Skip to content

Latest commit

 

History

History
131 lines (101 loc) · 4.58 KB

README.md

File metadata and controls

131 lines (101 loc) · 4.58 KB

The ensurer R package

Ensure values at runtime.

Build Status

ensurer is a small utility package for R which provides a simple and light-weight mechanism for ensuring certain aspects of values at runtime.

R does not provide any mechanism for type-safety and since it is not a compiled language, the risk of having unexpected results is there at runtime. R functions often accept different types for the same input and/or have different return types for different situations.

As an example, a query to a database or the scraping of a website might not return valid data, where "validity" can refer to a number of conditions. It might be a positive or certain number of records; that all cases are complete; that some column is weekly increasing; or simply that the result is a data.frame.

If one does not deal with these ambiguities and risks appropriately, some resulting errors may be hard to track down and may propagate in unexpected ways. It is desirable to get an error as soon as a value does not have the correct type or does not satisfy certain criteria.

"Ensuring values" is here meant as a "contract", or a set of conditions, such that if a value does not comply an error is raised instantly (unless special behavior is specified for the failure). An ensuring contract (a function) is created with the ensures_that function (ideal for multiple use or readability with complex contracts).

It is also possible to ensure properties on the fly using ensure_that (ideal for simple, one-time contracts).

Using the magrittr pipe %>% greatly improves semantics of the functionality provided by this package, but it is not necessary.

This package is not meant as a substitute for unit testing, and great packages for this already exist, e.g. testthat by Hadley Wickham. The ensurer package is ideal for scripts or programs where runtime conditions may break the functionality, and where errors should be raised as soon and clear as possible. Although a side-effect, It is my experience that it also promotes better design decisions at outset, and helps catch coding errors early on.

Installation

To install the current development version use devtools:

devtools::install_github("smbache/ensurer")

To install the CRAN version:

install.packages("ensurer")

Basic Examples

The following example shows how to define a contract ensuring that its input is square, and how to use it.

library(magrittr) # for the pipe -> cleaner semantics
library(ensurer) 

# To reference the value being evaluated, use the `.` placeholder.
ensure_square <- ensures_that(NCOL(.) == NROW(.))

# try it out: 
diag(5) %>%
  ensure_square  # passes, so returns the diagonal matrix

# This won't work, and an error is raised.
matrix(1:20, 4, 5) %>% 
  ensure_square

# On the fly contracts:
matrix(1:4, 2, 2) %>%
  ensure_that(is.matrix(.), all(is.numeric(.)))

One can specify several conditions, each separated with a comma. Simple predicate functions can be used in abbreviated symbolic form, e.g as in the example below.

ensure_square <- ensures_that(is.matrix, 
                              NCOL(.) == NROW(.))

Note that all conditions are tested to provide the most feedback upon failure. If "short-circuits" are desired, one can add more (separate) ensuring contracts.

Types and type-safe functions (new, experimental, subject to change)

A "type" is defined as objects satisfying the criteria of a corresponding ensurer contract which has the prefix type_. For example, a foo type will satisfy the conditions of an ensurer contract type_foo. There are some built-in types, but new types are easily defined using ensures/ensures_that and adhering to the type_ naming convention.

Type-safe functions can be made with function_:

f <- function_(a ~ integer, b ~ character: "Hello, World!", {
  rep(b, a)
})

f(10) # fails
f(10L) # works
f(10L, "foo")

type_lm <- ensures_that("lm" %in% class(.) ~ "Value is not a linear model.")
safe_lmsummary <- function_(model ~ lm, {
  summary(model)
})

safe_lmsummary(lm(Sepal.Length ~ ., iris))

The type some allows any value and can be used to allow unsafe inputs. (TODO: make some default type when default values are provided; already default type when no default value is provided.)

Special features include

  • Customizing error behavior
  • Easily combining several contracts
  • Customizing error description
  • Customizing individual conditions' error messages

and more. For more information, see the package vignette.