Merge pull request #2 from martin-braun/xywh

allow unattended silent screenshot of given rect
adam-zethraeus · Jan 3, 2024 · 9afd193 · 9afd193
2 parents d5edf7f + ae67804
commit 9afd193
Show file tree

Hide file tree

Showing 4 changed files with 16 additions and 10 deletions.
diff --git a/README.md b/README.md
@@ -13,10 +13,11 @@ Takes file, stdin, or screencapture as input.
 > cd macOCR
 > swift build -c release
 > .build/release/ocr -h                                                                                      
-USAGE: ocr [--capture] [--stdin] [-i <i>]
+USAGE: ocr [--capture] [-r <r>] [--stdin] [-i <i>]
 
 OPTIONS:
   -c, --capture           Capture screenshot. 
+  -r <r>                  Rectangle to unattendedly capture (-r x,y,w,h), needs --capture.
   -s, --stdin             Read stdin binary data. 
   -i <i>                  Path to input image. 
   -h, --help              Show help information.

diff --git a/Sources/ocr/CLIArgs.swift b/Sources/ocr/CLIArgs.swift
@@ -7,12 +7,14 @@ struct CLIArgs {
         @Flag(name: [.customShort("c"), .long], help: "Capture screenshot.")
         var capture: Bool = false
 
+        @Option(name: [.customShort("r")], help: "Rectangle to unattendedly capture (-r x,y,w,h), needs --capture.")
+        var rectangle: String?
+
         @Flag(name: [.customShort("s"), .long], help: "Read stdin binary data.")
         var stdin: Bool = false
 
         @Option(name: [.customShort("i")], help: "Path to input image.", completion: CompletionKind.file(extensions: ["gif", "png", "jpg", "webp", "tiff"]))
         var input: String?
 
-
     }
 }
diff --git a/Sources/ocr/OCRApp.swift b/Sources/ocr/OCRApp.swift
@@ -14,14 +14,18 @@ class OCRApp {
     func run() {
         let args = CLIArgs.ocr.parseOrExit()
         let image: CGImage?
-        switch (args.input, args.stdin, args.capture) {
-        case (.some(let path), false, false):
+        switch (args.input, args.stdin, args.capture, args.rectangle) {
+        case (.some(let path), false, false, .none):
             image = imageFromURL(path.pathAsURL)
-        case (.none, true, false):
+        case (.none, true, false, .none):
             image = imageFromStdIn()
-        case (.none, false, true):
+        case (.none, false, true, .some(let rect)):
             let tempURL = URL(fileURLWithPath: "/tmp/ocr-\(UUID().uuidString).png")
-            image = imageFromURL(captureRegion(destination: tempURL))
+            image = imageFromURL(captureRegion(destination: tempURL, rect: rect))
+            try? FileManager.default.removeItem(at: tempURL)
+        case (.none, false, true, .none):
+            let tempURL = URL(fileURLWithPath: "/tmp/ocr-\(UUID().uuidString).png")
+            image = imageFromURL(captureRegion(destination: tempURL, rect: nil))
             try? FileManager.default.removeItem(at: tempURL)
         default:
             CLIArgs.ocr.exit(withError: ArgumentParser.ValidationError("Input not properly specified."))
@@ -86,7 +90,6 @@ class OCRApp {
         return image.cgImage(forProposedRect: &imageRect, context: nil, hints: nil)
     }
 
-
     func imageFromURL(_ inputURL: URL) -> CGImage? {
         if let ciImage = CIImage(contentsOf: inputURL),
            let cgImage = CIContext(options: nil)

diff --git a/Sources/ocr/ScreenCapture.swift b/Sources/ocr/ScreenCapture.swift
@@ -1,11 +1,11 @@
 import Foundation
 
-func captureRegion(destination: URL) -> URL {
+func captureRegion(destination: URL, rect: String?) -> URL {
     let destinationPath = destination.path as String
 
     let task = Process()
     task.launchPath = "/usr/sbin/screencapture"
-    task.arguments = ["-i", "-r", destinationPath]
+    task.arguments = rect != nil ? ["-R" + rect!, "-x", "-r", destinationPath] : ["-i", "-r", destinationPath]
     task.launch()
     task.waitUntilExit()