WP-2: PDF composer + PageLayout (two-pass, atomic write, zero annotations) #3

Merged
kua-agent merged 1 commits from wp2/pdf-composer-20260831 into feat/shotdeck-20260830 2026-08-31 11:01:03 +00:00
3 changed files with 937 additions and 0 deletions
+209
View File
@@ -0,0 +1,209 @@
import AppKit
import CoreGraphics
import CoreText
import Foundation
import ImageIO
import os
private let pdfLog = Logger(subsystem: "ai.flowmaster.shotdeck", category: "PDF")
public struct PDFComposer: Sendable {
public init() {}
@discardableResult
public func compose(
session: CaptureSession,
imageURL: (Capture) -> URL,
title: String,
to outputURL: URL
) throws -> Int {
if session.isEmpty {
throw ShotdeckError.pdfCompositionFailed(reason: "session has no captures")
}
var loaded: [(Capture, CGImage)] = []
loaded.reserveCapacity(session.captures.count)
for capture in session.captures {
let url = imageURL(capture)
if let image = Self.loadCGImage(from: url) {
loaded.append((capture, image))
} else {
pdfLog.error(
"skipping capture \(capture.id.uuidString, privacy: .public) seq \(capture.sequence, privacy: .public): image unreadable at \(url.path, privacy: .public)"
)
}
}
guard !loaded.isEmpty else {
throw ShotdeckError.pdfCompositionFailed(reason: "no readable images in session")
}
let pageCount = loaded.count
let tmpURL = outputURL
.deletingLastPathComponent()
.appendingPathComponent(".tmp-\(UUID().uuidString)-\(outputURL.lastPathComponent)")
do {
try Self.writePDF(
loaded: loaded,
pageCount: pageCount,
title: title,
sessionID: session.id,
to: tmpURL
)
let handle = try FileHandle(forWritingTo: tmpURL)
try handle.synchronize()
try handle.close()
if FileManager.default.fileExists(atPath: outputURL.path) {
try FileManager.default.removeItem(at: outputURL)
}
try FileManager.default.moveItem(at: tmpURL, to: outputURL)
return pageCount
} catch {
try? FileManager.default.removeItem(at: tmpURL)
if let shotdeck = error as? ShotdeckError {
throw shotdeck
}
throw ShotdeckError.pdfCompositionFailed(reason: error.localizedDescription)
}
}
public static func fileName(for session: CaptureSession) -> String {
"Shotdeck-\(DubaiTime.fileStamp(session.createdAt)).pdf"
}
private static func writePDF(
loaded: [(Capture, CGImage)],
pageCount: Int,
title: String,
sessionID: UUID,
to tmpURL: URL
) throws {
guard let consumer = CGDataConsumer(url: tmpURL as CFURL) else {
throw ShotdeckError.pdfCompositionFailed(reason: "cannot open output location")
}
let auxiliaryInfo: [String: Any] = [
kCGPDFContextCreator as String: "Shotdeck",
kCGPDFContextTitle as String: title,
kCGPDFContextSubject as String: sessionID.uuidString.lowercased(),
]
guard let context = CGContext(
consumer: consumer,
mediaBox: nil,
auxiliaryInfo as CFDictionary
) else {
throw ShotdeckError.pdfCompositionFailed(reason: "cannot open output location")
}
let lightGrey = CGColor(gray: 0.75, alpha: 1)
let black = CGColor(gray: 0, alpha: 1)
let timestampColor = CGColor(gray: 0.45, alpha: 1)
for pageIndex in 1...pageCount {
let (capture, cgImage) = loaded[pageIndex - 1]
let layout = PageLayout(capture: capture, pageIndex: pageIndex, pageCount: pageCount)
context.beginPDFPage(Self.pageInfo(mediaBox: layout.pageRect))
context.draw(cgImage, in: layout.imageRect)
context.setStrokeColor(lightGrey)
context.setLineWidth(0.5)
context.move(to: CGPoint(x: layout.headerRect.minX, y: layout.headerRect.minY))
