Skip to content

Latest commit

 

History

History
138 lines (96 loc) · 4.58 KB

README.querylang.md

File metadata and controls

138 lines (96 loc) · 4.58 KB

Grano Query Language

This repository contains an experimental query language implementation for grano. It is intended to augment the existing REST API with a more advanced way of accessing data.

Grano QL is inspired by MQL, the Metaweb Query Language used to query Freebase. Its query-by-example approach seems more appropriate for a web interface than SQL-inspired query languages such as Neo4J's CYPHER or RDF's SPARQL.

WARNING: At this moment, the plugin does not implement authorization checks. When enabled, it will make all data in your grano instance accessible, regardless of permissions.

Comments and Feedback

This document and the implementation in this repository are requests for comments - none of the features are fixed at this point; and any concerns by users are valuable feedback, even if they imply significant changes to the language.

Submitting queries

When installed the Grano QL API endpoint is available at:

  /api/1/query

Queries can be submitted via HTTP GET or POST request. For GET requests, a JSON string is expected to be submitted in the query query string argument. POST requests are expected to carry the payload as the body, using application/json as a content type.

Basic queries

A simple query could look like this:

{
  "properties": {
    "name": "Barack Obama"
  },
  "id": null
}

Which means: get the id of the entity with the name Barack Obama.

All grano QL queries run against entities, although we'll see below how they can be used to retrieve relations.

As seen above, submitting null for any field will attempt to retrieve it. As a shortcut, a default set of fields can be retrieved using a wildcard:

{
  "*": null
}

Similarly, a filter can be set by submitting a value for a given field:

{
  "id": "xxx",
  "*": null
}

This will retrieve the default set of fields for the entity with the specified id.

Lists vs. object retrieval

To further specify the desired return type, empty lists can be used to signify whether the query should return a single item or a list of items:

{
  "schemata": {} // get the first available item
  // vs.
  "schemata": [{}] // get all available items
}

This also applies to the root of the query object: queries can either aim to retrieve one or many entities. This will get all entities (pagination is used):

[{
  "*": null
}]

Available fields

Based on these simple entity queries, the grano graph can be traversed. The following fields are available:

  • id, status, created_at, updated_at as simple fields on entities.
  • author is an account, with these properties:
    • login the login name
    • full_name, id, created_at, updated_at
  • project the associated project, with these properties:
    • slug, label, id, created_at, updated_at
  • schemata a set of schemata, each with these properties:
    • name, label, hidden, created_at, updated_at
  • properties can have a set of properties defined as nested objects, with these properties on each:
    • name, value, source_url, active
  • inbound has a list of inbound relations. Each relation has these fields:
    • id, created_at, updated_at
    • schema, author and project, analogous to those on entities.
    • source is an entity, with all it's properties available.
  • outbound, a list of outbound relations. Same as inbound, except with a target entity instead of source.

Some of the nested objects can be abbreviated when filtering, e.g. you can filter for {"project": "test"} instead of {"project": {"slug": "test"}}. This applies to schemata names, author logins and property values.

Discussion

What's cool about this type of query:

  • Readable JSON, easily constructed by a web frontend application.
  • Resembles the representation in the REST API; query and result are basically the same thing.
  • Granular access to individual properties, or constellations of objects.

Potential Problems:

  • How do we tell the difference between null as in "return this value" and null as in "this property is null"?

Indexes

CREATE INDEX dbg1 ON grano_entity (project_id); CREATE INDEX dbg2 ON grano_entity_schema (entity_id); CREATE INDEX dbg3 ON grano_entity (id);