The libr package brings the concepts of data libraries, data dictionaries, and data steps to R.

These concepts have been available in SAS® software for decades. But they have not been available in R … until now!

The libr package also includes an enhanced equality operator to make data comparisons more intuitive.

Key Functions

The above concepts are implemented in the libr package with four key functions. They are:

How to Use

Let’s look at some simple examples of each of the four functions above. These examples will be using some sample data. The sample data is included in the libr package, and also available for download here.

The libname() Function

The libr libname() function is quite similar to the SAS® libname statement. The first parameter is the name of the library. The second parameter is a path to a directory the library will point to. The third parameter is the engine with which to read and write the data.


# Get path to sample data
pkg <- system.file("extdata", package = "libr")

# Define data library
libname(sdtm, pkg, "csv") 

The libname() function above will send two types of information to the console:

The summary print-out looks like this:

# library 'sdtm': 8 items
- attributes: csv not loaded
- path: C:/packages/libr/inst/extdata
- items:
  Name Extension Rows Cols     Size        LastModified
1   AE       csv  150   27  88.1 Kb 2020-09-18 14:30:23
2   DA       csv 3587   18 527.8 Kb 2020-09-18 14:30:23
3   DM       csv   87   24  45.1 Kb 2020-09-18 14:30:23
4   DS       csv  174    9  33.7 Kb 2020-09-18 14:30:23
5   EX       csv   84   11    26 Kb 2020-09-18 14:30:23
6   IE       csv    2   14    13 Kb 2020-09-18 14:30:23
7   SV       csv  685   10  69.9 Kb 2020-09-18 14:30:24
8   VS       csv 3358   17   467 Kb 2020-09-18 14:30:24

The summary displays what type of library it is, where it is located, and what data (if any) is already in the library directory. In this case, there are eight ‘csv’ files available.

For each of the eight files, the libname() function also displayed the column specifications used to import the data file. A column specification looks like this:


-- Column specification ------------------------------------------
  STUDYID = col_character(),
  DOMAIN = col_character(),
  USUBJID = col_character(),
  VSSEQ = col_double(),
  VSTESTCD = col_character(),
  VSTEST = col_character(),
  VSPOS = col_character(),
  VSORRES = col_double(),
  VSORRESU = col_character(),
  VSSTRESC = col_double(),
  VSSTRESN = col_double(),
  VSSTRESU = col_character(),
  VSBLFL = col_character(),
  VISITNUM = col_double(),
  VISIT = col_character(),
  VSDTC = col_date(format = ""),
  VSDY = col_double()

The column specification shows how the data was imported. Since ‘csv’ files do not contain well-defined data type information on each of the columns, the libname function has to guess at the data types. The column specification shows you what the guesses were. This is useful information. You should review these column specifications to see if the libname function guessed correctly. If it did not guess correctly, you can control the import data types by sending a specs() collection of import_spec() objects to the import_specs parameter on the libname() function. See the specs() documentation for an example and additional details.

The lib_load() Function

Observe that there is difference between the SAS® libname statement and the libr libname() function. The difference is that after the SAS® libname statement is called, the data is immediately available to your code using two-level (<library>.<dataset>) syntax.

With the libr function, on the other hand, the data is immediately available using list syntax on the library variable name. To get the two-level syntax, you first have to call the lib_load() function.

# # library 'sdtm': 8 items
# - attributes: csv loaded
# - path: C:/packages/libr/inst/extdata
# - items:
#   Name Extension Rows Cols     Size        LastModified
# 1   AE       csv  150   27  88.1 Kb 2020-09-18 14:30:23
# 2   DA       csv 3587   18 527.8 Kb 2020-09-18 14:30:23
# 3   DM       csv   87   24  45.1 Kb 2020-09-18 14:30:23
# 4   DS       csv  174    9  33.7 Kb 2020-09-18 14:30:23
# 5   EX       csv   84   11    26 Kb 2020-09-18 14:30:23
# 6   IE       csv    2   14    13 Kb 2020-09-18 14:30:23
# 7   SV       csv  685   10  69.9 Kb 2020-09-18 14:30:24
# 8   VS       csv 3358   17   467 Kb 2020-09-18 14:30:24

Notice on the console printout that the library is now “loaded”. That means the data has been loaded into the workspace, and is available using two-level syntax. If you are working in RStudio, the environment pane will now show all the datasets available in the library.

At this point, you can work with your data very much the same way as you would in SAS®. You can pass these datasets into statistical functions, or manipulate them with dplyr functions. Note that you can also work with individual variables on the datasets using dollar sign (“$”) syntax.

# Get total number of records
# [1] 87

# Get frequency counts for each arm
# ARM A          ARM B          ARM C          ARM D SCREEN FAILURE 
# 20             21             21             23              2 

The datasets will be available in the workspace for the length of your session. If you wish to unload them from the workspace, call the lib_unload() function. See the lib_load() and lib_unload() documentation for additional information on these functions.

To see more examples of the libr data management functions, refer to the articles on Basic Library Operations and Library Management.

The dictionary() Function

Once you have a library defined, you may want to examine the column attributes for the datasets in that library. Examining those column attributes can be accomplished with the dictionary() function. The dictionary() function returns a tibble of information about the data in the library.

