v0.0.0.9004
Pre-release
Key features added in this release
Python functions in analysis pipelines
Python functions can now be defined, and Python files sourced through the reticulate package. These functions which have a reference in the R environment can be registered to and AnalysisPipeline object. This means that an interoperable pipeline can be created comprising of R, Spark & Python functions.
A vignette on how to use Python functions has been added. The Interoperable pipelines vignette has also been updated to showcase pipelines with all 3 engines.
Improved formula parsing & execution
Some logical inconsistencies in formula semantics have been resolved. Now, for data functions, the data argument can be either one of 3 things:
- Not passed, meaning that it is a pronoun which should operate on the input that the pipeline object has been instantiated with
- A data frame explicitly passed
- A formula - which denotes an output of a previous function
Additionally, if the input type of the data argument of data function is one of R data frame, Spark DataFrame, or Pandas DataFrame; type conversion is automatically performed according to the engine of that data function.