• About
  • Built-in tag parsers
• Status
  • Feature ideas
• Installation
• Dependencies
• API
  • Basic example
• Syntax
  • Grammar
  • Node references
  • Prefixed IDs
  • Includes
  • EGF generation / serialization
• Authors
• License

Extensible Graph Format.

Striving to be both easily readable & writable for humans and machines, this line based, plain text data format and package supports:

• Definition of different types of graph based data (e.g. RDF-style or Labeled Property Graph topologies)
• Full support for cyclic references, arbitrary order (automatic forward declarations)
• Choice of inlining referenced nodes for direct access or via special node ref values
• Arbitrary property values (extensible via tagged literals and custom tag parsers a la EDN)
• Optionally prefixed node and property IDs with (also optional) auto-expansion via declared prefixes (for Linked Data use cases)
• Inclusion of sub-graphs from external files
• Loading of individual property values from referenced file paths
• Optionally GPG encrypted property values (where needed)
• Multi-line values
• Line comments
• Configurable parser behavior & syntax feature flags
• Hand-optimized parser, largely regexp free
• Configurable GraphViz DOT export

(Source for this example graph is further below)

The following parsers for tagged property values are available by default. Custom parsers can be provided via config options.

Note: In this reference implementation, the #file and #gpg tag parsers are only available in NodeJS.

ALPHA - bleeding edge / work-in-progress

Search or submit any issues for this package

You're strongly encouraged to update to at least v0.4.0 to avoid the potential of arbitrary code execution in older versions when decrypting #gpg-tagged property values. A security advisory will be published ASAP. A fix has been deployed already.

(Non-exhaustive list)

• VSCode syntax highlighting
• JSON -> EGF conversion
• Async tag parsing
• URL support for #file tag
• Tag declarations & tag parser import from URL (needs trust config opts)
• #md tag parser for markdown content
• #gpg fallback behavior options

yarn add @thi.ng/egf

ES module import:

<script type="module" src="https://cdn.skypack.dev/@thi.ng/egf"></script>

Skypack documentation

For Node.js REPL:

const egf = await import("@thi.ng/egf");

Package sizes (brotli'd, pre-treeshake): ESM: 2.63 KB

• @thi.ng/api
• @thi.ng/associative
• @thi.ng/checks
• @thi.ng/dot
• @thi.ng/errors
• @thi.ng/logger
• @thi.ng/prefixes
• @thi.ng/strings
• @thi.ng/transducers-binary

Generated API docs

TODO - Full docs forthcoming...

• api.ts - Data structures & options
• dot.ts - Graphviz export (via @thi.ng/dot)
• parser.ts - Main parser
• tags.ts - Tagged value parsers

; file: readme.egf

; prefix declaration (optional feature)
@prefix thi: thi.ng/

; a single node/subject definition
; properties are indented
; `thi:` prefix will be expanded
thi:egf
    type project
    ; tagged value property (here: node ref)
    part-of -> thi:umbrella
    status alpha
    description Extensible Graph Format
    url https://thi.ng/egf
    creator -> toxi
    ; multi-line value
    ; read as whitespace separated list/array (via #list)
    tag #list >>>
graph
extensible
format
linked-data
<<<

thi:umbrella
    type project
    url https://thi.ng/umbrella
    creator -> toxi

toxi
    type person
    name Karsten Schmidt
    location London
    account -> toxi@twitter
    account -> postspectacular@gh

toxi@twitter
    type account
    name @toxi
    url http://twitter.com/toxi

postspectacular@gh
    type account
    name @postspectacular
    url http://github.com/postspectacular

import { parseFile } from "@thi.ng/egf";

