Skip to content

Canonicalization

Before owlcompare compares anything, it canonicalizes both ontologies into a normalized form. Without this step, every diff would be dominated by spurious blank-node relabeling and triple-reordering noise. This page will document each normalization stage and why it matters. You can inspect the output of this stage directly with owlcompare canonicalize.

This page is being expanded

The outline below anchors the scope; the full per-stage detail is coming.

What this page will cover

  • Why canonicalize at all — making two semantically-equal inputs produce byte-identical output, so the diff reflects meaning, not serialization.
  • Blank-node normalization — stable, content-based labels (RDFC-1.0), so anonymous nodes stop generating phantom changes.
  • Restriction reification — giving anonymous owl:Restriction structures deterministic identifiers so they can be compared.
  • RDF-list collapsing — turning rdf:first/rdf:rest chains into ordered lists.
  • Triple sorting — removing order sensitivity where order is semantically irrelevant.
  • Anonymous-structure decoding — owl:unionOf / owl:intersectionOf sets, datatype facet restrictions, and dcterms:isReplacedBy, decoded into structured changes instead of raw blank-node noise.
  • The --no-* flags — disabling individual stages for debugging.
  • Known limitations — named graphs and quad sources in v1.