OLDM is now structured as a small npm-workspaces monorepo. The split keeps the core object mapping package tree-shakeable and dependency-light, while preserving a beginner-friendly package that exposes a single oldm object and installs globalThis.oldm for browser use.
| Package | Role |
|---|---|
@muze-nl/oldm-core |
Core object/graph mapping, helpers, and classes. Explicit ESM exports only. No N3 dependency and no global side effects. |
@muze-nl/oldm-n3 |
N3 parser/writer adapter for OLDM. Depends on n3 and @muze-nl/oldm-core. |
@muze-nl/oldm |
Beginner-friendly package. Exports one default oldm object, provides default N3 parser/writer wiring, browser bundles, and sets globalThis.oldm. |
For the friendly default package:
npm install @muze-nl/oldmimport oldm from '@muze-nl/oldm'
const context = oldm.context()For explicit, tree-shakeable use:
npm install @muze-nl/oldm-core @muze-nl/oldm-n3import oldm, { one, many, Collection } from '@muze-nl/oldm-core'
import { n3Parser, n3PatchWriter, n3Writer } from '@muze-nl/oldm-n3'
const context = oldm({
parser: n3Parser,
writer: n3Writer,
patchWriter: n3PatchWriter
})Create a literal independently of any graph:
const name = oldm.literal('Auke', { language: 'nl-NL' })
const plainName = oldm.literal('Auke', { language: '' })
const birthday = oldm.literal('1972-09-20', {
type: 'http://www.w3.org/2001/XMLSchema#date'
})
const text = oldm.literal('https://example.org/')The helper creates a fresh boxed string or number, preserving optional language
and datatype metadata. Supply a full datatype IRI, or explicitly expand one
with context.fullURI('xsd$date'). URL-shaped text remains a literal when passed to graph
write helpers. Core users can import { literal } from @muze-nl/oldm-core.
See the literal reference for
options and compatibility with existing graph.setLanguage() calls.
Only property names are shortened. Class values (subject.a) and literal
datatypes (value.type, when present) are full IRIs in both source graphs and
combined context views. Linked objects retain their full IRI in .id.
Property names in source graphs prefer source prefixes. Combined context views prefer client-provided prefixes, then OLDM defaults, then source prefixes. Namespace aliases, such as OLDM's schema.org HTTP/HTTPS mapping, apply to property-name shortening; value IRIs retain their original namespace.
const context = oldm.context({
prefixes: { person: 'http://xmlns.com/foaf/0.1/' }
})
const graph = context.parse(turtle, profileUrl, 'text/turtle')
const me = context.get(`${profileUrl}#me`)
me.a // 'http://xmlns.com/foaf/0.1/Person', or an array of class IRIs
const personClass = context.fullURI('person$Person')
oldm.many(me.a).includes(personClass)graph.set/add/delete() and context write helpers still accept shorthand class
inputs. They expand the input once using the receiving graph or context's
prefixes. graph.setType() and context.setType() do the same for datatypes.
Unresolved shorthand such as missing$Person is rejected at these write
boundaries. Relative class/datatype references on a graph resolve against its
URL; standalone values and context datatype setters require absolute IRIs.
Absolute IRIs, including URNs, are accepted without namespace rewriting.
context.sources(id, 'a', className) accepts a full IRI or shorthand, preferring
context prefixes and falling back to non-conflicting source prefixes. Matching
compares exact IRIs: HTTP and HTTPS namespaces are distinct. For repeated
filtering, expand the requested class once outside the loop.
Source subjects remain ordinary objects. When assigning subject.a or literal
metadata directly, supply full IRIs. Context subject assignment still delegates
to the context write helpers. Turtle output may use prefixes or relative IRIs
without changing the represented value.
fullURI(value) always returns a valid absolute IRI or throws TypeError.
It expands a known prefix by exact string concatenation, preserving the entire
suffix, including an empty suffix. Relative references use RFC 3986 base
resolution. Absolute inputs are validated without normalization: case, Unicode,
percent escapes, and namespace identity are preserved.
shortURI(iri) returns a prefixed spelling, or the unchanged absolute IRI when
no prefix matches. It never returns a relative reference or applies namespace
aliases. Aliases still apply when mapping predicates to object property names.
relativeURI(iri, baseURI) returns a relative reference only when resolving it
against the base reproduces the exact input IRI; otherwise it returns the
absolute IRI. The base argument is optional: graph helpers default to
graph.baseURI, initially the graph URL without its fragment. Context helpers
use the optional baseURI context setting. Without a base, relativeURI
retains the absolute IRI, and fullURI rejects relative inputs.
For a graph whose base is https://example.org/data/card, with foo mapped to
that same IRI:
graph.fullURI('foo$/child') // 'https://example.org/data/card/child'
graph.fullURI('card/child') // 'https://example.org/data/card/child'
graph.fullURI('/child') // 'https://example.org/child'
graph.shortURI('https://example.org/data/card/child') // 'foo$/child'
graph.relativeURI('https://example.org/data/card/child') // 'card/child'An unknown $ prefix throws; use ./price$tag for a relative path that contains
$. With : as the separator, an unknown scheme:value can still be a valid
absolute IRI (for example urn:example). These API strings have no Turtle angle
brackets or backslash escapes.
