[joocer]

Adds Opteryx, which serves query results over Arrow Flight SQL from its hosted service. The Flight SQL server is built on pyarrow.flight (mabel-dev/flight.opteryx.app). I'm the maintainer.

[github-actions]

Preview URL: https://joocer.github.io/arrow-site

If the preview URL doesn't work, you may have forgotten to configure your fork repository for preview.
See https://github.com/apache/arrow-site/blob/main/README.md#forks for instructions on how to configure.

[kou]

I have some questions about the description:

1. The PR description says "uses PyArrow in its open-source engine (mabel-dev/opteryx)", but it seems that the engine behind opteryx.app is now opteryx-core, which says "neither PyArrow nor NumPy is present anywhere in the engine" in its README. It seems that opteryx-core exports results via the Arrow C Data Interface (draken/interop/draken_to_arrow.h) instead. Is my understanding correct?

2. Is the Arrow Flight SQL endpoint implemented with pyarrow.flight? I couldn't find the server implementation in public repositories. (mabel-dev/opteryx.app says it contains no service code.)

[joocer]

Thanks for checking.

1. Yes, that’s right. The engine behind opteryx.app is opteryx-core, which doesn’t use PyArrow; it supports exporting results through the Arrow C Data Interface, as you describe.
2. Yes. The Flight SQL server is built on pyarrow.flight, and it’s now public: https://github.com/mabel-dev/flight.opteryx.app/blob/main/app/flight/server.py (OpteryxFlightServer subclasses FlightServerBase). It runs at grpc+tls://flight.opteryx.app:443.

The entry only focuses on the flight service

[Opteryx][69]: Hosted SQL engine for querying files where they live.
Its Arrow Flight SQL endpoint, built on pyarrow.flight, returns query
results as Arrow record batches, so ADBC, JDBC and Python Flight SQL clients
can connect and read them directly.

[kou]

+1