tidyverse-elegant-r

name	tidyverse-elegant-r
description	Write elegant modern tidyverse R for this course repo—always = for assignment (never <-), dplyr 1.1+ (join_by, .by), native pipe, explicit library() per file, vectorized mutate/case_when, httr2 and purrr possibly/safely, rlang data-mask patterns. Use when editing 03_query_ai, 10_data_management, fixer, or refactoring API/data pipelines.

Tidyverse-elegant R (SYSEN5381 / dsai)

Project rule: .cursor/rules/tidyverse_elegant.mdc. This skill adds links, package map, and patterns.

Canonical links

• Tidyverse style guide — Syntax, Pipes, Functions, ggplot2
• Tidyverse packages
• Learning the tidyverse with AI — review pipelines step-by-step; remove unused packages and print() noise
• Writing performant code with tidy tools — profile with profvis; compare with bench::mark(); dplyr-first until a hotspot is proven
• Modern R Development Guide (gist) — join_by, .by, rlang, purrr 1.0+
• httr2 — req_error, Wrapping APIs, req_perform_iterative
• Formatting helpers: styler, lintr

Package map (load only what you use)

Package	Use for
dplyr	filter, mutate, summarise, joins, if_else, case_when, .by, join_by
tidyr	pivot_, nest/unnest, separate_, replace_na, complete
readr	read_csv, write_csv, read_lines, write_lines
tibble	tibble, as_tibble, tidy printing
purrr	map, walk, list_rbind, possibly, safely
stringr	str_*, vectorized strings in mutate
forcats	fct_*
lubridate	Dates and datetimes
ggplot2	Graphics
httr2	HTTP: request, req_, resp_
jsonlite	JSON with APIs
glue	Interpolation (when many pasted strings)
reprex	Minimal reproducible examples when debugging

Avoid library(tidyverse) as a default; attach the rows you need.

Modern dplyr patterns (gist-aligned)

# Joins — join_by() and join quality
x |>
  dplyr::inner_join(y, dplyr::join_by(id == id), multiple = "error", unmatched = "error")

# Per-operation grouping — .by
x |>
  dplyr::summarise(m = mean(value), .by = c(grp, year))

# purrr 1.0+ — row-bind results
purrr::map(splits, fit_one) |> purrr::list_rbind()

Vectorized vs scalar if

• Column logic: dplyr::case_when(), if_else(), coalesce(), tidyr::replace_na(), stringr inside mutate().
• Script flow: if (file.exists(".env")) readRenviron(".env") is fine once at top.

rlang snippets for function arguments

my_mean = function(data, var) {
  data |> dplyr::summarise(m = mean({{ var }}, na.rm = TRUE))
}

my_mean_by_name = function(data, var) {
  data |> dplyr::summarise(m = mean(.data[[var]], na.rm = TRUE))
}

httr2 + errors + mapped requests

• Single request: let req_perform() error, or tailor with req_error().
• Many URLs: wrap a small function in purrr::possibly() / safely(), or use req_perform_iterative(..., on_error = "return") when appropriate.
• For typed handling of HTTP conditions, rlang::try_fetch on specific httr2_http_* classes can be clearer than a huge tryCatch.

Performance workflow

1. Run profvis on realistic data.
2. Only optimize hot paths; use bench::mark() to compare; ensure equal outputs.
3. Prefer keeping dplyr in analysis scripts; consider vctrs / lower-level substitutes mainly for package internals or very tight loops (performant-packages).

Relation to tutorial coding_style.mdc

Tutorial rule prefers %>% and heavy commenting. Scoped elegant paths use |>, explicit library(), and vectorized table logic. Assignment is always = (never <-), same as arrows.mdc and the tutorial rule.