# Handling

The basics themselves only bring so much to error handling in R; they, however, enable a lot more.

# Raiser

One can pass a "raiser," a function that is used instead of the default function when the raise method is called. By default raise uses stop for errors and warning for warnings.

err <- e("print this")

err$raiser <- function(msg){
  stop(crayon::white(msg), call. = FALSE)

#> Error: print this

This function can also be passed to specific raise method for the odd cases.

err <- e("print this")

raise_ <- function(msg){
  stop(crayon::white(msg), call. = FALSE)



Use raise.e and raise.w to define global templates.

Defining templates is ideal when you need a consistent way to raise errors or warnings, e.g.: in shiny where one might want to show a notification instead of calling stop.

notify_ <- function(msg){
  shiny::showNotification(msg, type = "error")


# Class

One can check whether the object returned is an error or a warning with is.e and is.w respectively.

x <- e("This is an error")

[1] TRUE

# Bash

The function bash is analogous to tryCatch but will use err internally. It also allows passing e and w along to easily reuse error messages.

safe_log <- function(x){
 result <- bash(log(x))


Error in safe_log("a"): non-numeric argument to mathematical function

# Globals

You can also set the above as options so all errors will use the defined raiser.

  ERR_RAISER_ERROR = function(msg) cat(msg, "!!!\n"),
  ERR_RAISER_WARNING = function(msg) cat(x, "!\n")

# Resolve

Instead of checking the results of tryCatch with an if statement, one might want to use resolve which will check whether the result is an error or a warning and deal with it accordingly (stop or warning).

err <- e("Log only accepts numeric(s)")

safe_log <- function(x){
 result <- bash(log(x), e = err)


Error: Log only accepts numeric(s)

The resolve function accepts multiple objects, note that these are evaluated in order.

# nothing to resolve
x <- "just a string"

# should resolve
www <- w("Caution")
err <- e("Broken!")

resolve(x, www, err)
Warning: Caution
Error: Broken!

One can, of course, resolve objects that are not errors or warnings: nothing happens.

resolve("nothing wrong here")

# Defer Resolve

You can also use defer_resolve to defer the resolve when the function exits.

safe_log <- function(x){
 result <- bash(log(x), e = e("Gah!"))

 print("Doing something here")


[1] "Doing something here"
#> [1] 2.302585
[1] "Doing something here"
Error: Gah!

# Latch

Errors and warnings can also be latched onto objects so they can be dealt with later, functions such as is.e, resolve, and skip will also work on those objects.

x <- 1
problematic <- latch.e(x, e("Not right"))

[1] TRUE

This allows escalating the error or warning so it can be dealt with later, e.g.: the problematic object created above with latch.e can be passed to another function where it raises an error.

do_sth_with_x <- function(x){
 x + 1

[1] 2
Error: Not right

You can use unlatch to resolve these.

[1] 1

# Skip

The skip function is similar to resolve but instead of calling stop or warning it returns the error or warning object from the parent function. This is useful to escalate the error to the parent function.

# foo always returns an error
foo <- function(){

# bar calls foo
bar <- function(){
  x <- foo()
  print("This should not print")

# baz calls bar
baz <- function(){
  y <- bar()
  print("This should not print either")

Error: Problem!

In the above the print message in bar is never printed, neither is the one in baz: skip returned the error message which was then checked with resolve and the error is actually raised.


You can also set the w argument of the skip function to TRUE to skip warnings, e.g.: skip(x, w = TRUE)

You can call skip as many times as you want to escalate the error back as much as necessary.

# Rules and Checks

When writing code one often themselves writing a lot of conditions to check inputs, outputs, etc.

Together with an error often comes checks that define when it should be raised. Erratum lets one define those rules (with the rule field or addRule method), these rules can then be checked with the check method. Internally it runs the rules and if any of them does not return TRUE raises the error.

A rule is a function that accepts a single argument and must return a boolean.

err <- e("Must be a character of length 3 or more")

err$rule <- is.character
err$rule <- function(x){
  nchar(x) > 2

# passes

# not character

# less than 2 character

This avoids having to write and re-write checks and their error messages.

err <- e("Must be a numeric")
err$rule <- is.numeric

addOne <- function(x){
  x + 1

addTwo <- function(x){
  x + 2

# returns 3

# raises error