurrences(of: "-", with: "_"))\n")
+			}
+		}
+
+		contents.append("}")
+		try contents.write(toFile: path, atomically: true, encoding: .utf8)
+	}
+
 	/// Perform a depth-first search starting at the provided node
 	/// - Parameter node: the node whose children to search through
 	/// - Returns: an array of nodes ordered by a depth-first search approach

Sources/GenIR/Dependency Graph/Node.swift

@@ -23,7 +23,9 @@ class Node {
 	/// Adds an edge to this node
 	/// - Parameter edge: the edge to add
 	func add(edge: Edge) {
-		edges.append(edge)
+		if edges.filter({ $0.to.name == edge.to.name }).count == 0 {
+			edges.append(edge)
+		}
 	}
 }
 

Sources/GenIR/GenIR.swift

@@ -59,6 +59,9 @@ struct IREmitterCommand: ParsableCommand {
 	@Flag(help: "Runs the tool without outputting IR to disk (i.e. leaving out the compiler command runner stage)")
 	var dryRun = false
 
+	@Flag(help: "Output the dependency graph as .dot files to the output directory - debug only")
+	var dumpDependencyGraph = false
+
 	mutating func validate() throws {
 		if debug {
 			logger.logLevel = .debug
@@ -121,7 +124,12 @@ struct IREmitterCommand: ParsableCommand {
 		)
 		try runner.run(targets: targets)
 
-		let postprocessor = try OutputPostprocessor(archive: archive, output: output, targets: targets)
+		let postprocessor = try OutputPostprocessor(
+			archive: archive,
+			output: output,
+			targets: targets,
+			dumpGraph: dumpDependencyGraph
+		)
 		try postprocessor.process(targets: &targets)
 	}
 

Sources/GenIR/OutputPostprocessor.swift

@@ -8,7 +8,7 @@
 import Foundation
 import PBXProjParser
 
-private typealias FilenameAndSize = (String, Int)
+private typealias SizeAndCreation = (Int, Date)
 
 /// The `OutputPostprocessor` is responsible for trying to match the IR output of the `CompilerCommandRunner` with the products in the `xcarchive`.
 /// The `CompilerCommandRunner` will output IR with it's product name, but doesn't take into account the linking of products into each other.
@@ -26,17 +26,24 @@ class OutputPostprocessor {
 
 	private let graph: DependencyGraph
 
-	private var seenConflictingFiles: [URL: [FilenameAndSize]] = [:]
+	private var seenConflictingFiles: [URL: [SizeAndCreation]] = [:]
 
 	private let manager: FileManager = .default
 
-	init(archive: URL, output: URL, targets: Targets) throws {
+	init(archive: URL, output: URL, targets: Targets, dumpGraph: Bool) throws {
 		self.output = output
 		self.archive = archive
 
 		dynamicDependencyToPath = dynamicDependencies(in: self.archive)
 
 		graph = DependencyGraphBuilder.build(targets: targets)
+		if dumpGraph {
+			do {
+				try graph.toDot(output.appendingPathComponent("graph.dot").filePath)
+			} catch {
+				logger.error("toDot error: \(error)")
+			}
+		}
 	}
 
 	/// Starts the OutputPostprocessor
@@ -78,8 +85,8 @@ class OutputPostprocessor {
 	) throws -> Set<URL> {
 		let chain = graph.chain(for: target)
 
-		logger.info("Chain for target: \(target.nameForOutput):\n")
-		chain.forEach { logger.info("\($0)") }
+		logger.debug("Chain for target: \(target.nameForOutput):\n")
+		chain.forEach { logger.debug("\($0)") }
 
