datacheckr is an R package to check data frame's rows, column names, column classes, values, unique keys and joins.

Why Another Data Checking Package?

There are several existing R packages for checking data frames including assertr, assertive and datacheck. They are great for checking data in scripts but they have several limitations when embedded in functions in packages.

Informative Error Messages

library(assertr)
assert(mtcars, within_bounds(0,1), mpg)
#> Error: 
#> Vector 'mpg' violates assertion 'within_bounds' 32 times (e.g. [21] at index 1)

The error message is not that helpful for a user who is not familiar with the internals of a function that has just thrown that error.

The same test using the datacheckr::check_data() function produces an error message which is more likely to allow the end user to diagnose the problem.

library(datacheckr)
check_data(mtcars, list(mpg = c(0,1)))
#> Error: the values in column mpg in mtcars must lie between 0 and 1

Intuitive Checks

data1 <- data.frame(
  Count = c(0L, 3L, 3L, 0L), 
  LocationX = c(2000, NA, 2001, NA), 
  Extra = TRUE)

The following datacheckr code states that data1 should have a column Count of non-missing integers with values of 0, 1 or 3, should not have a column Comments and can include a column LocationX with missing values between 1012 and 2345.

check_data(data1, list(
  Count = c(0L, 1L, 3L), 
  Comments = NULL, 
  LocationX = c(NA, 2345, 1012),
  LocationX = NULL))

To produce similar functionality with assertr would require something like (please file an issue if the code below can be improved)

library(magrittr) # for the piping operator
data1 %>% assert(in_set(0, 1, 3), Count) %>%
  assert_rows(num_row_NAs, within_bounds(0,0.1), Count)
stopifnot(!"Comments" %in% colnames(data1))
if ("LocationX" %in% colnames(data1))
  data1 %>% assert(within_bounds(1012, 2345), LocationX)

A Single Function Call

The above checks can be performed on several data frames by simply repeatedly calling check_data()

data3 <- data2 <- data1

values <- list(
  Count = c(0L, 1L, 3L), 
  Comments = NULL, 
  LocationX = c(NA, 2345, 1012),
  LocationX = NULL)

check_data(data1, values)
check_data(data2, values)
check_data(data3, values)

The same tests using assertr would require the assertr code above to be copied and pasted three times which is tedious to produce and read; and as a result error prone.

Installation

install.packages("datacheckr")

# install.packages("devtools")
devtools::install_github("poissonconsulting/datacheckr")

More Information

For more information view vignette("datacheckr") for An Introduction to checkdatar.