context.addLine(to: CGPoint(x: layout.headerRect.maxX, y: layout.headerRect.minY))
context.strokePath()
drawText(
layout.pageNumberText,
font: layout.pageNumberFont,
color: black,
at: layout.pageNumberOrigin,
in: context
)
drawText(
layout.timestampText,
font: layout.timestampFont,
color: timestampColor,
at: layout.timestampOrigin,
in: context
)
drawText(
"PASS",
font: layout.tickLabelFont,
color: black,
at: layout.passLabelOrigin,
in: context
)
drawText(
"FAIL",
font: layout.tickLabelFont,
color: black,
at: layout.failLabelOrigin,
in: context
)
strokeTickBox(layout.passTickBox, in: context, color: black)
strokeTickBox(layout.failTickBox, in: context, color: black)
context.endPDFPage()
}
context.closePDF()
}
private static func pageInfo(mediaBox: CGRect) -> CFDictionary {
var box = mediaBox
let data = Data(bytes: &box, count: MemoryLayout<CGRect>.size)
return [kCGPDFContextMediaBox as String: data] as CFDictionary
}
private static func drawText(
_ string: String,
font: NSFont,
color: CGColor,
at origin: CGPoint,
in context: CGContext
) {
let attributes: [CFString: Any] = [
kCTFontAttributeName: font,
kCTForegroundColorAttributeName: color,
]
guard let attributed = CFAttributedStringCreate(
nil,
string as CFString,
attributes as CFDictionary
) else {
return
}
let line = CTLineCreateWithAttributedString(attributed)
context.textMatrix = .identity
context.textPosition = origin
CTLineDraw(line, context)
}
private static func strokeTickBox(_ box: CGRect, in context: CGContext, color: CGColor) {
let path = CGPath(
roundedRect: box,
cornerWidth: 2,
cornerHeight: 2,
transform: nil
)
context.addPath(path)
context.setStrokeColor(color)
context.setLineWidth(1.0)
context.strokePath()
}
private static func loadCGImage(from url: URL) -> CGImage? {
guard let source = CGImageSourceCreateWithURL(url as CFURL, nil),
CGImageSourceGetCount(source) > 0
else {
return nil
}
return CGImageSourceCreateImageAtIndex(source, 0, nil)
}
}
+149
View File
@@ -0,0 +1,149 @@
import AppKit
import CoreGraphics
import CoreText
import Foundation
public struct PageLayout: Sendable, Equatable {
public let isLandscape: Bool
public let pageRect: CGRect
public let headerRect: CGRect
public let imageBoxRect: CGRect
public let imageRect: CGRect
public let pageNumberText: String
public let pageNumberOrigin: CGPoint
public let timestampText: String
public let timestampOrigin: CGPoint
public let passLabelOrigin: CGPoint
public let failLabelOrigin: CGPoint
public let passTickBox: CGRect
public let failTickBox: CGRect
/// Constructed at access time so `PageLayout` stays Sendable under Swift 6
/// (`NSFont` is not Sendable). Same `NSFont.systemFont` values used for measurement.
public var pageNumberFont: NSFont { NSFont.systemFont(ofSize: 9, weight: .semibold) }
public var timestampFont: NSFont { NSFont.systemFont(ofSize: 9, weight: .regular) }
public var tickLabelFont: NSFont { NSFont.systemFont(ofSize: 8, weight: .semibold) }
private static let marginAll: CGFloat = 18
private static let headerHeight: CGFloat = 26
private static let headerImageGap: CGFloat = 10
private static let tickBoxSize: CGFloat = 13
private static let tickGroupGap: CGFloat = 8
private static let portraitPage = CGSize(width: 595, height: 842)
private static let landscapePage = CGSize(width: 842, height: 595)
/// `pageIndex` and `pageCount` are 1-based / total, over ACTUALLY-WRITTEN pages
/// (see SPEC decision D-D), not `capture.sequence` / `session.captures.count`.
