peerdb
peerdb copied to clipboard
PeerDB-io
Add opensearch connector
@heavycrystal yes! Thanks for the review.
You're absolutely right that this shares a lot of ground with the existing Elasticsearch connector. I even brought this up in the Community Slack before implementing it, as my initial thought was also to shim the client based on the backend selecting OpenSearch or Elasticsearch dynamically.
At the moment, I don't have strong opinions either way. That said, my reasoning for keeping them separate for now is that it helps clarify and isolate the nuances between the two. Especially as their APIs and behaviors are already diverging in subtle ways. For example, the client libraries differ, and certain API endpoints and auth mechanisms are evolving independently.
That said, OpenSearch still claims compatibility with Elasticsearch 7.10 clients, so it might indeed make sense to unify them into a single connector for now. Using a client shim and perhaps probing the server version or exposing a configuration flag to let the user choose.
Happy to revisit the structure based on your thoughts.
@heavycrystal yes! Thanks for the review.
You're absolutely right that this shares a lot of ground with the existing Elasticsearch connector. I even brought this up in the Community Slack before implementing it, as my initial thought was also to shim the client based on the backend selecting OpenSearch or Elasticsearch dynamically.
At the moment, I don't have strong opinions either way. That said, my reasoning for keeping them separate for now is that it helps clarify and isolate the nuances between the two. Especially as their APIs and behaviors are already diverging in subtle ways. For example, the client libraries differ, and certain API endpoints and auth mechanisms are evolving independently.
That said, OpenSearch still claims compatibility with Elasticsearch 7.10 clients, so it might indeed make sense to unify them into a single connector for now. Using a client shim and perhaps probing the server version or exposing a configuration flag to let the user choose.
Happy to revisit the structure based on your thoughts.
I think it would be easier to maintain if it was a single connector with shims and switches for different behaviour. I don't anticipate drastic changes happening, and we can always revisit this if they do. You can take a look at the work we did for MySQL/MariaDB for example, there's quite a few differences but it's still a single connector.
You should be able to reuse the same tests and just test both flavours, let me know if this works
