Perhaps you can have better support for composite data flow variables. E.g. modelling a moving window multivariate time-series, where you have a series of vectors and as new vectors are added calculations are triggered only for values which need re calculation. Also if a calculation is dependent on a slice updates are propagated in the most optimal way that only the needed data is recalculated.

The composite values can be:

• Same type:
  • Vectors
  • Matrices
  • Arrays
  • Series
  • Data frames
  • Slices
• Different types:
  • Tuples (also series with different types)
  • Tables (also Data Frames with different types)
  • Views (slices with different types)