WARNING: This package is in early stages of development and is likely to change.

This package is modeled after the recipes framework consisting of a recipe and a set of steps to apply. The goals of the package are to improve speed, decrease memory consumption, increase consistency between steps, and decrease some boiler plate code for step additions. The first three goals are likely to be achieved but the fourth might not given that the package implements two APIs (one for R6 and one for S3). Speed and memory improvements come from the collapse package, Rcpp code, using algorithms that scale better, and attempts to reduce copying data.

It diverges in a few ways:

• based on R6

• focus on long datasets (millions of rows)

• focus on memory usage

• focus on speed

• steps tailored to groundwater applications

• decrease the number of dependencies and foreign functions

• more flexible output options (list, matrix, data.frame, data.table, tibble)

• statistically less robust

• mixes modelling and feature engineering

• API changes

  • uses terms instead of … for variable selection and selections are wrapped in c() when more than one is required.
  • R6 and standard R interfaces

• Documentation
• Add alt text for figures in vignettes
• Assess default bake step for steps
• improve use of …
• Explore convolution methods for Laplace solutions
• Optimize Laplace solutions
• Fix selectors
• Feature naming
• Increase test coverage
• Steps
  • step_temporary_deployment
  • step_read_transducer
• Tests for selectors
• Increase speed
• Decrease memory consumption