When creating a reusable R-script, I try to make it as flexible as possible to my future self. This means not hard-coding anything in the code and taking as many user inputs as possible.
However sometimes, it is not possible to capture the flexibility I desire without being overwhelmed by too many command-line options. In such a case, I would prefer to pass an expression that can be directly evaluated by the code (not safe!). For instance, suppose you want to pass arbitrary expressions to filter a data frame.
Consider the following code:
iris %>% filter(Petal.Width < 4)
This would return a tibble with rows where Petal.Width is less than 4 from the iris
dataset. But what if you want the user (and yourself) to specify how the rows are filtered
arbitrarily. The following code does not work:
filter_exp <- "Sepal.Width < 3"
iris %>% filter(filter_exp)
Neither does unquoting the expression using !! works:
filter_exp <- "Sepal.Width < 3"
iris %>% filter(!!filter_exp)
The approach here is first to parse the string into an R expression and then use the !!
operator to format it to tidyr specification. To make this easy, there's a R-package
called friendlyeval that lets you do it quite easily.
For example,
iris %>%
filter(!!friendlyeval::treat_string_as_expr("Petal.Width < 4")
friendlyeval has a few other evaluation options as well, and I encourage you to explore
that.
Happy hacking!