fasterr

Summary

This package optimizes three commonly used R functions by rewriting them in C++ for better performance using Rcpp. The C++ implementations are faster, especially when working with large datasets or vectors, while keeping the functionality consistent with the original R functions. This package is designed to make workflows faster without changing the functionality of functions.

Installation

The current version of fasterr can be installed by:

remotes::install_github("sds270-f24/fasterr")

You can load this package by:

library(fasterr)

Original Functions

• tidyr::full_seq()
  • This function creates the full sequence of values in a vector. It takes in a numeric vector, a period, and the numerical tolerance for checking periodicity. Our fasterr version does not take in numerical tolerance, instead not taking into account any tolerance.
• tidyr::fill()
  • This function cleans an inputted data.frame by filling NA values. It takes in a direction as input, with the default being “down.” Our fasterr version does not take in a direction, using the default, meaning NA values are filled with the last valid non NA value as you move down the data.frame.
• stats::na.omit()
  • This function cleans an inputted data.frame, removing rows with NA values. The function stores information about which rows have been removed in an attribute of the cleaned data.frame.

Optimized Functions

Example usage and time demonstrations.The times are measured in microseconds.

full_seq()

seq <- c(1, 5, 7, 9, 17)
step <- 2
    
full_seq_results <- microbenchmark::microbenchmark(
    tidyr_full_seq = tidyr::full_seq(seq, step),
    fasterr_full_seq = fasterr:::full_seq(seq, step),
    times = 100
  )
#> Warning in microbenchmark::microbenchmark(tidyr_full_seq = tidyr::full_seq(seq,
#> : less accurate nanosecond times to avoid potential integer overflows

summary(full_seq_results)
#>               expr   min      lq       mean median      uq       max neval
#> 1   tidyr_full_seq 14842 15682.5 1567408.27  16113 16748.5 155117473   100
#> 2 fasterr_full_seq   861   984.0    3358.31   1148  1271.0    222015   100
ggplot2::autoplot(full_seq_results)

fill_forward()

fill_results <- microbenchmark::microbenchmark(
    tidyr_fill = tidyr::fill(airquality, "Ozone", "Solar.R"),
    fasterr_fill_forward = fasterr:::fill_forward(airquality,c("Ozone", "Solar.R")),
    times = 100
  )

summary(fill_results)
#>                   expr      min       lq       mean  median        uq      max
#> 1           tidyr_fill 1668.577 1697.092 1979.76741 1779.40 1932.3505 9542.258
#> 2 fasterr_fill_forward    2.296    2.665    4.31197    3.69    4.4075   29.561
#>   neval
#> 1   100
#> 2   100
ggplot2::autoplot(fill_results)

omit_na()

omit_na_results <- microbenchmark::microbenchmark(
    stats_na.omit = stats::na.omit(airquality),
    fasterr_omit_na = omit_na(airquality),
    times = 100
  )

summary(omit_na_results)
#>              expr    min      lq     mean  median     uq     max neval
#> 1   stats_na.omit 42.599 43.7265 46.97042 45.2845 46.904  94.669   100
#> 2 fasterr_omit_na 26.568 27.9825 31.06119 28.8230 30.053 106.846   100
ggplot2::autoplot(omit_na_results)

Contributors

• Emma Ruckle
• Molly Daniel