tidy-r

name	tidy-r
description	Modern tidyverse patterns, style guide, and migration guidance for R development. Use this skill when writing R code, reviewing tidyverse code, updating legacy R code, or enforcing consistent style. Covers native pipe usage, join_by() syntax, .by grouping, pick/across/reframe, filter_out/when_any/when_all, recode_values/replace_values/replace_when, tidyselect helpers, .data/.env pronouns, stringr, naming conventions, and readr.
license	CC-BY-4.0
metadata	{"author":"Ulrich Atz","r_version":">= 4.5.0","tidyverse_version":">= 2.0.0","dplyr_version":">= 1.2.0"}
allowed-tools	Read, Edit, Write, Grep, Glob, Bash, mcp__r-btw__*

Modern Tidyverse R Reference

Code from blog posts and StackOverflow often uses deprecated APIs, magrittr pipes, or base R patterns where a modern tidyverse function exists. This guide encodes the current recommended approach.

Reference files

Consult the appropriate reference file for detailed patterns and examples:

Topic	Reference file	When to consult
Joins	joins.md	Merging data, *_join, join_by, matching rows, lookup tables
Grouping & columns	grouping.md	.by, group_by, across, pick, reframe, column operations
Recoding & replacing	recode-replace.md	recode_values, replace_values, replace_when, filter_out, when_any, when_all
Strings	stringr.md	String manipulation, regex, str_* functions, text processing
Tidy selection	tidyselect.md	Column selection helpers, where(), all_of(), any_of(), boolean ops, .data/.env pronouns
Style	tidyverse-style.md	Naming, formatting, spacing, error messages, cli::cli_abort
Migration	migration.md	Updating old code, base R conversion, deprecated functions

For requests that span multiple topics (e.g., "rewrite this old code" touches migration + style), read multiple files.

Related skills

tidy-r is the default for in-memory tidyverse work. Reach for a companion skill when the task outgrows it -- the data-frame workflow and |> style carry over:

Reach for	When
collapse-r	Performance matters on large or heavily-grouped in-memory data, or you need weighted statistics or panel/time-aware ops (lags, growth, between/within). f-prefixed verbs.
duckplyr-r	Local data too big for RAM, or reading Parquet/CSV/JSON, while keeping dplyr syntax (DuckDB engine, dplyr-identical results).
dbplyr-r	Data lives in a remote/connection database (Postgres, Snowflake, BigQuery, SQL Server); dplyr is translated to SQL and run server-side.
r-btw-cli	Look up R help/vignettes, run R CMD check / tests / document(), or search CRAN from the command line.

Core principles

1. Use modern tidyverse patterns -- Prioritize dplyr 1.2+ features, native pipe, and current APIs
2. Write readable code first -- Optimize only when necessary
3. Follow tidyverse style guide -- Consistent naming, spacing, and structure

Quick reference

Pipe and lambda

• Always |>, never %>%
• Use _ placeholder for non-first arguments: x |> f(1, y = _). The placeholder must be named and used exactly once.
• Always \(x), never function(x) or ~ in map/keep/etc.

Code organization

Use newspaper style: high-level logic first, helpers below. Don't define functions inside other functions unless they are very brief.

Grouping

• Prefer .by for per-operation grouping; use group_by() when grouping must persist across multiple operations
• Never add ungroup() before or after .by -- it always returns ungrouped data
• Consolidate multiple mutate(.by = x) calls into one when they share the same .by; keep separate only when .by differs or a later column depends on an earlier one
• Place .by on its own line for readability

Joins

• Use join_by(), never c("a" = "b")
• Use relationship, unmatched, na_matches for quality control
• Use tidylog:: prefix for join verification

Recoding and replacing (dplyr >= 1.2.0)

Task	Function
Recode values (new column)	recode_values()
Replace values in place	replace_values()
Conditional update in place	replace_when()
Complex conditional (new column)	case_when()
Drop rows (NA-safe)	filter_out()
OR conditions	when_any()
AND conditions	when_all()

NA handling

• if_else() and case_when() accept plain NA (since dplyr 1.1.0) -- no need for NA_character_, NA_real_, etc.
• Load tidyna to make mean, sum, sd, etc. ignore NA by default. Avoid repetitive na.rm = TRUE.

Error handling

Use cli::cli_abort() with problem statement + bullets, never stop().

R idioms

• TRUE/FALSE, never T/F
• message() for info, never cat()
• map_*() over sapply() for type stability
• set.seed() with date-time, never 42
• qs2::qs_save()/qs2::qs_read(), never qs

Example

library(tidyverse)

penguins <- penguins |>
  filter_out(is.na(sex)) |>
  mutate(size = case_when(
    body_mass > 4500 ~ "large",
    body_mass > 3500 ~ "medium",
    .default = "small"
  ))

# Coordinates for spatial join below
island_coords <- tribble(
  ~island,      ~latitude,
  "Biscoe",     -65.5,
  "Dream",      -64.7,
  "Torgersen",  -64.8
)

island_summary <- penguins |>
  summarise(
    mean_flipper = mean(flipper_len),
    mean_mass = mean(body_mass),
    n = n(),
    .by = c(species, island)
  ) |>
  left_join(
    island_coords,
    by = join_by(island),
    unmatched = "error"
  ) |>
  arrange(species, island)

Best practices

1. Name variables as nouns, functions as verbs in snake_case
2. Explain "why" in comments, not "what"
3. Place .by on its own line for readability
4. Use .unmatched = "error" in case_when() and recode_values() for defensive programming
5. Use recode_values() over case_match() (dplyr 1.2+ preferred API)
6. Use replace_when() over case_when() with .default when updating a column in place
7. Prefer filter_out() over negated filter() for NA-safe row removal
8. Load tidyna early to eliminate na.rm = TRUE clutter
9. Use tidylog:: for joins to verify row counts and match quality
10. Use qs2 for serialization with .qs2 extension