qr_scanner_fast 0.0.2

A low-latency hybrid QR scanner that races ZXing with an ML Kit fallback.

Changelog

All notable changes to this package are documented in this file.

0.0.2

Changed

• Explicitly declared Android and iOS support in the package metadata.

0.0.1

Initial release of qr_scanner_fast.

Added

• Native full-frame QR scanning for Android and iOS.
• A hybrid decoding strategy in which multiple decoders race and the first valid result wins.
• A live pipeline and a separate historical snapshot pipeline, both active from the moment the camera starts.
• A bounded in-memory snapshot buffer for checking QR codes that were visible only briefly while the camera or code was moving.
• Configurable frame sampling, snapshot sampling, fallback delay, decoder intervals, snapshot buffer size, duplicate cooldown, inverted-code attempts, and stop-after-detection behavior.
• QrScannerFastController with pause, resume, restart, torch, scanner status, and torch state controls.
• A customizable scanner overlay that does not restrict the analyzed camera area.
• QrScanResult with the decoded payload, decoder engine, and timestamp.
• A standalone isolate-based ZxingQrDecoder.decodeLuminance utility.
• Camera permission handling for Android and iOS setup documentation.
• A complete example application, unit tests, and English documentation with architecture diagrams and performance guidance.

Android

• Added a native CameraX preview and ImageAnalysis pipeline.
• Added native ZXing and Google ML Kit QR decoding.
• Added independent executors for live ZXing, snapshot ZXing, and snapshot ML Kit so decoding does not block the camera analyzer.
• Added KEEP_ONLY_LATEST backpressure and capacity-one live ZXing queues to keep latency bounded.
• Added sharpness scoring for historical snapshots. When the buffer is full, sharper recent frames can replace weaker frames.
• Added ML Kit potential-barcode detection and automatic zoom suggestions for QR codes that are visible but still too small to decode.
• Added potential-region mapping from ML Kit bounding boxes to priority ZXing crops with quiet-zone margins.
• Added a 1280x720 analysis target, continuous autofocus, and a preferred 30 FPS camera range when supported by the device.
• Added cleanup for CameraX bindings, ML Kit scanners, frame queues, permission waiters, method-channel handlers, and decoder executors when the platform view is disposed.

iOS

• Added a native AVFoundation preview and video-output pipeline.
• Added Apple Vision as the primary live QR locator and decoder.
• Added native Google ML Kit live and historical-snapshot decoding.
• Added a bounded historical snapshot ring processed independently from the live stream.
• Added Vision candidate-region cropping with quiet-zone margins.
• Added an isolate-based Dart ZXing fallback for grayscale candidate regions; full camera frames remain on the native side.
• Added continuous autofocus, continuous exposure, a preferred 30 FPS capture rate, torch control, and camera-session lifecycle handling.

Changed

• Replaced the original Flutter camera pipeline with native CameraX on Android and AVFoundation on iOS.
• Removed the Flutter camera, google_mlkit_barcode_scanning, and google_mlkit_commons dependency chain. Native ML Kit SDKs remain integrated directly by each platform.
• Removed obsolete Dart ML Kit worker code and unused camera configuration.
• Removed unused loadingBuilder, startWithTorch, and QrScanResult.displayValue API members.
• Simplified the Dart ZXing worker to accept tightly packed grayscale images, matching the active iOS ROI pipeline.

Requirements

• Flutter 3.38.0 or newer.
• Dart 3.10.0 or newer.
• Android API 24 or newer.
• iOS 15.5 or newer.