Are different database systems going to be supported data sources?

[Smurphy000]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
Are there plans to add a way to support connect/reading from a database (Postgres, MySQL, etc)?

Describe the solution you'd like
I am currently thinking in terms of Spark, where I have the ability to specify a format where that format could also be a package like com.microsoft.sqlserver.jdbc.spark which connects to the database via jdbc

If this is already supported in some way, where can I find this?

[Igosuki]

Hi, disclaimer not a maintainer. No it's not currently supported. I definitely see a use for it to join database tables with columnar data, although I find it is subject to the multiple-writer database anti-pattern, so you must have strong test suites to make sure changing the db (which is often owned by an API) won't affect the ETL process.
One would have to create an experimental crate, implement a generic TableProvider for RDBMs systems, validate the SQL statements against the db, and adapt arrow to whatever data structure is used for sql statements.
Interesting projects that come to mind for this are tokio-rs/rdbc and sqlx, there is also the arrow jdbc adapter and ongoing tasks like https://issues.apache.org/jira/browse/ARROW-7744

[jorgecarleitao]

fyi, https://github.com/sfu-db/connector-x already supports arrow, so it is "only" a matter of gluing it together.

[Igosuki]

@jorgecarleitao didn't know about this, so just need to implement a table provider and add the sources !

[houqp]

Yes, I would recommend implementing a table provider plugin using connector-x as a self-contained crate :)

[Jefffrey]

DataFusion has support for custom table providers, see the guide: https://datafusion.apache.org/library-user-guide/custom-table-providers.html

Can also refer to this repo for some implementations: https://github.com/datafusion-contrib/datafusion-table-providers