In WinVector/wrapr: Wrap R Tools for Debugging and Parametric Programming

wrapr::let() for beginners

wrapr::let() is not interesting if you have no idea why you would need it or how to use it.

wrapr::let() is designed to solve a single problem when programming in R: when in writing code you don't yet know the name of a variable or a data.frame column you are going to use. That is: the name of the column will only be available later when your code is run.

For example suppose we have a data.frame "d" defined as follows:

d <- data.frame(x = c(15, 30, 40))
print( d )

If we know the name of the column we can access it as follows:

print( d$x )

If we don't know the name of the column (such as would be the case when writing a function) we write code like the following:

getColumn <- function(df, columnName) {
  df[[columnName]]
}

print( getColumn(d, 'x') )

This works because R takes a lot of trouble to supply parametric interfaces for most use cases.

We only run into trouble if code we are trying to work with strongly prefers non-parametric (also called non-standard-evaluation or NSE) interfaces.

A popular package that heavily emphasizes NSE interfaces is dplyr. It is the case that dplyr supplies its own methods to work around NSE issues ("underbar methods" and lazyeval for dplyr 0.5.0, and rlang/tidyeval for dplyr 0.7.0). However, we will only discuss wrapr::let() methodology here.

The issue is: it is common to write statements such as the following in dplyr:

suppressPackageStartupMessages(library("dplyr"))

d %>% mutate( v2 = x + 1 )

If we were writing the above in a function it would plausible that we would not know the name of the desired result column or the name of the column to add one to. wrapr::let() lets us write such code easily:

library("wrapr")

addOneToColumn <- function(df, 
                           result_column_name, 
                           input_column_name) {
  let(
    c(RESULT_COL = result_column_name,
      INPUT_COL = input_column_name),

    df %>% mutate( RESULT_COL = INPUT_COL + 1 )

  )
}

d %>% addOneToColumn('v2', 'x')

Again, writing the function addOneToColumn() was the goal. The issue was that in such a function we don't know what column names the user is going to end up supplying. We work around this difficulty with wrapr::let()

All the user needs to keep in mind is: wrapr::let() takes two primary arguments:

• The aliases mapping concrete stand-in names (here shown in ALL_CAPS style) to the variables holding the actual variable names.
• The block of code to re-write and execute (which can be a single statement, or a larger block with braces).

WinVector/wrapr documentation built on Aug. 29, 2023, 4:51 a.m.