Public account of Tencent Music's development team, focusing on technology sharing and communication.
The article explains how to set up Jest for front‑end unit testing, demonstrates basic test writing, async handling, hooks, snapshot and React component testing, shows coverage configuration and thresholds, compares Jest’s built‑in features to Mocha’s limitations, and offers tips on concurrency, mocking, and test‑driven development.
The article explains musical beat detection fundamentals, detailing traditional onset‑strength and dynamic‑programming algorithms (as in librosa), compares time‑domain and spectral methods, showcases deep‑learning advances, and describes practical applications such as audio visualisation, rhythm games, and QQ Music’s Super‑DJ automatic remix pipeline.
The article reviews four mobile‑focused machine‑learning frameworks—NCNN, TensorFlow Lite, PyTorch Mobile (Caffe2) and FeatherKit—detailing their size, speed, and resource trade‑offs, and explains Q Music’s edge‑inference pipeline, optimization strategies, and the challenges of performance variability on heterogeneous mobile devices.
Edge deep learning inference on mobile devices faces hardware and software fragmentation, diverse CPUs, GPUs, DSPs, and NPUs, and limited programmability; optimization techniques such as model selection, quantization, and architecture‑specific tuning enable real‑time performance, with most inference on CPUs, GPUs offering 5–10× speedups, and co‑processor support varying across Android and iOS.
This article walks through Android TV’s Leanback library—its purpose, how to add the dependency, and core components such as BaseGridView, Presenter, and ObjectAdapter—using the QQ Music TV app as a case study, while weighing its design advantages against maturity and update‑frequency drawbacks.
The article explains how resource obfuscation with andResGuard removes META‑INF service files needed for the Android Main dispatcher, causing coroutines to silently fail after withContext calls, and shows that preserving the META‑INF/services directory restores proper coroutine execution.
The article explains BufferQueue’s internal design and shows how Android developers typically use its wrapper classes—Surface as the producer and SurfaceTexture as the consumer—detailing their constructors, dequeue/queue workflows, lock/unlock mechanisms, and a complete SurfaceView example that illustrates buffer production and consumption by SurfaceFlinger.
Android’s BufferQueue, the core producer‑consumer mechanism behind the display pipeline, manages GraphicBuffer slots through explicit FREE‑DEQUEUED‑QUEUED‑ACQUIRED transitions, using shared memory and Binder to let producers (e.g., Views, MediaCodec) enqueue frames and consumers (e.g., SurfaceFlinger) acquire and release them efficiently.
This article analyzes the challenges of reproducing external‑network bugs, outlines common failure causes, and presents a complete design for a JavaScript SDK that records environment data, AJAX calls, errors, and user actions, stores them in IndexedDB, and visualizes the timeline for efficient troubleshooting.
The project replaces karaoke effects with ASS‑based lyrics on Android, using a backend to generate ASS files and a JNI renderer that rasterizes and composites bitmap fragments; optimizations like transparent‑fragment filtering, opacity early‑out, NEON SIMD, copy elimination, and double‑buffered decoding cut frame time from 52 ms to under 3 ms, raise frame rate from 7 fps to 15 fps, and shrink memory usage from 180 MB to under 10 MB.
