Picture two reviewers staring at the same iXBRL filing. One squints at the on-screen rendering and says the report looks fine. The other has Arelle open in a second window and frowns at a presentation tree that does not match the cash-flow statement. They are not really looking at the same thing. The reviewer with the validator is reading a graph; the reviewer in the browser is reading a typeset PDF that happens to know XML.
Linkbases are how XBRL stops being a flat list of facts and starts being a structured report. If concepts are the nouns of the language, linkbases are the grammar that holds them together: which concept is a child of which, which adds up to which, which has a label in Dutch, which cites IAS 38 paragraph 24. There are five standard linkbases plus the footnote linkbase that lives inside instances, and every one of them is, mechanically, a directed graph.
The graph has three building blocks, all of them XLink primitives. A locator (link:loc) is a node that points at a concept declaration through an xlink:href. A resource (link:label, link:reference, link:footnote) is a node carrying inline content, like a label string or a reference to an accounting standard. An arc (link:presentationArc, link:calculationArc, and so on) is an edge that joins two nodes, and the edge wears an xlink:arcrole URI that names the relationship: parent-child for presentation, summation-item for calculation, hypercube-dimension for definition wiring, concept-label for labels.
Around the arcs sit extended links (link:presentationLink, link:calculationLink, ...), each marked with an xlink:role that names an Extended Link Role, an ELR. The ELR is what lets a single linkbase carry a Statement of Financial Position, a Statement of Profit or Loss, and a Statement of Cash Flows side by side without their arcs bleeding into each other.
It is worth borrowing five words from graph theory here, because they make the rest of XBRL fall into place and they are the same words computer scientists use when they talk about networks. A graph is just a set of vertices (the dots) joined by edges (the lines). When the edges have a direction, an arrow rather than a plain line, the graph is directed, and a linkbase arc is always directed: it runs from a parent to a child, from a summation total to its operand, from a concept to its label, never ambiguously both ways. When the edges carry a tag, the graph is labeled, and the arcrole URI is exactly that tag. A sequence of edges you can follow nose-to-tail is a path, and the set of everything reachable from a starting vertex by following paths is, formally, its reachable set. That last idea is not academic: the Discoverable Taxonomy Set from the previous chapter is precisely the reachable set of documents you get by starting at one entry-point schema and following every reference edge.
Two special shapes of directed graph matter enough to name. A tree is a graph where every vertex has exactly one parent (except a single root with none), so there is exactly one path from the root to any node. The presentation linkbase is authored as a tree, which is why it renders cleanly as an indented outline. A directed acyclic graph, a DAG, relaxes the tree rule: a vertex may have several parents, but you still can never return to where you started by following the arrows. The calculation linkbase must be a DAG, because a calculation that could loop back on itself, A sums B, B sums C, C sums A, describes an arithmetic with no answer. A loop like that is a cycle, and a cycle in a summation graph is not a style choice but a defect a validator will reject. Hold the hierarchy in mind: every tree is a DAG, and every DAG is a directed graph, but not the other way around.
So the question to ask, in front of any linkbase artifact you do not understand, is: which graph (which linkbase), which arcrole, and which ELR? The keyPoints below build that habit. Hold those three coordinates in your head and most of the rest follows.
Every linkbase is a directed graph. Locators are nodes, arcs are edges, extended links partition the graph by role (an Extended Link Role, or ELR).
An arc carries an arcrole URI that names the relationship: parent-child for presentation, summation-item for calculation, hypercube-dimension for definition, concept-label for labels.
Each linkbase has its own arc element type: link:presentationArc, link:calculationArc, link:definitionArc, link:labelArc, link:referenceArc, link:footnoteArc.
An ELR partitions the graph into per-statement views. The Statement of Financial Position has its own ELR; the Statement of Profit or Loss has another. This is how you switch between views in a renderer.
When in doubt, ask three questions: which graph (which linkbase), which arcrole, which ELR. The answers tell you what the relationship actually means.
<link:presentationLink xlink:type="extended"
xlink:role="http://example.org/role/StatementOfFinancialPosition">
<link:loc xlink:type="locator"
xlink:href="ifrs.xsd#ifrs-full_Assets" xlink:label="assets"/>
<link:presentationArc xlink:type="arc"
xlink:from="assets" xlink:to="assets-current"
xlink:arcrole="http://www.xbrl.org/2003/arcrole/parent-child"/>
</link:presentationLink> Assets
| arcrole = parent-child, order = 1
v
CurrentAssets ------------------+
| order = 1 | order = 2
v v
CashAndCashEquivalents TradeReceivables
vertex = a concept declaration (a link:loc points at it by xlink:href)
edge = a link:presentationArc from parent to child
label = the xlink:arcrole URI (here, .../arcrole/parent-child)
direction matters: the arrow runs parent -> child, never the reverse PRESENTATION CALCULATION DEFINITION
(parent-child) (summation-item) (dimension wiring)
Equity o o weight +1 o---[axis]---o
| ^ |
Reserves o Revenue o---+ has-hypercube /
| ^ dimension-domain
RetainedE. o Expenses o---+ weight -1 o---[member]-o
Same vertices. Different edge sets. A concept can be a child here,
a summation operand there, and a hypercube primary item in the third. TREE (presentation) DAG (calculation: shared child, still acyclic)
A A B
/ \ \ /
B C \ /
/ \ v v
D E C <- C feeds both A and B; no cycle
CYCLE (illegal in a summation graph)
A --> B
^ |
| v
+---- C A sums B, B sums C, C sums A: no value can settle.3 questions. Druk op 1 tot 4 om te kiezen, Enter om te bevestigen.
Door deze quiz af te ronden, markeer je deze sectie als voltooid.
Tip: kies met 1 tot en met 4 en bevestig met Enter.