Each input line is a chain of tasks separated by >, each task ACTION TARGET ATTRIBUTE…. Map actions to view types: summarize, discover, explore, browse → overview; identify, lookup, locate → detail; compare → comparison; present → presentation; annotate → annotation layer; record → provenance log; derive → new attribute. Each view type has a requirement text:
- overview:
an overview of all items (aggregate or dense chart) - detail:
selection of one item + a detail view - comparison:
juxtaposed or superimposed views of the compared items - presentation:
a separate, static, annotated artifact for another audience - annotation layer:
notes attached to marks; provenance log:saved states of the analysis; new attribute:a computed column before encoding
For each step print 3. identify features district -> detail: selection of one item + a detail view. Then interactive views: 3 (overview, detail, comparison) (distinct non-presentation view types in order of first use), links needed: overview->detail, detail->comparison (each change of view type between consecutive steps, ignoring changes to/from presentation, without repeats; or none), and, if the chain contains present, note: the presentation step needs a different artifact, not the dashboard. Print an empty line after each chain.