Both Turtle writers prefer a valid prefixed name, then an exact relative
reference, then an absolute IRI. They emit @base explicitly so relative output
keeps its meaning when the document is moved. Turtle local-name escaping is
handled by the writer: internal foo$/child becomes foo:\/child. N-Triples and
N-Quads continue to use absolute IRIs; PATCH output keeps its absolute fallback.
Replace comparisons such as many(subject.a).includes('foaf$Person') with a
comparison against context.fullURI('foaf$Person'). Compare literal datatypes
against full IRIs too. Shape declarations can keep shorthand class/datatype
inputs; oldm-shape resolves them against the subject's graph or context.
Pass full datatype IRIs to standalone literal() calls. Existing serialized
Turtle needs no migration. Application-owned JSON that contains shorthand class
or datatype values must resolve those values with its recorded prefixes when
loading. OLDM does not silently rewrite stored application documents.
A context keeps a registry of every parsed graph. Each context.parse() call still returns a Graph for the parsed resource, while the context exposes a combined view over all graphs loaded into that context.
const context = oldm.context()
const profile = context.parse(profileTurtle, profileUrl, 'text/turtle')
const settings = context.parse(settingsTurtle, settingsUrl, 'text/turtle')
profile.get(`${profileUrl}#me`) // data from only the profile graph
context.get(`${profileUrl}#me`) // merged data from all graphs
context.graphs // [profile, settings]
context.graph(profileUrl) // profile
context.data // combined subject list
context.subjects // combined subject map by full URI
context.sources(context.get(`${profileUrl}#me`))
// [profile, settings] when both graphs contain data for that subject
context.sources(context.get(`${profileUrl}#me`), 'vcard$fn')
// [profile] when only the profile graph contains that propertyThe combined view merges named subjects by IRI and keeps the original graph views separate. context.sources(subject, predicate, value) can be used to inspect which graph contributed a subject, property, or specific property value.
For source-aware writes, use the graph-specific helpers when you know the resource you want to edit:
profile.set(`${profileUrl}#me`, 'vcard$fn', 'Auke')
profile.add(`${profileUrl}#me`, 'schema$knowsAbout', 'Linked Data')
profile.delete(`${profileUrl}#me`, 'schema$knowsAbout', 'Old value')Or use context-level helpers with an explicit graph:
context.set(`${profileUrl}#me`, 'vcard$fn', 'Auke', { graph: profile })
context.add(`${profileUrl}#me`, 'schema$knowsAbout', 'Solid', { graph: profileUrl })When no graph is passed, context.set/add/delete() uses a conservative default: the subject's exact graph URL, the subject document URL without a fragment, the only graph that currently contains the subject, the configured defaultGraph, or the only graph in the context. Direct property assignment on a named subject from context.get(...) uses the same resolver, so simple edits can stay object-like:
const me = context.get(`${profileUrl}#me`)
me.vcard$fn = 'Auke'
delete me.vcard$nicknameIf there is no obvious source graph, OLDM throws and asks you to choose one explicitly with context.set/add/delete(..., { graph }) or graph.set/add/delete(...).
When a context has a patch writer, graph.patch() returns a Solid N3 Patch string for the changes between the original parsed source and the graph's current Turtle output.
const profile = context.parse(profileTurtle, profileUrl, 'text/turtle')
profile.set(`${profileUrl}#me`, 'vcard$fn', 'Auke C.')
const patch = await profile.patch()
await fetch(profileUrl, {
method: 'PATCH',
headers: { 'Content-Type': 'text/n3' },
body: patch
})The default @muze-nl/oldm package wires in the N3 patch writer. Direct @muze-nl/oldm-core users can pass patchWriter: n3PatchWriter.
Owned anonymous values are handled conservatively. Fresh blank nodes and RDF collections can be inserted. Changing or deleting an existing owned blank node or collection replaces the named-subject edge into that anonymous value and the complete old anonymous closure. Shared or cyclic anonymous values are rejected so callers can fall back to full document replacement.
The friendly package builds browser bundles into packages/oldm/dist/: an ESM bundle for module scripts and a classic global IIFE bundle for plain script tags.
For modern module scripts:
<script type="module">
import oldm from 'https://cdn.jsdelivr.net/npm/@muze-nl/oldm/dist/oldm.min.js'
const context = oldm.context()
</script>For a classic script tag that creates globalThis.oldm:
<script src="https://cdn.jsdelivr.net/npm/@muze-nl/oldm/dist/oldm.global.min.js"></script>
<script>
const context = oldm.context()
</script>Install dependencies from the repository root:
npm installRun all package tests:
npm testBuild the browser bundles for @muze-nl/oldm:
npm run build-dev
npm run buildRun coverage for all workspaces:
npm run coverageimport oldm from '@muze-nl/oldm'
const context = oldm.context({
prefixes: {
schema: 'https://schema.org/',
vcard: 'http://www.w3.org/2006/vcard/ns#',
foaf: 'http://xmlns.com/foaf/0.1/'
}
})
const url = 'https://example.org/profile/card#me'
const response = await fetch(url)
const text = await response.text()
const source = context.parse(text, url, 'text/turtle')
console.log(source.primary.vcard$fn)MIT. See LICENSE.
An experimental small Turtle 1.1 parser/writer adapter lives in packages/oldm-turtle as @muze-labs/oldm-turtle. It is not the default yet; see docs/TURTLE_PARSER_INVESTIGATION.md for supported syntax, size comparison, and small Solid-style benchmark results.