Towards Modin 1.0

This page contains the details of the criteria by which we will decide and declare that Modin is stable enough for a 1.0 release.

Feature requirements

• Backend agnostic Execution API
  • Supported on the backends: Ray, Dask, Python Multiprocessing
• Data-manipulation front-end API for DSLs
  • Supported on the DSLs: pandas, SQL
  • (mostly) full pandas API support
• Dataframe algebra translation layer
• Query execution both in eager and lazy modes
• Data partitioning and placement API
  • Partition protocol, high level API for partitioning and placement
• Fault tolerance and error handling
  • Logging
  • Fault tolerance for a node or worker failures

Performance

• Reasonable performance on common operators
• Not slower than pandas on medium dataset sizes (500MB)

Non-functional requirements

• Microbenchmarks, end-to-end benchmarks, and dashboard
• Full documentation both for programmers and users
• Code complying to the adopted style-guide and docstring format
• Dataframe exchange protocol