anngvu/retold

Exploring interop between linked data modelling frameworks/formats

Hey there 👋

I'm a bioinformatics engineer at a Seattle-based company.

Many of my past projects are on GitLab.

I'm interested in knowledge graphs, knowledge engineering, well-designed systems, FAIR data, functional and pretty data viz.

What I'm currently LIFTing (Learning In Free Time)

• Rust
• Clojure

retold's Issues

Basic analysis -- entity deps

Add basic analysis functionality (primarily for schematic/LinkML data models):

Identify objects dependent on something else. So perhaps ability to run a command like bb ./retold --deps -m model.jsonld --entity ClassA to get everything that depends on ClassA.

Add `sms:` prefix to context

Though schematic doesn't add its sms: prefix to context, retold can probably add it when doing the translation, for better interop with other tools (e.g. TopQuadrant) that later need to work with the JSON-LD.

Basic analysis --stats

Add basic analysis functionality (primarily for schematic/LinkML data models:

Summarize numbers of props, classes, and enums overall (as well as per class? -- this might be more #19). bb ./retold --stats -m model.jsonld

Handle more translation of LinkML constraints to schematic validation rules syntax

Currently, data modelers who write LinkML but are schematic-conscious include schematic validation rules directly as annotations in YAML for the most straightforward interop. However, there are LinkML modelers who do not necessarily think of schematic JSON-LD as a main target and specify constraints in pure LinkML. So an example min-max constraint in LinkML like:

age:
  description: Human age in years.
  range: float
  minimum_value: 0
  maximum_value: 120

would be translated to schematic syntax: inRange 0 120 error.

However, not all translations are possible. A lot of translation from LinkML constraints to schematic validation rules will be approximate or impossible, so this will cover the easy subset that can be translated well.

Implement limited compilation of dynamic enums

See

• https://linkml.io/linkml/schemas/enums.html#dynamic-enums
• https://linkml.io/linkml/schemas/enums.html#tooling-to-support-dynamic-enums

The limited implementation would allow compilation of dynamic enums only from classes defined within the same data model and not an external ontology.

Implement translation of LinkML class inheritance for schematic JSON-LD

In LinkML, classes can inherit from a base or abstract class. This is a nice modeling feature that would be useful if it can be translated to schematic JSON-LD.

Handling of alternative `attributes` specification

Attributes are an alternative convenience feature in LinkML (see https://linkml.io/linkml/faq/modeling.html#when-should-i-use-attributes-vs-slots). Currently, retold does not handle classes that use attributes vs slots so the translation is incomplete.

Stricter generation of IDs during conversion to JSON-LD

We're currently converting things to IDs the way schematic does it, which is mainly by removing spaces. However, that still leaves illegal URIs as mentioned here ncihtan/data-models#77. Might be worthwhile to explore a flag feature (for situations when someone does care) where we have an alternate, stricter method of generating IDs.

Add version for JSON-LD

Though version is not officially in the schematic JSON-LD specifications retold can add it for downstream use.

Analysis --diff

Compare two data models to see diff, e.g. bb ./retold --diff -m modelA.jsonld modelB.jsonld