You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Clarify purpose of hierarchy, consider a more general relationships construct to complement it. #31
Status: WIP on extending hierarchy to support a "layered DAG" construct. More general relationships approach for non-hierarchy rels TBD
Schema metadata includes a 'hierarchy' attribute, used for hierarchy validation.
I think this basically serves as an optional convenience to get hierarchy validation automatically, but schemas generally don't depend on it.
A more graph-based schema (or hybrid with partial hierarchy) in theory should also work, but the onus is currently on the schema maintainer to enforce those relationship rules another way.
It might make sense to also allow expression of relationships with corresponding automatic validation.
Will need to think about how relationships are expressed in markdown and in schema. A simple start might be to write some jsonata rules that are parameterized, which could be generally useful or render the need for a specific relationships spec unnecessary
changed the title [-]Clarify purpose of hierarchy, consider a more general relationships construct to complement it.[/-][+]Clarify purpose of hierarchy, consider a more general relationships construct to complement it.
Will need to think about how relationships are expressed in markdown and in schema. A simple start might be to write some jsonata rules that are parameterized, which could be generally useful or render the need for a specific relationships spec unnecessary[/+]on Mar 7, 2026
changed the title [-]Clarify purpose of hierarchy, consider a more general relationships construct to complement it.
Will need to think about how relationships are expressed in markdown and in schema. A simple start might be to write some jsonata rules that are parameterized, which could be generally useful or render the need for a specific relationships spec unnecessary[/-][+]Clarify purpose of hierarchy, consider a more general relationships construct to complement it.[/+]on Mar 7, 2026
While more general relationships are good to think about, the more proximate opportunity is to expand from a hierarchical tree to a hierarchical/layered DAG. https://g.co/gemini/share/5ff3bbec3ffb
In short, we keep the hierarchy concept, add a way to specify, express, validate and visualise the layers that have many to many relationships.
Specify:
hierarchy array (existing)
a way to indicate per node type pair which ends specifies (parents specify array of children or children specify array of parents), and field name
Express: single wikilink or array matching above
Validate: ensure resolved nodes models and accurately reflects (I'm thinking it should normalise children refs to parent refs on child nodes); ensure we only read relationships as specified ( if expecting children don't allow parents on the other end)
Visualise: ok for Miro, show command may need a dag based format instead of bullets, or to denormalise.
This issue may also depend on #33 (Support full cross-schema resolution of refs) - if we want relationships to reference schema definitions across files, we'll need full cross-schema resolution first.
Activity
[-]Clarify purpose of hierarchy, consider a more general relationships construct to complement it.[/-][+]Clarify purpose of hierarchy, consider a more general relationships construct to complement it. Will need to think about how relationships are expressed in markdown and in schema. A simple start might be to write some jsonata rules that are parameterized, which could be generally useful or render the need for a specific relationships spec unnecessary[/+][-]Clarify purpose of hierarchy, consider a more general relationships construct to complement it. Will need to think about how relationships are expressed in markdown and in schema. A simple start might be to write some jsonata rules that are parameterized, which could be generally useful or render the need for a specific relationships spec unnecessary[/-][+]Clarify purpose of hierarchy, consider a more general relationships construct to complement it.[/+]mindsocket commentedon Mar 8, 2026
While more general relationships are good to think about, the more proximate opportunity is to expand from a hierarchical tree to a hierarchical/layered DAG. https://g.co/gemini/share/5ff3bbec3ffb
In short, we keep the hierarchy concept, add a way to specify, express, validate and visualise the layers that have many to many relationships.
Specify:
Express: single wikilink or array matching above
Validate: ensure resolved nodes models and accurately reflects (I'm thinking it should normalise children refs to parent refs on child nodes); ensure we only read relationships as specified ( if expecting children don't allow parents on the other end)
Visualise: ok for Miro, show command may need a dag based format instead of bullets, or to denormalise.
mindsocket commentedon Mar 9, 2026
Reviewing the collection of modelling issues, here are some observations:
Scope Overlaps & Gaps:
Suggestions:
mindsocket commentedon Mar 9, 2026
This issue may also depend on #33 (Support full cross-schema resolution of refs) - if we want relationships to reference schema definitions across files, we'll need full cross-schema resolution first.
feat: implement layered DAG support for multi-parent hierarchies (#31)
mindsocket commentedon Mar 17, 2026
Most of this has been address in #52 . New issues cover potential refactor and extension to ways of supporting relationships.