A manager for translating an ORM into Elasticsearch
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
elasticmanager
sample
.gitignore
LICENSE
MANIFEST
MANIFEST.in
README.md
setup.py

README.md

elasticmanager

A manager for translating an ORM into Elasticsearch

This package is still in active development. Currently it is being built as an mixin for the Django ORM. However, once stable, this will be abstracted away to be able to plugin to other ORMs.

This pacakge requires elasticsearch-dsl which you can get:

pip install elasticsearch-dsl

Getting Started

  1. To get started, install from pip.

    pip install elasticmanager

  2. Add elasticmanager to your settings.py (this enables management commands).

  3. Put the name of your index in settings.py: ELASTICSEARCH_INDEX = 'myindex'

  4. Subclass elasticmanager.ElasticModel.

    from django.db import models
    from elasticmanager.models import ElasticModel
    class Visitor(ElasticModel, models.Model):
       name = models.CharField(max_lendth=100)
       created = models.DateTimeField(auto_now_add=True, editable=False)
  5. Create a doctypes.py in the same app as the corresponding models.py. Add a DocType to this file with the same name as your model. (See Elasticsearch DSL documenation for more information on creating a DocType)

    from django.conf import settings
    from elasticmanager.doctypes import BaseDocType
    from . import models
    class Visitor(BaseDocType):
    
       pk = field.Keyword()
       name = field.Keyword()
       created = field.Date()
    
       class Meta:
           model = models.Visitor
           index = settings.ELASTICSEARCH_INDEX

    Note that the Meta information for the doctype should link to the model and the index.

  6. Run ./manage.py rebuild_mapping

  7. Run ./manage.py rebuild_indexing (if there are items in the DB that need to be inexed)

There is also another management command (./manage.py remove_index <NAME>) that can be used to delete an entire index. Helpful during development stages.

Currently, all calls to the .save() method on an instance of the model will trigger the .save() on the doctype, and therefore will keep the index in Elasticsearch up to date.

Making calls to query and filter

Calls somewhat approximate the default Django syntax.

Get all instances

visitors = Visitor.elastic.all()              # Return everything from Elasticsearch

Get one instance

visitor = Visitor.elastic.get(pk=123)         # Return an instance from Elasticsearch given a specific key
visitor = Visitor.elastic.first()             # Return first instance in a queryset
visitor = Visitor.elastic.last()              # Return last instance in a queryset

Count the number of instances

count = Visitor.elastic.count()               # Count the number of instances in a queryset

Filter/query This lines up with the query and filter methods. See the linked documentation for more information.

johns = Visitor.elastic.filter(name="John")   # Return everything from Elasticsearch

Chaining applies

first_john = Visitor.elastic.filter(name="John").first()

Under the hood, there is an execute() method that is committing the search. This can be called by itself, and probably should be if you are (for example) manipulating a queryset. But, you should not need to call that if you are, for example, calling count() or if you are iterating over the result set.

johns = Visitor.elastic.filter(name="John")
for john in johns:
    print(john)

# or

johns = Visitor.elastic.filter(name="John")
johns.execute()
print(johns.results)

This requirement will be removed with some more fine tuning to work more intuitively.

Plans for the future

  • Abstraction from using models.Manager
  • Fix the class factory so that a Model can be automatically transformed into a DocType without having to define it in doctypes.py.
  • Tests
  • Aggregations
  • Additional Exception handling
  • More management commands
  • More powerful API

If you have any questions, thoughts, complaints, compliments, let me know.