Pipes and the tidyverse mental model

R
How to think in tidyverse pipelines: understanding pipes, data flow, and readable transformations in R.
Author

Pablo Benavides-Herrera

Modified

February 13, 2026

This document is optional, but strongly recommended.

If you are new to R or come from another language, understanding pipes will make the entire course easier, clearer, and more enjoyable.

1 What a pipe really means

In tidyverse-style R, the pipe operator |> means:

“take the result of the previous step and pass it as the input of the next one”

A useful way to read a pipeline is to mentally replace |> with the word “then”.

For example:

df |>
  filter(value > 0) |>
  summarise(mean_value = mean(value))

can be read as:

“take df, then filter rows where value > 0, then compute the mean of value.”

Tip

If you can read a pipeline out loud using “then” and it still makes sense,
you are probably writing clear tidyverse code.

For more on style, including guidance on pipes, see
R for Data Science (2e) — Workflow: code style (Pipes).

2 Tidyverse code with and without pipes

df_filtered <- filter(df, value > 0)
df_transformed <- mutate(df_filtered, log_value = log(value))
df_grouped <- group_by(df_transformed, id)
df_summary <- summarise(df_grouped, mean_value = mean(log_value))
  • Explicit intermediate objects
  • More typing
  • Easier to lose the “story” of the data
df |>
  filter(value > 0) |>
  mutate(log_value = log(value)) |>
  group_by(id) |>
  summarise(mean_value = mean(log_value))
  • One clear data flow
  • No temporary variables
  • Easier to read, explain, and debug

3 The tidyverse mental model

A tidyverse pipeline usually follows this pattern:

  • start with a dataset
  • filter rows
  • create or transform variables
  • group the data
  • summarise

This is not a strict rule, but a very common and effective structure.

Note

Most tidyverse verbs return a new tibble.
The original data is never modified unless you explicitly reassign it.

If you want a broader overview of how the pieces fit together, see
R for Data Science (2e) — Introduction.

4 Pipes are not magic

The pipe does not change what functions do.
It only changes how inputs are passed.

4.1 Without pipes

summarise(
  group_by(
    mutate(
      filter(df, value > 0),
      log_value = log(value)
    ),
    id
  ),
  mean_value = mean(log_value)
)

4.2 With pipes

df |>
  filter(value > 0) |>
  mutate(log_value = log(value)) |>
  group_by(id) |>
  summarise(mean_value = mean(log_value))
Tip

If you can rewrite a pipeline as nested function calls,
then the pipeline is doing exactly what you think it is.

5 Pipes and column references

Inside tidyverse verbs, column names are used directly.

df |>
  summarise(mean_value = mean(value, na.rm = TRUE))

No $, no indexing, no extra syntax.

Note

This works because tidyverse uses data-masking:
column names are looked up automatically inside the data.

You do not need to master these details for this course.

If you are curious about what’s happening under the hood, see
R for Data Science (2e) — Functions.

6 Why we avoid $ in this course

  • It breaks the pipeline mental model
  • It mixes different styles of R
  • It becomes confusing with grouped data

For clarity and consistency, we will stick to tidyverse verbs and pipelines.

7 Common pipe mistakes

7.1 Breaking the pipeline

df |>
  filter(value > 0)

mutate(log_value = log(value))

The second line does not receive the result of the pipeline.

Correct:

df |>
  filter(value > 0) |>
  mutate(log_value = log(value))

7.2 Forgetting reassignment

df |>
  mutate(x = x * 2)

Nothing changes.

Correct:

df <- df |>
  mutate(x = x * 2)
Warning

If you forget to reassign, the pipeline runs but the result is discarded.

8 Pipes and readability

Pipelines are about clarity, not cleverness.

Prefer this:

df |>
  filter(value > 0) |>
  group_by(id) |>
  summarise(mean_value = mean(value))

Over this:

df |> filter(value > 0) |> group_by(id) |> summarise(mean_value = mean(value))
Tip

One verb per line is a good default.

For more examples of filter(), mutate(), group_by(), and summarise(), see
R for Data Science (2e) — Data transformation.

9 When not to use pipes

Pipes work best for linear transformations.

They are less useful for:

  • complex branching logic
  • deeply nested conditionals
  • non–data-centric code

Even so, in this course we will often prefer pipelines for consistency.

df |>
  summarise(mean_value = mean(value, na.rm = TRUE))

10 Pipes in this course

Throughout the course:

  • all data manipulation examples will use pipes
  • tidyverse verbs will be preferred
  • clarity will be valued over clever tricks

If something feels confusing, read the pipeline top to bottom, step by step —
using “then” as you go.

11 Final takeaway

Tip

If you understand pipes, you understand most of tidyverse-based R.

Everything else in the course builds on this idea.

Back to top