// enable prefix expansion in parser
const graph = parseFile("readme.egf", { opts: { prefixes: true } }).nodes;

console.log(Object.keys(graph));
// [
//  'thi.ng/egf',
//  'thi.ng/umbrella',
//  'toxi',
//  'toxi@twitter',
//  'postspectacular@gh'
// ]

console.log(graph.toxi);
// {
//   '$id': 'toxi',
//   type: 'person',
//   name: 'Karsten Schmidt',
//   location: 'London',
//   account: [
//     {
//       '$ref': 'toxi@twitter',
//       deref: [Function: deref],
//       equiv: [Function: equiv]
//     },
//     {
//       '$ref': 'postspectacular@gh',
//       deref: [Function: deref],
//       equiv: [Function: equiv]
//     }
//   ]
// }

// in this example inlining of referenced nodes is disabled (default)
// therefore refs are encoded as objects implementing the `IDeref` interface
// to obtain the referenced node
console.log(graph.toxi.account[0].deref());
// {
//   '$id': 'toxi@twitter',
//   type: 'account',
//   name: '@toxi',
//   url: 'http://twitter.com/toxi'
// }

EGF is a UTF-8 plain text format and largely line based, though supports multi-line values. An EGF file consists of node definitions, each with zero or more properties and their (optionally tagged) values. EGF does not prescribe any other schema or structure and it's entirely up to the user to e.g. allow properties themselves to be defined as nodes with their own properties, thus allowing the definition of LPG (Labeled Property Graph) topologies as well.

; Comment line

; First node definition
node1
    ; property with string value
    prop1 value
    ; property with reference to another node
    prop2 -> node2
    ; property with tagged value
    prop3 #tag value
    prop4 <<< long, potentially
multiline
value >>>
    prop5 #tag <<< tagged multi-line value >>>

node2
    ; property comment
    prop1 value
...

A full grammar definition is forthcoming. In the meantime, please see a somewhat outdated older version and related comments in #234 for more details.

Properties with reference values to another node constitute edges in the graph. References are encoded via property -> nodeid.

The following graph defines two nodes with circular references between them. Each node has a literal (string, by default) property name and a reference property knows to another node (via its ID). The order of references is arbitrary and the parser will automatically produce forward declarations for nodes not yet known.

alice
    name Alice
    knows -> bob

bob
    name Robert
    knows -> alice

Using default parser options, this produces an object as follows. Note, the references are encoded as objects with a $ref property and implement the IDeref and IEquiv interfaces defined in the @thi.ng/api package.

{
  alice: {
    '$id': 'alice',
    name: 'Alice',
    knows: {
      '$ref': 'bob',
      deref: [Function: deref],
      equiv: [Function: equiv]
    }
  },
  bob: {
    '$id': 'bob',
    name: 'Robert',
    knows: {
      '$ref': 'alice',
      deref: [Function: deref],
      equiv: [Function: equiv]
    }
  }
}

// access bob's name via alice
graph.alice.knows.deref().name
// "Robert"

If node resolution is enabled (via the resolve option) in the parser, the referenced nodes will be inlined directly and produce circular references in the JS result object. In many cases this more desirable and fine, however will stop the graph from being serializable to JSON (for example).

{
  alice: <ref *1> {
    '$id': 'alice',
    name: 'Alice',
    knows: { '$id': 'bob', name: 'Robert', knows: [Circular *1] }
  },
  bob: <ref *2> {
    '$id': 'bob',
    name: 'Robert',
    knows: <ref *1> {
      '$id': 'alice',
      name: 'Alice',
      knows: [Circular *2]
    }
  }
}

To enable namespacing and simplify re-use of existing data vocabularies, we're borrowing from existing Linked Data formats & tooling to allow node and property IDs to be defined in a prefix:name format alongside @prefix declarations. Such prefix IDs will be expanded during parsing and usually form complete URIs, but could expand to any string. The various (50+) commonly used Linked Data vocabulary prefixes bundled in @thi.ng/prefixes are available by default, though can be overridden, of course...