Continuing from the example above, let’s look at the dictionary for the ‘sdtm’ library created previously.

# # A tibble: 130 x 10
#    Name  Column  Class     Label Description Format Width Justify  Rows   NAs
#    <chr> <chr>   <chr>     <chr> <chr>       <lgl>  <int> <chr>   <int> <int>
#  1 AE    STUDYID character NA    NA          NA         3 NA        150     0
#  2 AE    DOMAIN  character NA    NA          NA         2 NA        150     0
#  3 AE    USUBJID character NA    NA          NA        10 NA        150     0
#  4 AE    AESEQ   numeric   NA    NA          NA        NA NA        150     0
#  5 AE    AETERM  character NA    NA          NA        72 NA        150     0
#  6 AE    AELLT   logical   NA    NA          NA        NA NA        150   150
#  7 AE    AELLTCD logical   NA    NA          NA        NA NA        150   150
#  8 AE    AEDECOD character NA    NA          NA        43 NA        150     0
#  9 AE    AEPTCD  numeric   NA    NA          NA        NA NA        150     0
# 10 AE    AEHLT   character NA    NA          NA        63 NA        150     0
# # ... with 120 more rows

The resulting dictionary table shows the name of the dataset, the column name, and some interesting attributes related to each column. As you can see, the libr dictionary table is overall quite similar to a SAS® dictionary table. See the dictionary() function documentation for more information.

The datastep() Function

People with experience in SAS® software know that it is sometimes advantageous to process row-by-row. In SAS®, row-by-row processing done with a data step. The data step is one of the most fundamental operations when working in SAS®.

The libr package offers a datastep() function that simulates this style of row-by-row processing. The function includes several of the most basic parameters available to the SAS® datastep: keep, drop, rename, retain, and by. Here is a simple example, again using the data from the library already defined above:

age_groups <- datastep(sdtm.DM, 
                       keep = c("USUBJID", "AGE", "AGEG"), { 
                         if (AGE >= 18 & AGE <= 29)
                           AGEG <- "18 to 29"
                         else if (AGE >= 30 & AGE <= 44)
                           AGEG <- "30 to 44"
                         else if (AGE >= 45 & AGE <= 59)
                           AGEG <- "45 to 59"
                           AGEG <- "60+"
# # A tibble: 87 x 3
#    USUBJID      AGE AGEG    
#    <chr>      <dbl> <chr>   
#  1 ABC-01-049    39 30 to 44
#  2 ABC-01-050    47 45 to 59
#  3 ABC-01-051    34 30 to 44
#  4 ABC-01-052    45 45 to 59
#  5 ABC-01-053    26 18 to 29
#  6 ABC-01-054    44 30 to 44
#  7 ABC-01-055    47 45 to 59
#  8 ABC-01-056    31 30 to 44
#  9 ABC-01-113    74 60+     
# 10 ABC-01-114    72 60+     
# # ... with 77 more rows

Notice that the datastep() function kept only those variables specified on the keep parameter. The data step itself is passed within the curly braces. You can put any number of conditional statements and assignments inside the curly braces, just like a SAS® data step. Also like a SAS® data step, you do not need to ‘declare’ new variables. Any name not identified as an R function name is assumed to be a new variable, and will be created automatically on the input data.

The datastep function also supports “first.” and “last.” functionality through use of the by parameter. See additional examples on the datastep() help page and in the data step article.

The %eq% Function

In Base R, there are two comparison functions: The double-equal (==) infix operator, and the identical() function. The weakness of the infix operator is that it will not test for NULL or NA values. The weakness of the identical() function is it will compare everything, including attributes assigned to the objects you are comparing. If there is even a small difference in one of these attributes, the identical() function will return FALSE. This behavior is inconvenient, especially when working with dplyr functions, as these functions assign many attributes to the datasets they are manipulating.

The %eq% function contained in the libr packages attempts to provide a more intuitive comparison. The %eq% is an infix operator that will test for NA and NULL values, and will ignore attributes in any data comparison. These qualities make it more suitable for the types of comparisons normally encountered when working with data. The suitability of the %eq% operator can be illustrated by examining these methods side-by-side on the same input data.

# Set up input data
v1 <- c(1, 2, 3)
v2 <- c(1, 2, 3)
attr(v2, "label") <- "My Label"
v3 <- c(1, 2, NA)

# Comparing "equal" vectors
v1 == v2              # TRUE TRUE TRUE
identical(v1, v2)     # FALSE
v1 %eq% v2            # TRUE

# Comparing "unequal" vectors
v1 == v3              # TRUE TRUE NA
identical(v1, v3)     # FALSE
v1 %eq% v3            # FALSE

Note the different results returned by the three different comparison functions. The %eq% operator is similar to the identical() function in that it will always return a single TRUE or FALSE value. The difference is that it ignored the “label” attribute assigned to vector v2, and only looked at the data values. Since it saw all the data values were equal, the %eq% operator returned TRUE, while the identical() function returned FALSE.

You can see some more examples of the %eq% operator in the help documentation, or in the Enhanced Equality article.

Next Steps

For next steps, please review the examples provided in the vignette articles. Those articles include: