Compendium

Visual / Dataflow Languages Family Index

5 min read

Visual / Dataflow Languages Family Index

type: language-family-index family: visual-dataflow languages_catalogued: 22 tags: [language-reference, family-index, visual-programming, dataflow, no-code]

Visual / Dataflow Languages — Family Index

Family overview

Visual / dataflow languages express programs as directed graphs: nodes are operators (functions, filters, generators) and edges carry values, signals, or events. Execution is implicit in the topology — data arrives at a node, the node fires, results propagate downstream. This contrasts with the control-flow model of textual languages (sequence, branch, loop) and aligns naturally with signal processing, simulation, and reactive UI domains where the shape of computation mirrors the shape of the physical or perceptual system being modeled.

The family splits along two axes. Visual-only vs hybrid: pure graph editors (Pd, Quartz Composer, Origami) require all logic to live in patches; hybrids (TouchDesigner, Houdini, vvvv) embed textual escape hatches (Python, VEX, HLSL) for what graphs express awkwardly. Audience: engineering (LabVIEW G in test/measurement, Simulink in control systems, Modelica in multi-domain physics), creative-coding/music (Pd, Max/MSP, Reaktor, TouchDesigner), interaction-design (Origami Studio, ProtoPie), education (Scratch, Snap!, Blockly), and the no-code app-building strain (Bubble) which is dataflow-adjacent rather than strictly dataflow.

These languages have surprising industrial longevity. LabVIEW remains the lingua franca of physical-test labs forty years on. Simulink is mandatory tooling at every automotive Tier-1. Modelica drives EV powertrain and HVAC modeling at OEMs. The visual paradigm wins where domain experts (electrical engineers, mechanical engineers, sound designers, architects) are not full-time programmers but need executable models of their domain.

In our deep library

None of these are deep-noted — visual languages don’t reduce to grammar/semantics writeups the way textual ones do. Cross-reference adjacencies:

Tier 3 family table

LanguageFirst appearedDomainVendor / originStatus (2026)URL
LabVIEW G1986Test, measurement, instrumentationNational InstrumentsActive; dominant in T&Mhttps://www.ni.com/labview
Simulink1984Control systems, model-based designMathWorksActive; standard in automotive/aerohttps://www.mathworks.com/products/simulink.html
Pure Data (Pd)1996Real-time audio, multimediaMiller Puckette (open source)Active; FOSS sibling of Maxhttps://puredata.info/
Max/MSP1985 (Max), 1997 (MSP)Music, interactive mediaCycling ‘74Active; flagship for live audiohttps://cycling74.com/products/max
Quartz Composer2005Motion graphics, prototyping (macOS)AppleDeprecated; bundled in Xcode but unmaintainedhttps://developer.apple.com/library/archive/documentation/GraphicsImaging/Conceptual/QuartzComposerUserGuide/
vvvv2002Realtime graphics, installationsvvvv group (.NET-hosted)Active; “gamma” generation modernhttps://vvvv.org/
TouchDesigner2000Live visuals, projection, installationsDerivativeActive; widely used in AV productionhttps://derivative.ca/
Houdini VEX2002 (VEX 2)Procedural 3D, FX, shadingSideFXActive; textual-in-graph hybridhttps://www.sidefx.com/docs/houdini/vex/index.html
Modelica1997Acausal multi-domain physical modelingModelica AssociationActive; open standardhttps://modelica.org/
Modelica Standard Library (MSL)2002Reference component library for ModelicaModelica AssociationActive; v4.x currenthttps://github.com/modelica/ModelicaStandardLibrary
NoFlo2013Flow-based programming for JavaScriptHenri Bergius / communityMaintained; nichehttps://noflojs.org/
Node-RED2013IoT, integration, automation flowsIBM / OpenJS FoundationActive; ubiquitous in IoThttps://nodered.org/
OpenMusic1998Computer-aided compositionIRCAMActive; research-gradehttps://openmusic-project.github.io/
Reaktor Core2005 (Core layer)Modular synth / DSP designNative InstrumentsActive inside Reaktor 6https://www.native-instruments.com/en/products/komplete/synths/reaktor-6/
Bubble2012No-code web appsBubble GroupActive; no-code growthhttps://bubble.io/
Grasshopper2007Parametric / generative architectureRobert McNeel & Associates (Rhino)Active; AEC standardhttps://www.grasshopper3d.com/
Dynamo2011BIM / Revit visual scriptingAutodeskActive; bundled with Revithttps://dynamobim.org/
Scratch2007Education (children)MIT Media Lab Lifelong KindergartenActive; v3 web-basedhttps://scratch.mit.edu/
Snap!2011Education (with first-class procedures)UC BerkeleyActive; CS-principles courseshttps://snap.berkeley.edu/
Blockly2012Framework for block-based editorsGoogleActive; embeds in many toolshttps://developers.google.com/blockly
Origami Studio2014UI / interaction prototypingMeta (Facebook)Active but slowed; macOS onlyhttps://origami.design/
ProtoPie2017UI / interaction prototyping (cross-platform)Studio XIDActive; commercialhttps://www.protopie.io/

Notable threads

  • LabVIEW’s hardware moat. Python (with PyVISA, nidaqmx) has eaten general scripting in test labs, but LabVIEW persists because NI controls the GPIB/PXI/DAQ hardware stack and ships first-class drivers, FPGA targets (LabVIEW FPGA), and real-time deployment to PXI controllers. The language is the fence around the hardware ecosystem; replacing it means replacing instrumentation.
  • Modelica’s acausal trick. Unlike Simulink (causal — every block has fixed inputs/outputs), Modelica equations are acausal: a Resistor declares v = R*i and the compiler decides causality at simulation time. This lets engineers compose mechanical, electrical, hydraulic, and thermal components freely — a single circuit can mix all four domains, and the solver figures out the DAE. Foundation of EV powertrain modeling at most OEMs (via Dymola, OpenModelica, SimulationX).
  • Pd vs Max. Same author (Miller Puckette), divergent paths: Max became the polished commercial tool with broad adoption in interactive media and Ableton integration (Max for Live); Pd stayed FOSS and minimal, beloved in academic computer music and embedded audio (libpd ships inside countless apps and games). Patches are almost portable but not quite.
  • Flow-Based Programming as a recurring meme. J. Paul Morrison formalized FBP at IBM in the early 1970s — independent processes communicating over typed bounded queues. Largely ignored mainstream, but resurfaces every decade: NoFlo (2013), Node-RED (2013, IoT-flavored FBP), Apache NiFi (data-engineering-flavored FBP), and conceptually in modern stream processors. The idea that “components + ports + connections” generalizes better than function calls keeps proving correct in narrow domains.
  • The educational dataflow line. Scratch (2007) crystallized block-based programming for kids; Snap! extended it with first-class procedures, lambdas, and lists-of-lists (it’s genuinely closer to Scheme than to Scratch internally); Blockly is the framework underneath Microsoft MakeCode, App Inventor, Code.org, and dozens of vendor-specific editors. The blocks/text divide stops being binary in advanced cases — Snap! lets you write block-based recursion that would terrify a CS101 student.
  • No-code is mostly not dataflow. Bubble’s logic editor is event-driven imperative with a graph-shaped UI. Retool, Glide, Softr, etc. are form/table builders with formula languages. The genuinely dataflow no-code tools are rarer (n8n, Zapier paths, Pipedream are closer). Don’t conflate “visual builder” with “dataflow language” — most no-code is closer to spreadsheets-plus-UI than to Pd or Simulink.

Citations

Graph View