public init(capture: Capture, pageIndex: Int, pageCount: Int) {
let pageNumberFont = NSFont.systemFont(ofSize: 9, weight: .semibold)
let timestampFont = NSFont.systemFont(ofSize: 9, weight: .regular)
let tickLabelFont = NSFont.systemFont(ofSize: 8, weight: .semibold)
isLandscape = capture.isLandscape
pageRect = CGRect(origin: .zero, size: isLandscape ? Self.landscapePage : Self.portraitPage)
let W = pageRect.width
let H = pageRect.height
let contentMinX = Self.marginAll
let contentMaxX = W - Self.marginAll
let contentMinY = Self.marginAll
let contentMaxY = H - Self.marginAll
let contentWidth = contentMaxX - contentMinX
headerRect = CGRect(
x: contentMinX,
y: contentMaxY - Self.headerHeight,
width: contentWidth,
height: Self.headerHeight
)
imageBoxRect = CGRect(
x: contentMinX,
y: contentMinY,
width: contentWidth,
height: headerRect.minY - Self.headerImageGap - contentMinY
)
let pointSize = CGSize(
width: CGFloat(capture.pixelWidth) / capture.scale,
height: CGFloat(capture.pixelHeight) / capture.scale
)
let fitScale = min(
imageBoxRect.width / pointSize.width,
imageBoxRect.height / pointSize.height,
1.0
)
let drawnSize = CGSize(
width: pointSize.width * fitScale,
height: pointSize.height * fitScale
)
imageRect = CGRect(
x: imageBoxRect.midX - drawnSize.width / 2,
y: imageBoxRect.midY - drawnSize.height / 2,
width: drawnSize.width,
height: drawnSize.height
)
let metricsFont = timestampFont as CTFont
let ascent = CTFontGetAscent(metricsFont)
let descent = CTFontGetDescent(metricsFont)
let baselineY = headerRect.minY + (headerRect.height - (ascent + descent)) / 2 + descent
pageNumberText = "\(pageIndex) / \(pageCount)"
pageNumberOrigin = CGPoint(x: headerRect.minX, y: baselineY)
timestampText = DubaiTime.stamp(capture.capturedAt)
let pageNumberWidth = Self.measuredWidth(pageNumberText, font: pageNumberFont)
timestampOrigin = CGPoint(x: headerRect.minX + pageNumberWidth + 10, y: baselineY)
let groupRightX = headerRect.maxX
let failBoxMinX = groupRightX - Self.tickBoxSize
let tickBoxY = headerRect.midY - Self.tickBoxSize / 2
failTickBox = CGRect(
x: failBoxMinX,
y: tickBoxY,
width: Self.tickBoxSize,
height: Self.tickBoxSize
)
let failLabelWidth = Self.measuredWidth("FAIL", font: tickLabelFont)
let failLabelMaxX = failTickBox.minX - Self.tickGroupGap
failLabelOrigin = CGPoint(x: failLabelMaxX - failLabelWidth, y: baselineY)
let passBoxMaxX = failLabelOrigin.x - Self.tickGroupGap
passTickBox = CGRect(
x: passBoxMaxX - Self.tickBoxSize,
y: tickBoxY,
width: Self.tickBoxSize,
height: Self.tickBoxSize
)
let passLabelWidth = Self.measuredWidth("PASS", font: tickLabelFont)
let passLabelMaxX = passTickBox.minX - Self.tickGroupGap
passLabelOrigin = CGPoint(x: passLabelMaxX - passLabelWidth, y: baselineY)
}
public static func == (lhs: PageLayout, rhs: PageLayout) -> Bool {
lhs.isLandscape == rhs.isLandscape
&& lhs.pageRect == rhs.pageRect
&& lhs.headerRect == rhs.headerRect
&& lhs.imageBoxRect == rhs.imageBoxRect
&& lhs.imageRect == rhs.imageRect
&& lhs.pageNumberText == rhs.pageNumberText
&& lhs.pageNumberOrigin == rhs.pageNumberOrigin
&& lhs.timestampText == rhs.timestampText
&& lhs.timestampOrigin == rhs.timestampOrigin
&& lhs.passLabelOrigin == rhs.passLabelOrigin
&& lhs.failLabelOrigin == rhs.failLabelOrigin
&& lhs.passTickBox == rhs.passTickBox
&& lhs.failTickBox == rhs.failTickBox
}
private static func measuredWidth(_ text: String, font: NSFont) -> CGFloat {
let attributed = NSAttributedString(string: text, attributes: [.font: font])
let line = CTLineCreateWithAttributedString(attributed)
return CGFloat(CTLineGetTypographicBounds(line, nil, nil, nil))
}
}
@@ -0,0 +1,579 @@
import CoreGraphics
import Foundation
import ImageIO
import PDFKit
import ShotdeckCore
import Testing
import UniformTypeIdentifiers
enum TestFixtureError: Error { case cannotCreateBitmap, cannotCreatePNG, cannotFinalizePNG }
func writeSolidPNG(
pixelWidth: Int,
pixelHeight: Int,
red: UInt8,
green: UInt8,
blue: UInt8,
to url: URL
) throws {
let colorSpace = CGColorSpaceCreateDeviceRGB()
guard let context = CGContext(
data: nil,
width: pixelWidth,
