primal-grakn

A convenience wrapper around the official grakn-python client.

Features

• Less code / boilerplate.
• Response data looks and acts like primitive data structures (python lists / dicts etc.). Thus more immediately intelligble, accessable, and JSON-serialisable.
• Some added conveniences such as match_or_insert function.
• Still access all underlying grakn-python client functionality where needed.

Why

The grakn-python client provides a complete and efficient object-oriented method of interaction with a Grakn instance. It can require a lot of code and recursion to get data. This extension aims to provide convenience through reducing code involved in connecting to Grakn and working with response data. It reflects a manner of working with Grakn through python that I have found to be preferrable.

Usage

Example

import primal_grakn.primal_grakn as grakn

with grakn.Graph(uri='myuri', keyspace='mykeyspace') as graph:
    query = 'insert $a isa animal has name \"squirrel\";'  # Escape your quotes, or use a raw string
    concept_map = graph.execute('match $a isa animal; get;')
    print(concept_map)
        [{'a': {
            'id': 'V4144',
            'type': 'animal',
            'base_type': 'entity',
            'attributes': [{
                'id': 'V4216',
                'label': 'name',
                'value': 'squirrel'
            }]
        }}]
    print(concept_map.object)  # Get the underlying ConceptMap object
    print(concept_map['a'].object)  # Get the underlying Concept object
    graph.commit()  # Don't forget to commit changes if you make them. N.B. this also closes the session

API

primal_grakn.Graph

primal_grakn.ConceptDict

Dictionary respresenation of a Grakn Concept object.

An explanation about explanations

At the time of writing, the explanation data structures Grakn provides are undocumented. Briefly, the Grakn ConceptMap object exposes the set of facts as a tree. The top level of this tree includes the inferred facts from the response, and the compositional facts are nested within deeper levels.

At present, we provide two structures to access these facts: one is .explanation, which is the explanation tree as it is exposed by grakn-python, parsed into the form of a python dictionary. The second is .flat_explanation, where the tree is flattened into a list. I found this much more convenient for my purposes, as it meant I could filter the list for only the types of concepts I was interested in for my explanation, and then sort the list into the logical order (Grakn does not provide any ordering in its explanation output).

If you don't need information about the depth of inferences underlying your response, use .flat_explanation.

I feel this is an area where ripe for improvement both as regards the Grakn API and third party packages such as this.

Installation

Clone the repo

git clone https://github.com/cyclecycle/pygrakn.git

Requirements

• Grakn running.

• Official python-grakn client:

  pip install grakn

Contributions

Are welcome :)