r-style

R Coding Style

When writing R code, follow these conventions for consistency and readability.

Core Principles

Pipe-based workflows - Structure operations as pipelines rather than nested function calls or chains of temporary variables. Either |> or %>% is fine; be consistent within a script.

# Good - reads top to bottom
result <- data |>
  filter(year == 2026) |>
  group_by(site) |>
  summarize(mean_count = mean(count, na.rm = TRUE))

# Avoid - nested and harder to follow
result <- summarize(group_by(filter(data, year == 2026), site), mean_count = mean(count, na.rm = TRUE))

dplyr for data manipulation - Use filter(), mutate(), select(), summarize(), group_by(), arrange(), and rename() as the default verbs for working with data frames. Reach for across() when applying the same operation to multiple columns.

tidyr for reshaping - Use pivot_longer() and pivot_wider() instead of reshape(), melt(), or cast(). Use separate(), unite(), unnest() when cleaning messy columns.

ggplot2-based plotting - Build visualizations with ggplot2 as the default, not base plot(), barplot(), hist(), etc. Use + to layer geoms and customize with theme(), labs(), and scale_* functions. Other plotting packages that build on or complement ggplot2 are welcome - for example, plotly (including ggplotly() for interactive versions of ggplot2 charts), patchwork, or ggiraph.

purrr over loops - Prefer map(), map_dfr(), walk(), and map2() for iteration instead of for loops. This keeps code in the functional, pipe-friendly style and avoids manual index management. That said, a simple for loop is fine when it's genuinely clearer - for example, when each iteration has side effects or the logic is complex enough that map() would obscure it.

readr / readxl for reading data - Use read_csv() instead of read.csv(), read_excel() instead of xlsx::read.xlsx(). These return tibbles and handle types more predictably.

here::here() for repo file paths - Use here::here() to build paths to files within the project or repo. This keeps paths relative to the project root and works regardless of the working directory, so there's no need for setwd(). Hardcoded absolute paths are fine for network drives (e.g., "Z:/Shared/RRC/...") since those aren't relative to the repo.

# Good - relative to project root
data <- read_csv(here::here("data", "survey_results.csv"))

# Good - network drive path is absolute, that's fine
output_path <- "Z:/Shared/RRC/M/DATA/SEM/SEM (YR5) 2026/field_tracking"

# Avoid - setwd() and manual relative paths
setwd("C:/Users/someone/repos/my-project")
data <- read_csv("data/survey_results.csv")

When Base R Is Fine

Being practical matters more than being pure. Base R is perfectly acceptable when:

• A simple operation doesn't benefit from pipes (e.g., nrow(df), names(df), length(x))
• String operations are simple enough that paste0() or gsub() is clearer than loading stringr
• You're writing a quick one-off check, not production code
• Package dependencies matter and the base R version is equally readable

The guiding question: would a colleague reading this code find the tidyverse version easier to follow? If yes, use tidyverse. If it's a wash, either is fine.

Package Preferences at a Glance

Task	Prefer	Instead of
Read CSV	readr::read_csv()	read.csv()
Read Excel	readxl::read_excel()	xlsx::read.xlsx()
Write Excel	writexl::write_xlsx()	xlsx::write.xlsx(), openxlsx::write.xlsx()
Repo file paths	here::here()	setwd(), manual relative paths
Filter rows	dplyr::filter()	subset(), df[df$x > 1, ]
Add/modify columns	dplyr::mutate()	transform(), df$new <- ...
Summarize	dplyr::summarize()	aggregate(), tapply()
Reshape long	tidyr::pivot_longer()	reshape(), melt()
Reshape wide	tidyr::pivot_wider()	reshape(), dcast()
Iterate	purrr::map()	lapply(), for loops
String ops	stringr::str_*()	grep(), gsub(), sub()
Plotting	ggplot2, plotly, etc.	plot(), barplot(), hist()
Dates	lubridate::ymd(), etc.	as.Date(), strptime()