Build lookup indexes

The model’s collections are flat arrays whose cross-references are plain string ids (see IDs and references). For anything beyond one-off lookups, build a Map index once — a single O(n) pass that pays off on every subsequent lookup.

import { parseArchiModel } from '@cda/archi-semantic-core';

const model = parseArchiModel(xml);

const elementById = new Map(model.elements.map((e) => [e.id, e]));
const relationshipById = new Map(model.relationships.map((r) => [r.id, r]));
const folderById = new Map(model.folders.map((f) => [f.id, f]));
const objectById = new Map(model.diagramObjects.map((o) => [o.id, o]));
const profileById = new Map(model.profiles.map((p) => [p.id, p]));

Resolving relationship endpoints

A relationship’s source or target can be an element or another relationship (see Relationships), so resolve from both indexes:

function nodeName(id: string): string | null {
  return elementById.get(id)?.name ?? relationshipById.get(id)?.name ?? null;
}

Resolving the visual layer

A connection’s sourceId/targetId point at diagram objects, not semantic elements. Follow the chain:

function elementForObject(objectId: string) {
  const object = objectById.get(objectId);
  return object?.archimateElementId
    ? elementById.get(object.archimateElementId)
    : undefined;
}

View-local navigation

Views precompute their containment (see Views), so iterating a view’s diagram objects is direct:

function objectsInView(viewId: string) {
  const view = model.views.find((v) => v.id === viewId);
  if (!view) return [];
  return view.diagramObjectIds
    .map((id) => objectById.get(id))
    .filter(Boolean);
}

When to skip indexes

For a single traversal of one collection, find/filter are fine and simpler. Build indexes when you will look things up repeatedly — typically in any renderer, impact-analysis walk, or report generator.