Repository navigation
Collection instances (columns blocks) and generic arrays (type Array, class Name < Array) #42
Closed as not planned
Description
Context
RS 3001 §Instance collection supports named collections with planned instances (columns { }), and generic array types (type Array, class Name < Array). None are implemented; today only ad-hoc named instance collections (instances "Name" { … }) exist.
Scope
- grammar:
columns { }planned-instance sugar; array generic types;< Arrayparent form (parent_class already parses< name) - model/builder: typed collections; expansion into hydrated instances
- executor/CSV: column mapping against planned columns
- RS 3001 alignment check: §Instance collection + §Value collection semantics
- specs + fixtures
Activity
[-]Collection instances () and generic arrays (, )[/-][+]Collection instances (columns blocks) and generic arrays (type Array, class Name < Array)[/+]ronaldtse commentedon Oct 6, 2026
Determination (audit against the authoritative spec):
columns {},type Array, andclass Name < Arrayare not in RS 3001 — the spec's par. Instance collection defines onlyinstances {name} {block}withinstancemembers, all of which is implemented and now vendored as corpus fixtures 21–23 (PR #48).These constructs are therefore extension-track: they need spec adoption first (the same audit-then-amend process as visibility in #38) before lutaml-lml implements them. Repurposing this issue to track that adoption; the EA/XMI tabular round-trip demand is the motivating use case to research.
ronaldtse commentedon Oct 7, 2026
Audit conclusion (2026-10-07): no grounded demand — closing as not planned.
Researched every plausible consumer:
columns {}or generic arrays.lutaml-yaml2uml(the tool whose dialect inspired this item): it converts the metanorma-style datamodel view YAML (relations/classes structure) into LutaML text — no tables, no columns anywhere. lutaml-uml's own TODO (TODO.next/08-broken-executables.md) additionally flags it as a misplaced executable unrelated to that gem.ExecutorCSV adapter does column-mapped CSV import → hydrated instances, outside the grammar — which is where tabular data belongs (an interchange concern, not a modeling-language construct).type Array/class Name < Arraylikewise have no claiming consumer in the family today. If a concrete demand appears (e.g., a consumer needing typed array attributes in normative models), reopen with that use case and run the visibility-style adoption process: audit → spec amendment → implementation.