An NPM package that can be added to a Node.js/NPM codebase as follows:

$ npm add @lionweb/core

It contains:

• several base types
• the LionCore M3, including the builtins language
• functions for (de-)serialization

The following is a list of links to potential starting points:

• Implementation of the LionCore metametamodel (M3): see the specific README.
• Metamodel-generic/-aspecific code regarding:
  • TypeScript type definitions.
  • Representation of references.
  • Serialization.

• Add functions mapValues and instantiableClassifiers.
• Add a method SymbolTable.languageMatching to look up a language.

• The deserializer now deserializes all nodes in the serialization chunk, not just the root nodes (identified as having parent set to null). (This fixes issue #145.)
  • Also: deserializeChunk is renamed to deserializeSerializationChunk for naming consistency, although deserializeChunk is retained as an alias.
• Add a method metaType to M3 types, to deduce classifiers' names from. (That fixes issue #143.)

• Add a method Classifier.metaPointer.

• Introduce a isBuiltinNodeConcept function that checks whether a classifier happens to be the Node concept built into LionCore.
• Add a reference annotates to Annotation.

• Introduce a type SymbolTable to encapsulate lookup of languages' entities and their features.
  • Provide a naive and a memoising implementation of the SymbolTable abstraction, and use the latter in the deserializer to make that (a bit) more performant.

This is the first version corresponding to a release of LionWeb (version: 2023.1) as a whole.

• Remove JSON Schema generation functionality – this wasn't up-to-date at all.
• Implement persistence of annotations on Nodes, including (de-)serialization.
• Added an optional field resolveInfoFor to the ExtractionFacade type, to compute the field resolveInfo field of a serialized reference target. The default used is to check whether a node instance implements INamed and takes the value of the name field.
• Breaking change: serializeLanguage and deserializeLanguage are now "multi-lingual" in the sense that a serialization chunk can contain multiple Languages. As a result, these functions have been renamed to (the plural) serializeLanguages (with varargs), and deserializeLanguages (returning Language[]).
• (Bugfix:) Deserialize Annotations.
• Breaking change: Rework ID and key generation in the LionCore LanguageFactory. As part of that:
  • Add String[s]Mapper types, a StringsMapper instance last, and functions concatenator and chain to produce instances of them.
  • Remove M3Node.keyed and the "awkward" key generation in the InstantiationFacade for the LionCore language.
  • Remove types IdGenerator, and KeyGenerator and everything related to that.
  • Expose keys (instead of qualified names) for LionCore M3 and the LionCore built-ins.
• Add a type DynamicINamed and an M3-function conceptsOf, primarily for generation purposes.
• Move asText and ID checking to @lionweb/utilities.
• Expose a helper function nameSorted.
• Rename ConceptInterface → Interface.
• Change names: ReadModelAPI → ExtractionFacade, WriteModelAPI → InstantiationFacade.
• Change name of concept fields to classifier, and their types (where applicable) to Classifier.
• Fix license field in package.json to "Apache-2.0".
• Replace all occurrences of "LIon" (with uppercase 'I') with "Lion" (with lowercase 'i').
• Split the ModelAPI interface into a read- and write-part: ReadModelAPI vs. WriteModelAPI.
• Make fixes w.r.t. multi-language models.
• Migrate from Deno to Node.js.
• Implement import of enums in Ecore importer.

No changelog has been kept for previous versions, regardless of whether these were published or not.

Build it from source as follows:

npm run build