How JD Built a High‑Performance Cross‑Platform Image Library for HarmonyOS with Taro

JD launched the development of a HarmonyOS app in early 2024 and identified image loading performance as a critical bottleneck for its e‑commerce platform, prompting the creation of a custom image library.

Background

Because the app heavily relies on images, JD prioritized the image library as a core capability in the first phase of its HarmonyOS infrastructure.

Technical Implementation

Pre‑research

Investigation revealed that the HarmonyOS platform loads network images mainly via two approaches:

System – Image component

The built‑in component offers basic loading but suffers from limited performance, lack of AVIF support, no quality‑monitoring system, and no control over download, decode, or cache pipelines.

Third‑party – ImageKnife

Inspired by Android’s Glide, ImageKnife is implemented in ArkTS but still falls short in performance, stability, and extensibility.

Given these limitations, JD decided to develop its own image library using C++ as the core, enabling cross‑platform reuse for iOS and Android and reducing the overhead of frequent ArkTS ‑ C++ communication.

JDImage Capabilities

Architecture Design

To address cross‑platform compatibility, JD adopted a modular, layered architecture inspired by iOS/Android libraries such as SDWebImage, Glide, and Fresco. The library is split into a Core layer (implemented in C++) and a Client layer for platform‑specific adaptations.

Core Modules

The library consists of independent modules for cache, decoder, source fetching, performance monitoring, and the image component.

Image Cache

Memory cache uses an LRU algorithm to store decoded bitmaps and dynamically adjust memory usage.

Disk cache stores downloaded images in the app sandbox, supports multithreaded reads/writes, and provides size management.

Image Source Fetching

HTTP: Uses Cornet for high‑performance parallel network requests.

Base64: Directly decodes Base64‑encoded images.

Decoder

System decoder supports PNG, JPG, GIF, WebP, SVG, etc.

AVIF decoder integrates libavif to handle JD’s extensive AVIF usage.

Performance Monitoring

Load performance: records stage timings, cache hit rates, format usage, and resource consumption.

Exception monitoring: captures errors with unique codes and categorizes them as warnings or critical failures.

Image Component

Wraps the system Image component, rendering decoded PixelMap objects via the src attribute and offering extra properties such as lazy, sourceSize, and priority for performance tuning.

Performance Optimizations

Image Loading Pipeline

The pipeline, inspired by Fresco, allows flexible task ordering, module invocation, thread scheduling via HarmonyOS FFRT, duplicate task aggregation, load monitoring, and cancel/retry mechanisms.

Server‑Side Image Processing

JD’s image server can perform format conversion, quality reduction, scaling, rounding, and grayscale via URL parameters. The pipeline automatically appends these parameters (e.g., using .avif) to reduce network payload.

Quality Assurance

Online Monitoring

Exception monitoring with distinct error codes and severity levels.

Performance monitoring of load success rates and timing for each pipeline stage.

Exception Recovery

Pipeline retry with strategy based on error type.

Component fallback to the system Image component if custom decoding fails.

Online configuration‑driven degradation for specific users, devices, or app versions.

Future Plans

Network optimizations such as weak‑network handling and HTTPDNS.

Further performance gains: higher cache reuse, memory pool usage, GIF optimization, low‑end device tweaks.

Deeper integration with JD’s image server for dynamic scaling and format selection.

Expand support to iOS, Android, and additional platforms.

Enhance image quality via advanced interpolation, super‑resolution, HDR, etc.

Conclusion

JD’s self‑built HarmonyOS image library, through modular design and a flexible pipeline, resolves cross‑platform development challenges while delivering high performance and stability. The library now powers JD’s HarmonyOS app and sets a foundation for future multi‑platform expansion.