1. Returning multiple values from functions
R functions can only return a single value, but there are two ways to get around this rule. You can return several objects in a list, or you can store objects as attributes.
2. Getting the session information
Let's write a function to retrieve details about the current R session. It's a simple version of the sessionInfo function. We want the version of R, the name and version of the operating system, and the packages that have been loaded.
3. Defining session()
Wrapping this into a function is straightforward, since we don't have any arguments to worry about. The problem is, how do we return three different things?
4. Defining session()
The answer is to collect all the values to be returned into a list.
5. Calling session()
Calling this function returns what you might expect: the three things you retrieved, in a list.
6. Multi-assignment
Suppose the user actually did want each of the return values separately, instead of in the list. They can do this using the zeallot package's multi-assignment operator. In Python, this is called unpacking variables.
The multi-assignment operator is the usual assignment left arrow, but wrapped in percent signs. On the left, each of the variables to be assigned is wrapped in a call to the c function.
Now there are three variables available containing the individual values.
Note that no changes to the function were needed to make this work. It's up to the user to decide if they want separate variables or not.
7. Attributes
The other technique for returning multiple values involves attributes. Let's see how they work.
Here's a named vector of month numbers. You can see all the attributes for this variable using attributes. In this case, the only attribute is called names.
The attr function can be used to retrieve a specific attribute. Using attr to retrieve the names is equivalent to calling the names function.
attr can also be used to set attributes. You give it the object and the name of the attribute to get, then use the assignment arrow to give it a new value. Here, the names are changed to be the full month names.
8. Attributes of a data frame
Let's look at an example with a data frame. Here's R's Orange dataset, converted into a tibble. It contains the age and circumference of orange trees.
It has three attributes: names for each column; row names, which are just the row numbers; and a class. It's a local tibble, tee-bee-ell-dee-eff, which is a tibble, tee-bee-ell, which in turn is a data frame.
9. Attributes added by group_by()
Now let's see what happens when you group the data frame using dplyr's group_by function.
The names, row names, and class attributes are still there, but group_by has added a groups attribute to keep track of which row is in which group.
10. When to use each technique
dplyr has a philosophy that all its functions should accept a data frame as their first argument, and return a data frame.
That means that group_by couldn't return a list. To comply with the package philosophy, it returned a data frame, then returned the group metadata as an attribute to that data frame.
In general, you should include additional return values as attributes when you need the return value to be a particular type.
If you don't have this constraint, you should return the values in a list.
11. broom
In the next exercise, you'll make use of three functions in the broom package, for tidying model objects into data frames.
glance returns model-level values like the degrees of freedom.
tidy returns coefficient-level values like p-values.
augment returns observation-level values like residuals.
12. Let's practice!
Let's return lots of values!