High Performance Data Visualization for the Web [PyCon DE & PyData Berlin 2024]

Perspective is a high-performance data visualization library combining C++ engine (compiled to WebAssembly) with Python backend and TypeScript/Rust UI plugins

Supports both static and streaming data with real-time updates, originally built for finance industry use cases

Key features include:

• Interactive filtering and sorting
• Cross-filtering between multiple linked visualizations
• Configurable views (grids, charts, maps)
• Virtualization for handling large datasets
• Apache Arrow for efficient data transfer

Architecture options:

• Pure client-side with embedded engine
• Client-server setup for larger datasets (4GB+)
• Streaming support via websockets

Empowers end users to explore data themselves through:

• Interactive pivoting and aggregations
• Configurable visualizations
• JSON-based view configurations that can be saved/loaded
• Familiar interface for Excel users

Built on modern web technologies:

• Web Components for UI
• WebAssembly for performance
• Apache Arrow for data transfer
• WebSockets for streaming

Integrates with Python ecosystem:

• FastAPI/Starlette for web server
• CSP library for stream processing
• Supports Kafka integration
• Panel/Streamlit integration options

Focuses on tabular/time-series data with support for:

• Real-time updates
• Large datasets through virtualization
• Complex aggregations and pivoting
• Multiple synchronized views of same data