Common Code 15 — Strings and Factors

stringr for text cleaning, forcats for plot order

packages

setup

How to arrange and work with factors

Author

Bill Perry

Published

July 5, 2026

Strings and factors

Two related but distinct problems: strings (stringr) deal with the content of text — trimming whitespace, detecting patterns, fixing capitalisation. Factors (forcats) deal with the order of categories — controlling which group appears first on a plot axis or in a table.

⬇️ Download the companion R script: 15_strings_factors.R

library(tidyverse)
library(palmerpenguins)
library(nycflights13)

PART 1 · Strings with `stringr`

1 · Case conversion

The single most common data-cleaning fix: inconsistent capitalisation makes "Sunny", "SUNNY", and "sunny" three different groups.

str_to_lower("SLIMY SCULPIN")      # "slimy sculpin"
str_to_upper("cottus cognatus")    # "COTTUS COGNATUS"
str_to_title("slimy sculpin")      # "Slimy Sculpin"

# Fix all character columns after import
penguins |>
  mutate(across(where(is.character), str_to_lower))

2 · Trimming whitespace

"sunny " and "sunny" are different strings to R. This causes phantom extra factor levels and failed joins.

str_trim("  sunny  ")         # "sunny"  — removes leading and trailing spaces
str_squish("sunny   shady")   # "sunny shady" — also collapses internal spaces

# Apply after every import
penguins |>
  mutate(across(where(is.character), str_trim))

Whitespace is invisible and causes real bugs

After importing from Excel, always run tabyl(species) or count(species) to check for phantom duplicate levels. "sunny" and "sunny " will each appear as a separate row — the giveaway.

3 · Detecting patterns

str_detect("brook trout", "trout")   # TRUE
str_detect("sculpin",     "trout")   # FALSE

# Filter rows where species name contains "trout"
tibble(species = c("brook trout","lake trout","sculpin","walleye")) |>
  filter(str_detect(species, "trout"))

# Case-insensitive search
str_detect("Brook Trout", regex("trout", ignore_case = TRUE))  # TRUE

4 · Replacing patterns

str_replace("bill_length_mm",    "_mm", "")    # "bill_length" — first match
str_replace_all("length_mm_mm",  "_mm", "")    # "length"      — all matches

# Clean column names that have units embedded like "Length (mm)"
tibble(`Length (mm)` = 1:3, `Mass (g)` = 4:6) |>
  rename_with(~ str_replace_all(.x, "\\s*\\(.*\\)", "")) |>  # remove (units)
  rename_with(str_to_lower) |>
  rename_with(~ str_replace_all(.x, " ", "_"))

5 · Extracting substrings

# Extract a pattern from a field label like "trout_S01_2026"
tibble(label = c("trout_S01_2026","sculpin_S02_2026","bass_S01_2025")) |>
  mutate(
    species = str_extract(label, "^[a-z]+"),     # first word
    site    = str_extract(label, "S\\d+"),        # "S" + digits
    year    = str_extract(label, "\\d{4}$")       # 4-digit year at end
  )

6 · Splitting and combining strings

# Split "Lake_Superior_Site01" into components
tibble(id = c("Lake_Superior_Site01","Lake_Michigan_Site02")) |>
  separate_wider_delim(
    cols  = id,
    delim = "_",
    names = c("water_body_1","water_body_2","site")
  ) |>
  unite("lake", water_body_1, water_body_2, sep = " ")

# Pad site codes to consistent width
str_pad(1:5, width = 2, pad = "0")    # "01" "02" "03" "04" "05"

separate_wider_delim() vs separate()

separate_wider_delim() is the modern tidyr function (tidyr ≥ 1.3). The older separate() still works but is superseded. Use separate_wider_delim() in new code.

PART 2 · Factors with `forcats`

7 · Why factors control plot order

By default, R puts categorical variables in alphabetical order on plot axes. That is almost never the order you want. Factors let you set the order explicitly.

