11.3K ·GitHub
What is this?
Perspective is an interactive data visualization and analytics component built for large, real-time and streaming datasets. It ships as a Custom Element UI (grid, charts, pivots) backed by a streaming query engine that runs in-browser via WebAssembly or server-side in Python, Node.js and Rust.
Perspective is an interactive data visualization and analytics component built for large, real-time and streaming datasets. It ships as a Custom Element UI (grid, charts, pivots) backed by a streaming query engine that runs in-browser via WebAssembly or server-side in Python, Node.js and Rust.Is this a UI widget or the whole analytics backend?It's both: a Custom Element/Web Component UI (with React bindings) plus the query engine itself, written in C++ and compiled to WebAssembly, Python and Rust, so the same engine can run client-side or server-side.
It's both: a Custom Element/Web Component UI (with React bindings) plus the query engine itself, written in C++ and compiled to WebAssembly, Python and Rust, so the same engine can run client-side or server-side.How do I add it to a web app or a notebook?Drop in the @perspective-dev/viewer web component (or the React bindings) and connect it to a Client that runs in-process, in a Web Worker, or over WebSocket to a server. For notebooks, install the Python package and use the Jupyter widget built on anywidget.
Drop in the @perspective-dev/viewer web component (or the React bindings) and connect it to a Client that runs in-process, in a Web Worker, or over WebSocket to a server. For notebooks, install the Python package and use the Jupyter widget built on anywidget.Do I need Python or just Node for the browser version?Pure client-side use only needs the JS/WASM package in a browser. Server-side use adds bindings for Python (aiohttp, Starlette, Tornado), Node.js or Rust, and building from source requires pnpm, binaryen, emsdk and LLVM.
Pure client-side use only needs the JS/WASM package in a browser. Server-side use adds bindings for Python (aiohttp, Starlette, Tornado), Node.js or Rust, and building from source requires pnpm, binaryen, emsdk and LLVM.Can it handle datasets bigger than a browser's WASM memory limit?Yes — there's a dedicated 64-bit memory64 WebAssembly build specifically for in-browser datasets larger than 4GB, and large datasets can also be virtualized server-side so only visible rows stream to the client.
Yes — there's a dedicated 64-bit memory64 WebAssembly build specifically for in-browser datasets larger than 4GB, and large datasets can also be virtualized server-side so only visible rows stream to the client.Can Perspective query a database directly instead of loading data into it?Yes, 'virtual servers' let Perspective's UI run directly against external engines like DuckDB, ClickHouse, PostgreSQL and Polars, translating view configs into native queries with no ETL or data copy required.
Yes, 'virtual servers' let Perspective's UI run directly against external engines like DuckDB, ClickHouse, PostgreSQL and Polars, translating view configs into native queries with no ETL or data copy required.What chart types and visuals does it support?It includes a WebGL charting engine with 15+ chart types, a virtual-scrolling editable data grid, tile-based geographic maps, pivot tables and full theming support.
It includes a WebGL charting engine with 15+ chart types, a virtual-scrolling editable data grid, tile-based geographic maps, pivot tables and full theming support.Can the client connect to a remote engine instead of running everything locally?Yes, the same Client API can connect in-process, inside a Web Worker, or remotely over WebSocket to a server-side engine, and datasets can be mirrored to the browser or virtualized server-side.
Yes, the same Client API can connect in-process, inside a Web Worker, or remotely over WebSocket to a server-side engine, and datasets can be mirrored to the browser or virtualized server-side.Does it support streaming updates, or just static datasets?Tables update incrementally and views tick in real time, with reactive joins across tables and streaming read/write support for Apache Arrow, CSV and JSON.
Tables update incrementally and views tick in real time, with reactive joins across tables and streaming read/write support for Apache Arrow, CSV and JSON.What are similar projects?redash, bokeh, superset, WrenAI, metabase, dataease, apexcharts.js, pyecharts
redash, bokeh, superset, WrenAI, metabase, dataease, apexcharts.js, pyechartsWhich topics is it in?bi, datagrid, webcomponents, business-intelligence, web-components, duckdb, clickhouse, jupyter, webassembly, charts, analytics, visualization
bi, datagrid, webcomponents, business-intelligence, web-components, duckdb, clickhouse, jupyter, webassembly, charts, analytics, visualization