Interacting with elasticsearch in django

Question

I want to use elasticsearch for a new project and from my research there seems to be 3 viable solutions:

1. Dont use a wrapper and communicate directly with elasticsearch
2. Use elasticsearch-py
3. Use elasticsearch-dsl-py

I like solution 1 because it doesn't need dependencies and I can focus on learning the native syntax/api as opposed that of a wrapper as in 2 or 3. Are there any convincing reasons to use 2 or 3 over 1?

Update

I ended up using elasticsearch-py as it offers various benefits like connection pooling and persistence. I found elasticsearch-dsl-py too abstract and verbose

Answer 1

I would suggest that there is no reason to talk directly to Elasticsearch when an official Python client is available. The Python client does a lot of the heavy lifting for you - otherwise you will spend a lot of time/effort in converting Python data to ES and vice versa.

As regards the choice between elasticsearch-dsl-py and elasticsearch-py:

elasticsearch-dsl-py is a wrapper for the Query DSL only (plus a few other things). It doesn't provide access to the whole Elasticsearch API (e.g., Cluster API, Indices API, Bulk API etc.). It says in the docs:

To use the other Elasticsearch APIs (eg. cluster health) just use the underlying client.

It is highly likely that you will need to use both libraries in any large application. elasticsearch-dsl-py itself uses elasticsearch-py.

I agree with your comment about Haystack - it's Elasticsearch backend leaves a lot to be desired.