height: pixelHeight,
bitsPerComponent: 8,
bytesPerRow: pixelWidth * 4,
space: colorSpace,
bitmapInfo: CGImageAlphaInfo.premultipliedLast.rawValue
) else { throw TestFixtureError.cannotCreateBitmap }
context.setFillColor(CGColor(
red: CGFloat(red) / 255,
green: CGFloat(green) / 255,
blue: CGFloat(blue) / 255,
alpha: 1
))
context.fill(CGRect(x: 0, y: 0, width: pixelWidth, height: pixelHeight))
guard let image = context.makeImage(),
let destination = CGImageDestinationCreateWithURL(
url as CFURL,
UTType.png.identifier as CFString,
1,
nil
)
else { throw TestFixtureError.cannotCreatePNG }
CGImageDestinationAddImage(destination, image, nil)
guard CGImageDestinationFinalize(destination) else { throw TestFixtureError.cannotFinalizePNG }
}
private let fixtureDate = Date(timeIntervalSince1970: 1_777_482_180)
private func makeCapture(
sequence: Int,
pixelWidth: Int,
pixelHeight: Int,
scale: CGFloat = 1,
capturedAt: Date = fixtureDate,
id: UUID = UUID(),
fileName: String? = nil
) -> Capture {
Capture(
id: id,
sequence: sequence,
fileName: fileName ?? String(format: "%03d.png", sequence),
pixelWidth: pixelWidth,
pixelHeight: pixelHeight,
scale: scale,
capturedAt: capturedAt
)
}
private func makeSession(
captures: [Capture],
id: UUID = UUID(),
createdAt: Date = fixtureDate
) -> CaptureSession {
CaptureSession(
id: id,
createdAt: createdAt,
state: .open,
captures: captures,
pdfFileName: nil
)
}
private func makeScratchDirectory() throws -> URL {
let url = FileManager.default.temporaryDirectory
.appendingPathComponent("shotdeck-pdf-\(UUID().uuidString)", isDirectory: true)
try FileManager.default.createDirectory(at: url, withIntermediateDirectories: true)
return url
}
private func writePNG(
for capture: Capture,
red: UInt8,
green: UInt8,
blue: UInt8,
in directory: URL
) throws -> URL {
let url = directory.appendingPathComponent(capture.fileName)
try writeSolidPNG(
pixelWidth: capture.pixelWidth,
pixelHeight: capture.pixelHeight,
red: red,
green: green,
blue: blue,
to: url
)
return url
}
private func imageURLMap(_ urls: [UUID: URL]) -> (Capture) -> URL {
{ capture in
urls[capture.id] ?? URL(fileURLWithPath: "/nonexistent/\(capture.id.uuidString).png")
}
}
private func openDocument(_ url: URL) throws -> PDFDocument {
let document = try #require(PDFDocument(url: url))
return document
}
private func pdftoppm(pdf: URL, prefix: URL) throws {
let process = Process()
process.executableURL = URL(fileURLWithPath: "/opt/homebrew/bin/pdftoppm")
process.arguments = ["-png", "-r", "110", pdf.path, prefix.path]
let err = Pipe()
process.standardError = err
try process.run()
process.waitUntilExit()
guard process.terminationStatus == 0 else {
let message = String(data: err.fileHandleForReading.readDataToEndOfFile(), encoding: .utf8) ?? ""
throw ShotdeckError.pdfCompositionFailed(reason: "pdftoppm failed: \(message)")
}
}
private struct RGBABitmap {
let width: Int
let height: Int
let bytesPerRow: Int
let data: Data
func pixel(x: Int, y: Int) -> (UInt8, UInt8, UInt8)? {
guard x >= 0, y >= 0, x < width, y < height else { return nil }
let offset = y * bytesPerRow + x * 4
guard offset + 2 < data.count else { return nil }
return (data[offset], data[offset + 1], data[offset + 2])
}
func containsNonWhite(x0: Int, y0: Int, x1: Int, y1: Int) -> Bool {
let xStart = max(0, x0)
let yStart = max(0, y0)
let xEnd = min(width, x1)
let yEnd = min(height, y1)
guard xStart < xEnd, yStart < yEnd else { return false }
for y in yStart..<yEnd {
for x in xStart..<xEnd {
if let (r, g, b) = pixel(x: x, y: y), r != 255 || g != 255 || b != 255 {
return true
}
}
}
return false
}
static func loadPNG(at url: URL) throws -> RGBABitmap {
let colorSpace = CGColorSpaceCreateDeviceRGB()
let source = try #require(CGImageSourceCreateWithURL(url as CFURL, nil))
let image = try #require(CGImageSourceCreateImageAtIndex(source, 0, nil))
let width = image.width
let height = image.height
let bytesPerRow = width * 4
var data = Data(count: bytesPerRow * height)
try data.withUnsafeMutableBytes { raw in
guard let base = raw.baseAddress else {
throw TestFixtureError.cannotCreateBitmap
