R Expression Add-on

  collapse = TRUE,
  comment = "#>"
gl <- glue::glue
`%>%` <- magrittr::`%>%`

1. %?<-%: Assign if invalid

When coding in R, the data checking is actually a headache. For example, to check if a variable aa exists and not NULL, otherwise set a default value to be 1, the check looks like:

if( exists('aa') && !is.null(aa) ){
  aa <- 1

Most of time we are repeating ourselves. With %?<-% operator, we just need:

aa %?<-% 1

The powerful part is the left-hand side can be any expression. For example,

l %?<-% list()

l$aa %?<-% 1


If the value exists, then %?<-% does nothing (not even evaluate the expressions on the right-hand side)

# e already exists
e <- list(aa = 1)

# %?<-% will not evaluate rhs, nor assign values
  e %?<-% { Sys.sleep(10); list(aa = 2) }

2. JavaScript-style of creating functions

In modern JavaScript, function can be created via (arg) => { ... }. For example,

const li = ['A', 'T', 'G', 'C'];
li.map((el, ii) => {
  return(`The index for ${el} is ${ii}`);

dipsaus provides functions iapply, and %=>%, together with glue package, we can apply elements like this:

# gl <- glue::glue
# `%>%` <- magrittr::`%>%`

li <- c('A', 'T', 'G', 'C')
li %>% iapply(c(el, ii) %=>% {
  gl('The index for {el} is {ii}')

%=>% collect the left-hand side elements as arguments and right-hand side expression as body and create function:

c(a, b=a^2, ...) %=>% {
  print(c(a , b,...))

3. Match nested calls match_calls

The function match.call provided by base package let us format calls with formals matched.

match.call(textInput, call = quote(textInput('inputId', 'label', 'aaa')))

This is already powerful as we can parse the expressions using as.list() to get the input parameters. However, when encounter the nested calls like shiny UI components, match.call does not work well. We can't see the matched results inside of the nested functions.

match.call(tagList, call = quote(tagList(
      tags$li(textInput('inputId', 'label', 'aaa'))

match_calls solves this problem by recursively calling match.call:

match_calls(call = tagList(
      tags$li(textInput('inputId', 'label', 'aaa'))
), recursive = TRUE)

It can also change modify the calls. For example, we want to add ns to input ID in shiny modules, then the following replave_args changes "inputId" to ns("inputId")

match_calls(call = tagList(
      tags$li(textInput('inputId', 'label', 'aaa'))
), recursive = TRUE, replace_args = list(
  'inputId' = function(v, ...){
    as.call(list(quote(ns), v))

4. Pipe-friendly functions

Pipe functions can simplify the workflow and make R code more readable. The most popular pipe %>% allows the left-hand elements to be the first input of the right-hand side functions. dipsaus provides several pipe-friendly functions.

No-operations no_op

no_op takes whatever input in, and returns the input, with side effects. For example, we want to plot the results from the pipe and continue the analysis, usually this is what happens:

x %>% 
  do_something(...) ->


x_tmp %>% 
  do_others(...) ->

With no_op, the pipe becomes:

x %>% 
  do_something(...) %>% 
  no_op(plot, ylim = c(0,100)) %>% 
  do_others(...) ->

Here's an example

par(mfrow = c(1,2))

(1:10) %>% 
  iapply(c(el, ii) %=>% {
    rnorm(20, el, ii)
  }, simplify = FALSE) %>% 
  unlist %>% 

  # Begin no-ops, result will not change
    # Use expression and "." to refer the data
  }) %>% 
    # Use function and pass ... to function
    plot, x = seq(0,1,length.out = 200), 
    type = 'p', ylim = c(-20,20), pch = 16,
    xlab = 'Time', ylab = 'Value', las = 1
  ) %>% 
  no_op(hist, xlab = 'Values', main = 'Histogram') ->



This is a wrapper of aggregate function. When using formula, aggregate requires the formula to be the first element. If the pipe results are data.frame and we want to use formula, it's super inconvenient.

## S3 method for class 'formula'
aggregate(formula, data, FUN, ..., subset, na.action = na.omit)

do_aggregate allows the first element to be data frames while using formula:

ToothGrowth %>%
  do_aggregate(len ~ ., mean)

Try the dipsaus package in your browser

Any scripts or data that you put into this service are public.

dipsaus documentation built on July 9, 2023, 5:43 p.m.