# Default: Adelie, Chinstrap, Gentoo (alphabetical — fine here by accident)
penguins |> count(species) |>
  ggplot(aes(x = species, y = n)) + geom_col()

8 · `fct_relevel()` — set a custom order

penguins |>
  mutate(species = fct_relevel(species,
                               "Gentoo","Chinstrap","Adelie")) |>
  count(species) |>
  ggplot(aes(x = species, y = n)) +
  geom_col(fill = "steelblue", alpha = 0.7) +
  labs(title = "Custom order: Gentoo, Chinstrap, Adelie") +
  theme_minimal()

9 · `fct_reorder()` — sort by a numeric value

The most useful forcats function. Sorts bars by their actual value — almost always better than alphabetical.

penguins |>
  drop_na(body_mass_g) |>
  group_by(species) |>
  summarise(mean_mass = mean(body_mass_g), .groups = "drop") |>
  mutate(species = fct_reorder(species, mean_mass)) |>   # sort ascending
  ggplot(aes(x = species, y = mean_mass, fill = species)) +
  geom_col(alpha = 0.7) +
  labs(x = NULL, y = "Mean body mass (g)",
       title = "Bars sorted by mean body mass") +
  theme_minimal() +
  theme(legend.position = "none")

Sort descending with desc()

mutate(species = fct_reorder(species, desc(mean_mass)))

Or for horizontal bar charts, coord_flip() combined with ascending sort puts the largest bar at the top — the most readable layout for many groups.

10 · `fct_infreq()` — sort by count

flights |>
  mutate(carrier = fct_infreq(carrier)) |>
  count(carrier) |>
  ggplot(aes(x = carrier, y = n)) +
  geom_col(fill = "steelblue", alpha = 0.7) +
  labs(title = "Airlines sorted by flight count (most to least)") +
  theme_minimal()

11 · `fct_lump_n()` — collapse rare categories

When you have many levels but only want to highlight the top ones:

flights |>
  mutate(carrier = fct_lump_n(carrier, n = 5)) |>
  count(carrier, sort = TRUE)
# Top 5 carriers kept; all others collapsed to "Other"

12 · Ecological example — species by body size

The canonical use case: fish or species ordered from smallest to largest on a horizontal axis.

fish_df |>
  group_by(species) |>
  mutate(species = fct_reorder(species, mass_g, .fun = mean)) |>
  ggplot(aes(x = species, y = mass_g, fill = species)) +
  geom_boxplot(alpha = 0.6) +
  coord_flip() +
  labs(x = NULL, y = "Mass (g)",
       title = "Fish species ordered by mean body mass") +
  theme_minimal() +
  theme(legend.position = "none")

Quick reference

Strings

Task	Code
Lowercase	`str_to_lower(x)`
Title case	`str_to_title(x)`
Trim spaces	`str_trim(x)`
Collapse internal spaces	`str_squish(x)`
Detect pattern	`str_detect(x, "pattern")`
Replace first match	`str_replace(x, "old", "new")`
Replace all matches	`str_replace_all(x, "old", "new")`
Extract pattern	`str_extract(x, "regex")`
Pad to width	`str_pad(x, width = 3, pad = "0")`
Apply to all chr cols	`mutate(across(where(is.character), str_trim))`

Factors

Task	Code
Custom order	`fct_relevel(f, "B","A","C")`
Sort by numeric variable	`fct_reorder(f, numeric_col)`
Sort by frequency	`fct_infreq(f)`
Collapse rare levels	`fct_lump_n(f, n = 5)`
Rename levels	`fct_recode(f, "New" = "Old")`

End of Common Code 15 — Strings and Factors. Next: Common Code 16 — Writing functions.

Other Formats

Strings and factors

PART 1 · Strings with stringr

1 · Case conversion

2 · Trimming whitespace

3 · Detecting patterns

4 · Replacing patterns

5 · Extracting substrings

6 · Splitting and combining strings

PART 2 · Factors with forcats

7 · Why factors control plot order

8 · fct_relevel() — set a custom order

9 · fct_reorder() — sort by a numeric value

10 · fct_infreq() — sort by count

11 · fct_lump_n() — collapse rare categories

12 · Ecological example — species by body size

Quick reference

PART 1 · Strings with `stringr`

PART 2 · Factors with `forcats`

8 · `fct_relevel()` — set a custom order

9 · `fct_reorder()` — sort by a numeric value

10 · `fct_infreq()` — sort by count

11 · `fct_lump_n()` — collapse rare categories