}
guard let context = CGContext(
data: base,
width: width,
height: height,
bitsPerComponent: 8,
bytesPerRow: bytesPerRow,
space: colorSpace,
bitmapInfo: CGImageAlphaInfo.premultipliedLast.rawValue
) else {
throw TestFixtureError.cannotCreateBitmap
}
context.interpolationQuality = .none
context.draw(image, in: CGRect(x: 0, y: 0, width: width, height: height))
}
return RGBABitmap(width: width, height: height, bytesPerRow: bytesPerRow, data: data)
}
}
private func pixelWindow(
for rect: CGRect,
pageHeightPt: CGFloat,
scale k: CGFloat
) -> (x0: Int, y0: Int, x1: Int, y1: Int) {
let pixelX0 = Int(floor(rect.minX * k))
let pixelX1 = Int(ceil(rect.maxX * k))
let pixelY0 = Int(floor((pageHeightPt - rect.maxY) * k))
let pixelY1 = Int(ceil((pageHeightPt - rect.minY) * k))
return (pixelX0, pixelY0, pixelX1, pixelY1)
}
@Test("Three-page basic compose")
func threePageBasicCompose() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let c1 = makeCapture(sequence: 1, pixelWidth: 1600, pixelHeight: 900)
let c2 = makeCapture(sequence: 2, pixelWidth: 900, pixelHeight: 1600)
let c3 = makeCapture(sequence: 3, pixelWidth: 1200, pixelHeight: 800)
let urls: [UUID: URL] = [
c1.id: try writePNG(for: c1, red: 200, green: 20, blue: 20, in: directory),
c2.id: try writePNG(for: c2, red: 20, green: 200, blue: 20, in: directory),
c3.id: try writePNG(for: c3, red: 20, green: 20, blue: 200, in: directory),
]
let session = makeSession(captures: [c1, c2, c3])
let output = directory.appendingPathComponent("three.pdf")
let pageCount = try PDFComposer().compose(
session: session,
imageURL: imageURLMap(urls),
title: "Three page",
to: output
)
#expect(pageCount == 3)
let document = try openDocument(output)
#expect(document.pageCount == 3)
let p0 = try #require(document.page(at: 0))
let p1 = try #require(document.page(at: 1))
let p2 = try #require(document.page(at: 2))
#expect(p0.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 842, height: 595))
#expect(p1.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 595, height: 842))
#expect(p2.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 842, height: 595))
}
@Test("Wide/tall page geometry exact")
func wideTallPageGeometryExact() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let wide = makeCapture(sequence: 1, pixelWidth: 1600, pixelHeight: 900)
let tall = makeCapture(sequence: 1, pixelWidth: 900, pixelHeight: 1600)
let wideURL = try writePNG(for: wide, red: 10, green: 10, blue: 10, in: directory)
let tallURL = try writePNG(for: tall, red: 10, green: 10, blue: 10, in: directory)
let wideOut = directory.appendingPathComponent("wide.pdf")
_ = try PDFComposer().compose(
session: makeSession(captures: [wide]),
imageURL: { _ in wideURL },
title: "Wide",
to: wideOut
)
let tallOut = directory.appendingPathComponent("tall.pdf")
_ = try PDFComposer().compose(
session: makeSession(captures: [tall]),
imageURL: { _ in tallURL },
title: "Tall",
to: tallOut
)
let widePage = try #require(openDocument(wideOut).page(at: 0))
let tallPage = try #require(openDocument(tallOut).page(at: 0))
#expect(widePage.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 842, height: 595))
#expect(tallPage.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 595, height: 842))
}
@Test("Square resolves to portrait (Capture.swift:39-42 isLandscape strict >)")
func squareResolvesToPortrait() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let square = makeCapture(sequence: 1, pixelWidth: 1000, pixelHeight: 1000)
let png = try writePNG(for: square, red: 80, green: 80, blue: 80, in: directory)
let output = directory.appendingPathComponent("square.pdf")
_ = try PDFComposer().compose(
session: makeSession(captures: [square]),
imageURL: { _ in png },
title: "Square",
to: output
)
let page = try #require(openDocument(output).page(at: 0))
#expect(page.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 595, height: 842))
}
@Test("Retina capture uses point size, not pixel size")
func retinaCaptureUsesPointSize() throws {
let capture = makeCapture(sequence: 1, pixelWidth: 2880, pixelHeight: 1620, scale: 2.0)
