DunneFlow · The dashboard
Document
Two lanes in one pane, the eight documents DunneFlow writes about your program and every document your program already carries.
Document view has two lanes, and they are two different subjects:
- The map: the eight documents DunneFlow writes about your program.
- The program's own: every document your program already carries.
The switch at the top picks between them and remembers which you were in.
Lane 1: the map#
| Document | Contents |
|---|---|
| Where to start | Every way into the program, ranked by how much each commands. |
| Routines | Every routine, with what goes in and comes out, and what its author said. |
| Data structures | Classes, dataclasses, enums, type aliases. |
| Named values | The constants, and what they are. |
| Boundaries | Everything that leaves the program, by kind. |
| Database | Every table, its schema where the code states one, and who reads and writes it. |
| Network | What it calls out to. |
| Files the program names | Paths the code mentions. |
These read as documents, but every row is a selection. Click one and the graph lights, the inspector fills, the trail records the hop, and Back brings you home. A row naming a database table selects the table, which selects all the code that reads and writes it. As in the table, a banner says where your selection is, with Go to it.
Every analysis also writes Markdown to map@<instant>/: fourteen documents and an index,
including these eight plus flows, governance, anomalies, the glossary, naming, traceability
and the mirror, which have their own places in the dashboard.
Lane 2: the program's own#
Every Markdown, reStructuredText and plain-text document DunneFlow found beside the code, grouped by kind, with its size in words and sections. The filter box narrows by path or title.
The note at the top says whether any document relates to your current selection, and if not, why. DunneFlow relates code to prose through the plan, ticket and decision ids they both name, never by matching words, which would produce confident nonsense at scale.
Reading a document#
IN THIS DOCUMENT is the outline; click a heading to jump to it. Rendering makes some deliberate choices:
- No image is ever fetched. An image would fire a network request with no click and no consent, from a tool whose posture is that nothing leaves your machine.
- Relative links are checked against the corpus before they are followed. If one cannot be followed you are told, rather than sent somewhere blank.
- Raw HTML is not rendered.
- Author comments are folded by default. HTML comments are counted and can be shown, but they are not body prose.
Following a citation back#
A glossary term that quotes your documentation offers Read it in the document →, which opens the document at the line the definition is on. The button appears only where the stored corpus can actually serve the version on screen.
Something unclear or out of date? Tell us.