; prefix declaration
@prefix thi: http://thi.ng/

thi:toxi
    rdf:type -> foaf:person

Result:

{
  'thi.ng/toxi': {
    '$id': 'thi.ng/toxi',
    'http://www.w3.org/1999/02/22-rdf-syntax-ns#type': {
      '$id': 'http://xmlns.com/foaf/0.1/person'
    }
  },
  'http://xmlns.com/foaf/0.1/person': {
    '$id': 'http://xmlns.com/foaf/0.1/person'
  }
}

Currently in NodeJS only, external graph definitions can be included in the main graph via the @include directive. Any @prefix declarations in the included file will only be available in that file, however will inherit any pre-existing prefixes declared in the main file.

Relative file paths will be relative to the path of the currently processed file:

 |- include
 |  |- sub1.egf
 |  |- sub2.egf
 |- main.egf

(These examples make use of the schema.org ontology)

; main.egf
; declare an empty prefix
@prefix : http://thi.ng/

@include include/sub1.egf

; use empty prefix for this node
:toxi
    rdf:type -> schema:Person

; sub1.egf
@include sub2.egf

:sub1.egf
    rdf:type -> schema:Dataset
    schema:dateCreated #date 2020-07-19

; sub2.egf

:sub2.egf
    rdf:type -> schema:Dataset
    schema:creator -> :toxi

Parsing the main.egf file (with node resolution/inlining and pruning) produces:

{
  'http://thi.ng/sub2.egf': {
    '$id': 'http://thi.ng/sub2.egf',
    'http://www.w3.org/1999/02/22-rdf-syntax-ns#type': { '$id': 'http://schema.org/Dataset' },
    'http://schema.org/creator': {
      '$id': 'http://thi.ng/toxi',
      'http://www.w3.org/1999/02/22-rdf-syntax-ns#type': { '$id': 'http://schema.org/Person' }
    }
  },
  'http://thi.ng/toxi': {
    '$id': 'http://thi.ng/toxi',
    'http://www.w3.org/1999/02/22-rdf-syntax-ns#type': { '$id': 'http://schema.org/Person' }
  },
  'http://thi.ng/sub1.egf': {
    '$id': 'http://thi.ng/sub1.egf',
    'http://www.w3.org/1999/02/22-rdf-syntax-ns#type': { '$id': 'http://schema.org/Dataset' },
    'http://schema.org/dateCreated': 2020-07-19T00:00:00.000Z
  }
}

Complying JS objects can be converted to EGF using the toEGF() function. This function takes an iterable of Node objects, optional prefix mappings and an optional property serialization function to deal with custom tagged values. The default property formatter (toEGFProp()) handles various values for built-in tags and can be used in combination with any additional user provided logic.

import { rdf, schema } from "@thi.ng/prefixes";

const res = toEGF([
    {
      $id: "thi:egf",
      "rdf:type": { $ref: "schema:SoftwareSourceCode" },
      "schema:isPartOf": { $id: "http://thi.ng/umbrella" },
      "schema:dateCreated": new Date("2020-02-16")
    },
    {
      $id: "thi:umbrella",
      "rdf:type": { $ref: "schema:SoftwareSourceCode" },
      "schema:programmingLanguage": "TypeScript"
    }
  ],
  // prefix mappings (optional)
  {
    thi: "http://thi.ng/",
    schema,
    rdf
  }
  // property serializer (optional)
  toEGFProp
);

@prefix thi: http://thi.ng/
@prefix schema: http://schema.org/
@prefix rdf: http://www.w3.org/1999/02/22-rdf-syntax-ns#

thi:egf
    rdf:type -> schema:SoftwareSourceCode
    schema:isPartOf -> thi:umbrella
    schema:dateCreated #date 2020-02-16T00:00:00.000Z

thi:umbrella
    rdf:type -> schema:SoftwareSourceCode
    schema:programmingLanguage TypeScript

• Karsten Schmidt

If this project contributes to an academic publication, please cite it as:

@misc{thing-egf,
  title = "@thi.ng/egf",
  author = "Karsten Schmidt",
  note = "https://thi.ng/egf",
  year = 2020
}

© 2020 - 2024 Karsten Schmidt // Apache License 2.0