let layout = PageLayout(capture: capture, pageIndex: 1, pageCount: 1)
let expectedScale = min(806.0 / 1440.0, 523.0 / 810.0, 1.0)
let expected = CGSize(width: 1440.0 * expectedScale, height: 810.0 * expectedScale)
#expect(abs(layout.imageRect.width - expected.width) < 0.01)
#expect(abs(layout.imageRect.height - expected.height) < 0.01)
#expect(layout.imageRect.width < 2000)
#expect(layout.isLandscape)
}
@Test("Zero PDFAnnotation objects on every page")
func zeroPDFAnnotationObjects() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let c1 = makeCapture(sequence: 1, pixelWidth: 800, pixelHeight: 600)
let c2 = makeCapture(sequence: 2, pixelWidth: 600, pixelHeight: 800)
let c3 = makeCapture(sequence: 3, pixelWidth: 1000, pixelHeight: 1000)
let urls: [UUID: URL] = [
c1.id: try writePNG(for: c1, red: 30, green: 30, blue: 30, in: directory),
c2.id: try writePNG(for: c2, red: 40, green: 40, blue: 40, in: directory),
c3.id: try writePNG(for: c3, red: 50, green: 50, blue: 50, in: directory),
]
let output = directory.appendingPathComponent("no-annots.pdf")
_ = try PDFComposer().compose(
session: makeSession(captures: [c1, c2, c3]),
imageURL: imageURLMap(urls),
title: "No annotations",
to: output
)
let document = try openDocument(output)
#expect(document.pageCount == 3)
for index in 0..<document.pageCount {
let page = try #require(document.page(at: index))
#expect(page.annotations.isEmpty)
}
}
@Test("Mixed orientations coexist, unnormalised")
func mixedOrientationsCoexist() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let wide = makeCapture(sequence: 1, pixelWidth: 1600, pixelHeight: 900)
let tall = makeCapture(sequence: 2, pixelWidth: 900, pixelHeight: 1600)
let square = makeCapture(sequence: 3, pixelWidth: 1000, pixelHeight: 1000)
let urls: [UUID: URL] = [
wide.id: try writePNG(for: wide, red: 90, green: 10, blue: 10, in: directory),
tall.id: try writePNG(for: tall, red: 10, green: 90, blue: 10, in: directory),
square.id: try writePNG(for: square, red: 10, green: 10, blue: 90, in: directory),
]
let output = directory.appendingPathComponent("mixed.pdf")
_ = try PDFComposer().compose(
session: makeSession(captures: [wide, tall, square]),
imageURL: imageURLMap(urls),
title: "Mixed",
to: output
)
let document = try openDocument(output)
#expect(document.pageCount == 3)
let p0 = try #require(document.page(at: 0))
let p1 = try #require(document.page(at: 1))
let p2 = try #require(document.page(at: 2))
#expect(p0.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 842, height: 595))
#expect(p1.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 595, height: 842))
#expect(p2.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 595, height: 842))
}
@Test("No cover page, no before/after, exactly one page per capture")
func noCoverPageExactlyOnePagePerCapture() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let c1 = makeCapture(sequence: 1, pixelWidth: 800, pixelHeight: 500)
let c2 = makeCapture(sequence: 2, pixelWidth: 500, pixelHeight: 800)
let urls: [UUID: URL] = [
c1.id: try writePNG(for: c1, red: 12, green: 12, blue: 12, in: directory),
c2.id: try writePNG(for: c2, red: 22, green: 22, blue: 22, in: directory),
]
let output = directory.appendingPathComponent("two.pdf")
let pageCount = try PDFComposer().compose(
session: makeSession(captures: [c1, c2]),
imageURL: imageURLMap(urls),
title: "Two",
to: output
)
#expect(pageCount == 2)
#expect(try openDocument(output).pageCount == 2)
}
@Test("Small image is never upscaled")
func smallImageIsNeverUpscaled() throws {
let capture = makeCapture(sequence: 1, pixelWidth: 200, pixelHeight: 150)
let layout = PageLayout(capture: capture, pageIndex: 1, pageCount: 1)
#expect(layout.imageRect.size == CGSize(width: 200, height: 150))
}
@Test("Empty session throws, produces nothing on disk")
func emptySessionThrowsProducesNothing() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let output = directory.appendingPathComponent("empty.pdf")
var threwCompositionFailed = false
do {
_ = try PDFComposer().compose(
session: makeSession(captures: []),
