Compatibility philosophy

archi-semantic-core parses Archi’s native .archimate XML, the format Archi itself reads and writes on disk — not the ArchiMate® Model Exchange File Format. That choice is deliberate and deserves an explicit explanation.

Faithful semantics over standards

The purpose of this package is a faithful, typed semantic representation of the model as Archi actually stores it. Native Archi semantics — raw vs. semantic type names, Junctions’ type attribute, Access’ 0–3 codes, Influence’s free-text strength, Label Expressions, nested-relative geometry, Specializations — are part of the data, not an obstacle.

An exchange format (Open Exchange, .xml variants) exists to interoperate across tools, and in doing so it normalizes: AccessTypeEnum uses "Access" where Archi uses "Unspecified"; Label Expressions are rendered away or lost; Specializations and Profiles have no equivalent; nested-relative coordinates don’t exist. A parser for an exchange format must throw that information away or reinterpret it. This package was built to not do either.

If you need Open Exchange compatibility, the right architecture is a transformer layer on top of ArchiModel — reading this package’s output and emitting the exchange format — not a fork that parses exchange XML.

Original Archi’s quirks, preserved

The parser is written against Archi’s own source code and test models, not against a spec. Where the format is ambiguous, Archi’s behavior wins. When Archi introduced a breaking change (Label Expression escaping at Archi 4.4), the parser tracks the behavior, so old and new models resolve to the same text.

Compatibility goals

• Forward compatibility with Archi versions — a model saved by a newer Archi with previously-unknown element types or relationship attributes still parses; nothing is rejected because a name was not in a hardcoded catalogue.
• Exactness — unknown native values are never guessed (see Junctions for the clearest example).
• Honesty — boundaries and limitations are documented on Known limitations.