R Documentation

Purpose

Write and manipulate R package documentation using roxygen2 and the Rd format. Covers structured documentation (sections, lists, tables), mathematical notation (\eqn/\deqn), and programmatic Rd parsing/extraction.

Roxygen2 Structure

Roxygen2 blocks map to Rd sections. The implicit structure is:

#' Title (first paragraph)
#'
#' Description (second paragraph).
#'
#' Details (third+ paragraphs).
#' @param x Argument docs.
#' @return What it returns.
#' @export
my_func <- function(x) {}

Explicit tags override implicit placement: @title, @description, @details.

Multiple @details blocks

@details can appear multiple times — all merge into a single Details section:

#' @details
#' First block.
#'
#' @param x Input.
#'
#' @details
#' Second block (merged with first in output).

Custom sections

Two equivalent approaches:

#' @section Algorithm:
#' Description of the algorithm.

Or with Markdown headings (requires Roxygen: list(markdown = TRUE) in DESCRIPTION):

#' @details
#' # Algorithm
#' Description of the algorithm.
#'
#' ## Phase 1
#' ...

Top-level # creates \section{}. ##/### create subsections. Headings only valid after @description or @details.

Lists, Tables, and Preformatted Blocks

Gotcha: \describe{\item{label}{desc}} uses two-arg \item. \itemize{\item} and \enumerate{\item} use bare \item. Mixing these up is a common error.

#' \tabular{lcr}{
#'   Name \tab Value \tab Unit \cr
#'   foo  \tab 42    \tab kg   \cr
#' }

Column alignment: l/c/r. Columns: \tab. Rows: \cr.

Use \preformatted{} for matrix output examples — preserves exact whitespace.

Mathematical Notation

Two macros for math in Rd:

Always use the two-argument form for anything beyond trivial expressions. The second argument is the plain-text fallback shown in terminal ?help.

Rendering by output format

Quick examples

#' The variance is \eqn{\sigma^2}{sigma^2}.
#'
#' The expected value:
#' \deqn{E[X] = \sum_{i} p_i x_i}{E[X] = sum(p_i * x_i)}

Symbol definition rule

YOU MUST define every symbol before or at its first use. Never introduce P, O, t, etc. in a formula without saying what they are in the surrounding prose.

#' Let \eqn{P_{ij}}{P[i,j]} be the joint probability and
#' \eqn{O_{ij}}{O[i,j]} the outcome value. Then:
#' \deqn{E[X] = \sum_{i,j} P_{ij} \cdot O_{ij}}{E[X] = sum(P[i,j] * O[i,j])}

For detailed math formatting syntax, multi-line equations, mathjaxr, and gotchas, load references/math-formatting.md.

Programmatic Rd Parsing

tools::parse_Rd() parses the entire Rd file into a nested tagged list. There is no partial parsing — selective reading happens post-parse by filtering tags.

Extract tag names

rd <- tools::parse_Rd("man/my_function.Rd")
tags <- vapply(rd, attr, character(1L), "Rd_tag")
# => "\\name", "\\alias", "\\title", "\\description", ...

Extract a section's text

idx <- which(tags == "\\description")
trimws(paste(unlist(rd[[idx[1L]]]), collapse = ""))

Render one section to text

fragment <- rd[which(tags == "\\description")]
class(fragment) <- "Rd"
tools::Rd2txt(fragment, out = stdout(), fragment = TRUE)

The fragment = TRUE flag is essential when rendering subsets.

For full patterns (batch extraction, HTML rendering, helper packages like gbRd/Rdpack), load references/rd-parsing.md.

Verification Workflow

After writing or editing roxygen2 documentation:

# 1. Regenerate Rd files
roxygen2::roxygenise()

# 2. Parse and validate (should produce no output = success)
rd <- tools::parse_Rd("man/MyTopic.Rd")
tools::checkRd(rd)

# 3. Render to text (verify terminal help looks right)
tools::Rd2txt(rd, out = stdout())

Check for:

• •Unbalanced braces (most common Rd error)
• •Missing ASCII fallbacks on complex \eqn/\deqn
• •Symbols used before definition
• •\preformatted{} blocks rendering with correct whitespace

Key Rules

• •$...$ and $$...$$ are NOT supported in Rd. Use \eqn{}/\deqn{}.
• •Markdown is NOT processed inside \eqn{}/\deqn{} (they are verbatim).
• •\\ for LaTeX newlines requires \\\\ in roxygen2 source.
• •\describe{\item{}{}} takes two-arg items. \itemize{\item} takes bare items.
• •Always verify in text output (Rd2txt), not just HTML — terminal users see the ASCII fallback.

References (Load on Demand)

• •references/math-formatting.md — Load when writing complex math (multi-line equations, Greek letters, mathjaxr, amsmath environments) or debugging math rendering issues.
• •references/rd-parsing.md — Load when programmatically extracting or manipulating Rd file contents (batch processing, section extraction, format conversion, helper packages).