imageURL: { _ in directory.appendingPathComponent("missing.png") },
title: "Empty",
to: output
)
} catch ShotdeckError.pdfCompositionFailed {
threwCompositionFailed = true
}
#expect(threwCompositionFailed)
#expect(!FileManager.default.fileExists(atPath: output.path))
let leftover = try FileManager.default.contentsOfDirectory(atPath: directory.path)
#expect(leftover.isEmpty)
}
@Test("A missing PNG is skipped, page count drops, document still opens")
func missingPNGIsSkipped() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let first = makeCapture(sequence: 1, pixelWidth: 1600, pixelHeight: 900)
let missing = makeCapture(sequence: 2, pixelWidth: 800, pixelHeight: 600)
let third = makeCapture(sequence: 3, pixelWidth: 900, pixelHeight: 1600)
let firstURL = try writePNG(for: first, red: 220, green: 10, blue: 10, in: directory)
let thirdURL = try writePNG(for: third, red: 10, green: 10, blue: 220, in: directory)
let missingURL = directory.appendingPathComponent("does-not-exist.png")
let urls: [UUID: URL] = [
first.id: firstURL,
missing.id: missingURL,
third.id: thirdURL,
]
let output = directory.appendingPathComponent("skip-missing.pdf")
let pageCount = try PDFComposer().compose(
session: makeSession(captures: [first, missing, third]),
imageURL: imageURLMap(urls),
title: "Skip missing",
to: output
)
#expect(pageCount == 2)
let document = try openDocument(output)
#expect(document.pageCount == 2)
let p0 = try #require(document.page(at: 0))
let p1 = try #require(document.page(at: 1))
#expect(p0.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 842, height: 595))
#expect(p1.bounds(for: .mediaBox) == CGRect(x: 0, y: 0, width: 595, height: 842))
// Distinct fill colors sampled from pdftoppm at each imageRect center.
let prefix = directory.appendingPathComponent("skip-missing")
try pdftoppm(pdf: output, prefix: prefix)
let page1PNG = directory.appendingPathComponent("skip-missing-1.png")
let page2PNG = directory.appendingPathComponent("skip-missing-2.png")
let bitmap1 = try RGBABitmap.loadPNG(at: page1PNG)
let bitmap2 = try RGBABitmap.loadPNG(at: page2PNG)
let k = 110.0 / 72.0
let layout1 = PageLayout(capture: first, pageIndex: 1, pageCount: 2)
let layout2 = PageLayout(capture: third, pageIndex: 2, pageCount: 2)
let c1 = try #require(bitmap1.pixel(
x: Int((layout1.imageRect.midX * k).rounded()),
y: Int(((layout1.pageRect.height - layout1.imageRect.midY) * k).rounded())
))
let c2 = try #require(bitmap2.pixel(
x: Int((layout2.imageRect.midX * k).rounded()),
y: Int(((layout2.pageRect.height - layout2.imageRect.midY) * k).rounded())
))
#expect(c1.0 > 150 && c1.1 < 80 && c1.2 < 80)
#expect(c2.2 > 150 && c2.0 < 80 && c2.1 < 80)
}
@Test("An unreadable (corrupt) PNG is treated exactly like a missing one (D-A)")
func unreadablePNGIsSkippedLikeMissing() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let good = makeCapture(sequence: 1, pixelWidth: 800, pixelHeight: 500)
let bad = makeCapture(sequence: 2, pixelWidth: 800, pixelHeight: 500)
let goodURL = try writePNG(for: good, red: 30, green: 140, blue: 30, in: directory)
let badURL = directory.appendingPathComponent("corrupt.png")
try Data([0x00, 0x01, 0x02, 0xFF, 0xD8, 0x00]).write(to: badURL)
let urls: [UUID: URL] = [good.id: goodURL, bad.id: badURL]
let output = directory.appendingPathComponent("skip-corrupt.pdf")
let pageCount = try PDFComposer().compose(
session: makeSession(captures: [good, bad]),
imageURL: imageURLMap(urls),
title: "Skip corrupt",
to: output
)
#expect(pageCount == 1)
let document = try openDocument(output)
#expect(document.pageCount == 1)
}
@Test("Document attributes round-trip exactly")
func documentAttributesRoundTrip() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let sessionID = try #require(UUID(uuidString: "deadbeef-dead-4eef-8ead-deadbeef0001"))
let capture = makeCapture(sequence: 1, pixelWidth: 640, pixelHeight: 480)
let png = try writePNG(for: capture, red: 40, green: 40, blue: 40, in: directory)
let output = directory.appendingPathComponent("attrs.pdf")
_ = try PDFComposer().compose(
session: makeSession(captures: [capture], id: sessionID),
