WP-2: PDF composer + PageLayout per SPEC-A1b (two-pass, atomic write, zero annotations)

This commit is contained in:
2026-08-31 14:42:00 +04:00
parent 97d21be470
commit 1b5816aa38
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))
}
}