extended relational API for R bindings (similar to duckdb-r)

[e-kotov]

Hi @paleolimbot ,

I'm exploring the possibility of building a dplyr-compatible interface on top of sedonadb for R, similar to what duckplyr provides for DuckDB.

Looking at the current R bindings, I noticed that sedonadb exposes a limited subset of DataFrame operations:

• select_indices() for column selection
• limit() for row limiting
• collect() / to_view() for materialization

In contrast, duckdb-r exposes a full relational algebra API that duckplyr uses:

• Expression builders: expr_reference(), expr_constant(), expr_function(), expr_comparison()
• Relation operations: rel_filter(), rel_project(), rel_aggregate(), rel_order(), rel_join()

This allows duckplyr to translate dplyr verbs directly into relational operations without going through SQL string generation.

Questions:

1. Are there any plans to expose more of DataFusion's DataFrame API (like filter(), aggregate(), sort()) through the R bindings?

2. Would there be interest in accepting contributions that add an expression/relational API similar to duckdb-r?

For now, I'm working on a SQL-based approach using sd_sql(), which works but requires R-to-SQL expression translation. A native relational API would be more elegant and potentially more performant (avoiding SQL parsing overhead).

[paleolimbot]

Yes and yes! In addition to being more elegant than SQL generation, there are some things that we need to do that can't quite be expressed in SQL yet (DataFusion's SQL support for extension types is still in progress).

My vacation programming this year was to try to get a framework for expression translation solid and hopefully I can have that ready for review in an evening or two ( #468 ). I would love any/all contributions! If you have a few steps in mind based on duckdb-r that would be helpful (particularly if there are low-level pieces that are missing that you'd like one of us to do).