#Data-flow diagrams

#What it is

Where data comes from, which steps transform it, where it is stored, and who finally gets it.

#When to use it

A data pipeline, the handling of one request.

#How to write it

Pamphlet's own syntax — declarations first, relationships second; all seventeen kinds share this skeleton:

:::dataflow[Compiling a document]
nodes:
  author = "Author"
  parse = round "parse"
  render = round "render"
  cache = cylinder "diagram cache"
  reader = "Reader"
flows:
  author -> parse : plan.md
  parse -> render : AST
  cache -> render
  render -> reader : plan.html
:::

This syntax does not render on GitHub. If you need it to, write the same diagram as a Mermaid fence instead.

#What comes out

Drawn to SVG at compile time and inlined into the output, so the reader downloads no drawing library and makes no network request. Colours follow the theme; scroll to zoom, drag to pan, double-click to reset.

#Traps

• The block names are fixed: nodes: and flows: — note it is flows: here, not the flowchart's edges:
• The six shapes are shared with flowcharts: box for external actors, round for processes, cylinder for stores
• Edge labels are the data flowing (plan.md, AST), not the action
• Without {dir=...} the default is LR, unlike the flowchart's TD