Fluid: towards transparent, self-explanatory research outputs
summary
Fluid implements a bidirectional dynamic dependency analysis in an untyped functional language runtime that emits a dynamic dependence graph alongside results. the runtime records which inputs and intermediate values compute each chart point, exposes triangle operators to traverse input→output and output→input, and surfaces those relations in a JS D3 frontend so clicking a point shows the exact rows and columns used and selecting an input highlights outputs that depend on it. prototype works but is slow and still WIP.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:07Fluid programming language
- 2:53Traditional data visualizations limits
- 3:57Fluid demo interactive exploration
- 4:40Tracing input dependencies
- 6:25Dependency relevance in Fluid
- 7:30Fluid’s dynamic dependence graph
- 8:13Formalizing dependency navigation
- 10:23Brushing and linking in visualization
- 11:49Generalizing dependency operators
- 12:30Fine-grained queries projection
- 13:33Matrix convolution example
- 14:36Future plans and examples
- 14:56Q&A
related talks
21mLinked Visualisations via Galois Dependencies
30mFlux for Data Visualization: A Preliminary Manifesto
29mFluid - Build Data Orchestration in Kubernetes
34mLucky in the Cloud With Diagrams
30mVisualization Driven Development
37mVisualizing Persistent Data Structures
35mTOWARDS THE GRAMMAR OF ANIMATION
11mFlowco: Mixed-Initiative Authoring of Reliable End-to-End Data Analyses via Dataflow Graphs and LLMs