This is a work in progress experiment PoC of rewriting OctoSQL in Rust, as I've been seeing 10-100x performance improvements using a more optimized design.

Development has moved to: https://github.com/cube2222/octosql/tree/rust-rewrite

Roadmap

The current roadmap is to achieve feature parity with the Go version of OctoSQL. When this is done, we'll decide which one gets to stay.

• Projection
• Filter
• GroupBy
  • Support all types
  • More aggregates
  • Wildcard
• Triggers
  • Counting
  • Delay
• Retractions
  • Add retractions to Projection (don't allow the user to remove the retraction column)
• Stream join
  • Float support (currently not tested and possibly wacky)
• Expressions
  • Evaluation in record context.
  • Evaluation in execution context of variables (if we're in a subquery, we need to understand both the current record, and variables stemming from record flows above us)
  • Common functions / arithmetics. (Currently only comparison operators)
• Map (evaluate expressions, this is the only place where expressions are evaluated in OctoSQL, everything else gets evaluated expressions passed from here by name)
  • Wildcard
• Watermarks
  • Metadata Message and Handling
  • Watermark trigger
  • Watermark generators
    • Start with Max difference
• Shuffle
• Subqueries
  • Handle all primitive types
  • Handle multiple columns/rows (Tuple values)
• Lookup join
• Physical Plan (fit for pattern matching)
  • Basic optimiser
  • Pushing down projections
  • Pushing down filters
• Logical Plan
• SQL
  • Temporal Extensions
  • Join
• Nice output printing
  • CSV
  • JSON
  • Live Table
• Data Sources
  • CSV
  • MySQL
  • PostgreSQL
  • JSON
  • Parquet
  • Excel
  • Kafka
• Table Valued Functions
  • Range
  • Tumble
• Durations and Dates support
  • SQL interval

Contributing

Contributions are welcome!

Please open an issue or contact me personally when you want to try to contribute, so that no work gets duplicated or thrown away.