kit

Fast data manipulation functions implemented in C for large datasets. Provides vectorized alternatives to base R functions with significant performance improvements.

Installation

# From CRAN
install.packages("kit")

# Development version
install.packages("kit", repos = "https://fastverse.r-universe.dev")

Features

Parallel Statistical Functions

Vector-valued functions operating in parallel over vectors or data frames:

• psum, pprod, pmean: Parallel sum, product, and mean (similar to pmin/pmax)
• pall, pany: Parallel all/any operations
• pcount, pcountNA: Count occurrences of values or NAs
• pfirst, plast: First/last non-missing values

x <- c(1, 3, NA, 5)
y <- c(2, NA, 4, 1)
psum(x, y, na.rm = TRUE)  # [1] 3 3 4 6
pmean(x, y, na.rm = TRUE) # [1] 1.5 3.0 4.0 3.0

Vectorized and Nested Switches

Fast vectorized conditional logic:

• iif: Fast replacement for ifelse() with attribute preservation
• nif: Nested if-else (SQL CASE WHEN equivalent)
• vswitch, nswitch: Vectorized switch statements

iif(x > 2, x, x - 1)  # Preserves attributes unlike base::ifelse
nif(x == 1, "one", x == 2, "two", default = "other")

Sorting

• psort: Parallel sort for character vectors
• topn: Efficient partial sort (top N values) without full sorting

topn(x, n = 6L, decreasing = TRUE)  # Much faster than order()[1:6]

Factors

• charToFact: Fast character-to-factor conversion
• setlevels: Change factor levels by reference

Unique Values and Counts

• funique, fduplicated: Fast unique/duplicated operations
• uniqLen: Fast equivalent to length(unique(x))
• count, countNA, countOccur: Count element occurrences

funique(iris$Species)  # Faster than base::unique
uniqLen(iris$Species)  # Faster than length(unique())

Miscellaneous

• fpos: Find matrix/vector positions within larger structures
• shareData, getData, clearData: Share data between R sessions

Documentation

Full documentation available at: https://fastverse.github.io/kit/

License

GPL-3