imageURL: { _ in png },
title: "Ben review — 2026-08-30",
to: output
)
let document = try openDocument(output)
let attributes = try #require(document.documentAttributes)
#expect(attributes[PDFDocumentAttribute.creatorAttribute] as? String == "Shotdeck")
#expect(attributes[PDFDocumentAttribute.titleAttribute] as? String == "Ben review — 2026-08-30")
#expect(attributes[PDFDocumentAttribute.subjectAttribute] as? String == "deadbeef-dead-4eef-8ead-deadbeef0001")
#expect(document.pageCount == 1)
#expect(try #require(document.page(at: 0)).annotations.isEmpty)
}
@Test("Rendered page is not blank (pdftoppm, pinned renderer + rounding)")
func renderedPageIsNotBlank() throws {
let directory = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: directory) }
let capture = makeCapture(sequence: 1, pixelWidth: 1600, pixelHeight: 900)
let png = try writePNG(for: capture, red: 128, green: 128, blue: 128, in: directory)
let output = directory.appendingPathComponent("not-blank.pdf")
_ = try PDFComposer().compose(
session: makeSession(captures: [capture]),
imageURL: { _ in png },
title: "Not blank",
to: output
)
let prefix = directory.appendingPathComponent("not-blank")
try pdftoppm(pdf: output, prefix: prefix)
let rendered = directory.appendingPathComponent("not-blank-1.png")
let bitmap = try RGBABitmap.loadPNG(at: rendered)
let layout = PageLayout(capture: capture, pageIndex: 1, pageCount: 1)
let k: CGFloat = 110.0 / 72.0
let imageWin = pixelWindow(for: layout.imageRect, pageHeightPt: layout.pageRect.height, scale: k)
let headerWin = pixelWindow(for: layout.headerRect, pageHeightPt: layout.pageRect.height, scale: k)
#expect(bitmap.containsNonWhite(x0: imageWin.x0, y0: imageWin.y0, x1: imageWin.x1, y1: imageWin.y1))
#expect(bitmap.containsNonWhite(x0: headerWin.x0, y0: headerWin.y0, x1: headerWin.x1, y1: headerWin.y1))
}
@Test("Right-edge tick group geometry matches D-12 exactly, both orientations")
func rightEdgeTickGroupGeometry() throws {
let landscape = makeCapture(sequence: 1, pixelWidth: 1600, pixelHeight: 900)
let portrait = makeCapture(sequence: 1, pixelWidth: 900, pixelHeight: 1600)
let landscapeLayout = PageLayout(capture: landscape, pageIndex: 1, pageCount: 1)
let portraitLayout = PageLayout(capture: portrait, pageIndex: 1, pageCount: 1)
#expect(landscapeLayout.failTickBox.maxX == 824)
#expect(landscapeLayout.headerRect.maxX == 824)
#expect(portraitLayout.failTickBox.maxX == 577)
#expect(portraitLayout.headerRect.maxX == 577)
#expect(landscapeLayout.failTickBox.width == 13)
#expect(landscapeLayout.failTickBox.height == 13)
#expect(portraitLayout.failTickBox.width == 13)
#expect(portraitLayout.failTickBox.height == 13)
#expect(landscapeLayout.passTickBox.maxX == landscapeLayout.failLabelOrigin.x - 8)
#expect(portraitLayout.passTickBox.maxX == portraitLayout.failLabelOrigin.x - 8)
}
@Test("Writes visual sample for review")
func writesVisualSampleForReview() throws {
let verifyDir = URL(fileURLWithPath: "/Users/benjaminhippler/mmd-projects/multi-agent-runs/shotdeck-20260830/verify")
try FileManager.default.createDirectory(at: verifyDir, withIntermediateDirectories: true)
let scratch = try makeScratchDirectory()
defer { try? FileManager.default.removeItem(at: scratch) }
let wide = makeCapture(sequence: 1, pixelWidth: 1600, pixelHeight: 900)
let tall = makeCapture(sequence: 2, pixelWidth: 900, pixelHeight: 1600)
let retina = makeCapture(sequence: 3, pixelWidth: 2880, pixelHeight: 1620, scale: 2.0)
let urls: [UUID: URL] = [
wide.id: try writePNG(for: wide, red: 196, green: 42, blue: 42, in: scratch),
tall.id: try writePNG(for: tall, red: 32, green: 96, blue: 176, in: scratch),
retina.id: try writePNG(for: retina, red: 36, green: 148, blue: 84, in: scratch),
]
let output = verifyDir.appendingPathComponent("wp2-sample.pdf")
let pageCount = try PDFComposer().compose(
session: makeSession(captures: [wide, tall, retina]),
imageURL: imageURLMap(urls),
title: "WP-2 visual sample",
to: output
)
#expect(pageCount == 3)
let prefix = verifyDir.appendingPathComponent("wp2-sample")
try pdftoppm(pdf: output, prefix: prefix)
}