dplyr pull() in R: Extract a Column as a Vector

The pull() function in dplyr extracts a single column from a data frame and returns it as a plain vector. It is the pipe-friendly equivalent of base R's $ and [[ operators.

pull(df, mpg)                          # extract mpg as vector
pull(df, "mpg")                        # quoted name also works
pull(df, 1)                            # by position (column 1)
pull(df, last_col())                   # last column
pull(df, mpg, name = car_name)         # named vector
df |> filter(cyl == 4) |> pull(mpg)    # in pipeline
mean(pull(df, mpg))                    # feed to a function

What pull() does in one sentence

pull() extracts ONE column from a data frame and returns a plain VECTOR. Unlike select(df, x) which returns a one-column data frame, pull(df, x) strips the data frame wrapper and gives you the column's values directly.

This matters when downstream code expects a vector (e.g., mean(), length(), sum()). A one-column data frame is NOT the same as a vector; passing one where the other is expected often errors or produces surprising output.

Syntax

pull() takes a data frame and a single column reference. The reference can be a bare name, a quoted string, or an integer position. An optional name argument turns the output into a named vector.

The full signature:

pull(.data, var = -1, name = NULL, ...)

.data is the data frame. var is the column to extract; default is -1 (the LAST column). name is an optional column whose values become names of the output vector.

Six common patterns

1. Extract a column by name

The result is a numeric vector, NOT a data frame.

2. Extract by position

Integer positions are 1-indexed. pull(df, last_col()) extracts the last column by position helper.

3. Inside a pipeline

pull() is the bridge from "data frame world" to "vector world". After pull, you can use base R functions that expect vectors: mean(), sum(), range(), etc.

4. Named vector with the name argument

Setting name = car turns the output into a named vector. Useful for lookup tables and direct subsetting like named_mpg["Mazda RX4"].

5. Pull last column (default)

pull() with no argument returns the LAST column. Convenient after a mutate() where the new column is at the end.

6. Pull with quoted column name

Quoted strings work the same as bare names. Useful inside functions where the column name is stored in a variable.

pull() vs base R column extraction

Three ways to extract a column as a vector. Choose based on context.

When to use which:

• Use pull() inside any dplyr pipeline.
• Use df$x or df[["x"]] for one-line scripts without other dplyr code.
• Use pull(df, "x") inside functions where the column name is a variable.

Common pitfalls

Pitfall 1: confusing pull() and select(). pull(df, x) returns a VECTOR. select(df, x) returns a one-column DATA FRAME. They behave differently when fed to other functions: mean(select(df, mpg)) errors, but mean(pull(df, mpg)) works.

Pitfall 2: pull() returns ONLY the last column by default. Calling pull(df) without specifying a column returns the LAST column. New users sometimes expect it to return all columns. To get a specific column, always specify: pull(df, mpg).

Pitfall 3: pulling from a grouped frame can return UNEXPECTED ORDER. group_by(df, g) |> pull(x) may not return values in the original row order, depending on dplyr version. If order matters, arrange() or ungroup() first.

Try it yourself

Related dplyr functions

After mastering pull(), look at:

• select() (with one column): returns a data frame, not a vector
• $ and [[ in base R: equivalent to pull but not pipe-friendly
• tibble::deframe(): convert a 2-column tibble to a named vector
• tibble::enframe(): inverse, convert a named vector to a tibble

For extracting multiple columns, use select() and keep the data frame structure. pull() is specifically for the single-column case.

FAQ

What is the difference between pull and $ in dplyr and base R?

pull(df, x) and df$x both return the same vector. The difference is composability: pull() works inside dplyr pipes (df |> filter(...) |> pull(x)), while $ requires the data frame on the left side and breaks the pipe flow.

How do I extract a single column as a vector in dplyr?

Use pull(df, column_name). The result is a vector you can pass to mean(), sum(), length(), etc. Use select(df, column_name) instead if you want a one-column data frame.

Can I use pull with a quoted string for the column name?

Yes. pull(df, "mpg") and pull(df, mpg) produce the same result. Use quoted strings inside functions where the column name is stored in a variable.

What does pull return for a grouped data frame?

A plain vector with no grouping. The vector contains values for all rows, but the group structure is gone. If grouping matters for the downstream operation, summarise within the group first, or ungroup() before pulling.

How do I get a named vector from pull?

Pass the name argument: pull(df, value_col, name = name_col). The result is a named vector where each name comes from name_col and each value from value_col. Useful for lookup tables.