 		// We want to process the chain, visiting each node _shallowly_ and copy it's dependencies into it's parent
 		var processed = Set<URL>()
@@ -164,48 +171,39 @@ class OutputPostprocessor {
 	///   - source: source file path
 	///   - destination: destination file path
 	private func copyFileUniquingConflictingFiles(source: URL, destination: URL) throws {
-		/// Returns the size of the path. Throws is attributes cannot be found for this file path.
-		/// - Parameter path: the path to get the file size of
-		/// - Returns: the size of the file at path, if it was able to be converted to an integer
-		func size(for path: URL) throws -> Int? {
-			do {
-				return try manager.attributesOfItem(atPath: path.filePath)[.size] as? Int
-			} catch {
-				logger.debug("Couldn't get size attribute for path: \(path.filePath)")
-				throw error
-			}
-		}
+		let destinationAttributes = try manager.attributesOfItem(atPath: destination.filePath)
+		let sourceAttributes = try manager.attributesOfItem(atPath: source.filePath)
 
 		guard
-			let destinationSize = try size(for: destination),
-			let sourceSize = try size(for: source)
+			let destinationSize = destinationAttributes[.size] as? Int,
+			let sourceSize = sourceAttributes[.size] as? Int,
+			let destinationCreatedDate = destinationAttributes[.creationDate] as? Date,
+			let sourceCreatedDate = sourceAttributes[.creationDate] as? Date
 		else {
+			logger.debug("Failed to get attributes for source: \(source) & destination: \(destination)")
 			return
 		}
 
 		let uniqueDestinationURL = manager.uniqueFilename(directory: destination.deletingLastPathComponent(), filename: source.lastPathComponent)
 
-		// TODO: Should we use all file attributes here? What about created date etc?
 		if seenConflictingFiles[source] == nil {
-			seenConflictingFiles[source] = [(source.lastPathComponent, sourceSize)]
+			seenConflictingFiles[source] = [(sourceSize, sourceCreatedDate)]
 		}
 
-		for (_, size) in seenConflictingFiles[source]! where size == destinationSize {
-			logger.debug("Ignoring copy of: \(destination.lastPathComponent) as the sizes where the same")
+		for (size, date) in seenConflictingFiles[source]! where size == destinationSize && date == destinationCreatedDate {
 			return
 		}
 
-		seenConflictingFiles[source]!.append((uniqueDestinationURL.lastPathComponent, destinationSize))
-
+		seenConflictingFiles[source]!.append((destinationSize, destinationCreatedDate))
+		logger.debug("Copying source \(source) to destination: \(uniqueDestinationURL)")
 		try manager.copyItem(at: source, to: uniqueDestinationURL)
 	}
 }
 
-// swiftlint:disable private_over_fileprivate
 /// Returns a map of dynamic objects in the provided path
 /// - Parameter xcarchive: the path to search through
 /// - Returns: a mapping of filename to filepath for dynamic objects in the provided path
-fileprivate func dynamicDependencies(in xcarchive: URL) -> [String: URL] {
+private func dynamicDependencies(in xcarchive: URL) -> [String: URL] {
 	let searchPath = baseSearchPath(startingAt: xcarchive)
 	logger.debug("Using search path for dynamic dependencies: \(searchPath)")
 
@@ -224,7 +222,7 @@ fileprivate func dynamicDependencies(in xcarchive: URL) -> [String: URL] {
 /// Returns the base URL to start searching inside an xcarchive
 /// - Parameter path: the original path, should be an xcarchive
 /// - Returns: the path to start a dependency search from
-fileprivate func baseSearchPath(startingAt path: URL) -> URL {
+private func baseSearchPath(startingAt path: URL) -> URL {
 	let productsPath = path.appendingPathComponent("Products")
 	let applicationsPath = productsPath.appendingPathComponent("Applications")
 	let frameworkPath = productsPath.appendingPathComponent("Library").appendingPathComponent("Framework")
@@ -254,4 +252,3 @@ fileprivate func baseSearchPath(startingAt path: URL) -> URL {
 	logger.debug("Couldn't determine the base search path for the xcarchive, using: \(productsPath)")
 	return productsPath
 }
-// swiftlint:enable private_over_fileprivate