OTI UML Json serialization
This project provides support for constructing a Json representation of a set of OTI Documents corresponding to UML Packages or Profiles. The Json representation is based on simple data structures corresponding to a normalized database schema where:
Every UML metaclass in the CMOF metamodel maps to a database table with a single primary key, the element's tool-specific ID and columns for all the scalar data properties in the metaclass (i.e. UML Properties typed by a primitive type or a UML Enumeration)
Every UML association in the CMOF metamodel maps to a database relation with two foreign keys corresponding to the primary keys of the directed association's first and second ends plus an index if the end is ordered.
This project provides a small library enabling a tool-specific OTI UML Adapter to map
Relationship to OMG's XMI serialization
At the OMG, UML models are serialized according to OMG's XMI serialization format, which is an XML Schema-based tree representation of the contents of the UML model. This serialization has been a significant source of complexity that led to several attempts to simplify it -- see for example OMG's XMI 2.5 Canonical XMI appendix.
Differences in terms of the serialization format:
XMI serialization is a tree, the depth of the tree corresponds to the nesting of UML element ownership
OTI UML Json serialization is very shallow; the depth is independent of the nesting of UML element ownership
Differences in terms of serialization/deserialization performance:
XMI serialization/deserialization is inherently slow because extensive cross-references within a single XMI document (i.e.
xmi:idref) limit the effectiveness of parallelism.
The normalized database tables & relations involved in OTI UML Json provide opportunities to exploit parallelism for serialization and deserialization.
Proof of concept example.
Loading a ~ 560K UML element (Europa Architecture + Framework, version from September 2015) in MagicDraw 18.0 SP5 takes ~ 4 minutes.
Serializing all the UML information for a significant subset of that model to Json is significantly faster:
JsonExportAsOTIDocumentSetConfiguration.overall (20 OTI document packages totalling 536596 elements) in 1 minute, 12 seconds, 982 milliseconds
Note that this serialization does not include (yet) the Json representation corresponding to stereotypes applied to UML elements and to their tag property values.