1. Work on your branch
2. Push your changes on that branch
3. Go to GitHub and open a pull request to merge that branch into main
4. Describe the work you have done
5. Request review
6. Act on feedback
7. Confirm pull request

A benchmark is a controlled test to see how well your code runs.

You are trying to quantify the efficiency of your code, so you can compare it against other methods.

You are also often looking for bottlenecks, or parts of your code that slow down the whole process.

You may recall we’ve actually done some simple benchmarks before!

When we were talking about parallelizing code, we used tictoc to see how long it took code to run.

That’s pretty helpful on it’s own!

library(tictoc)

num_vec = c(1, 2, 3, 4, 5, 6, 7, 8)

tic()

sapply(num_vec, FUN = function(num){
  
  # Say number
  print(paste0("The number is ", num,
          "! I'll wait that many seconds."))
  
  # wait that many seconds
  Sys.sleep(num)
  
  # return number
  return(num)
  
})

toc()

Running sequentially, takes about 37 seconds on my desktop.

CPU

RAM

Harddrive

The bench package provides a simple function, mark() to track how long code takes to run.

It will run the contained code several times to get an average.

It will also tell you how much memory the code uses.

num_vec = c(1, 2, 3, 4, 5, 6, 7, 8)

bench::mark({
  
  sapply(num_vec, FUN = function(num){
  
  # Say number
  print(paste0("The number is ", num,
          "! I'll wait that many seconds."))
  
  # wait that many seconds
  Sys.sleep(num)
  
  # return number
  return(num)
  
})
  
})

Still takes about 37 seconds on my desktop, but will run multiple times to average.

The difference is a semantic one.

Broadly, benchmarks look at individual functions, profiling looks at whole code flows.

Think of it as the difference between tracking a race and a marathon.

Benchmark == Race

Profile == Marathon

Here is the profile of a section of my research code on asset forfeiture.

The top part of the diagram shows how long each line of code takes, as well as memory usage.

The bottom plots how many functions are being run nested inside each other.

There are several tools to profile with in R. I will reccomend profvis.

It creates interactive profile plots that give you an at-a-glance view of code sequences.

I’ve included a small example here:

profvis::profvis({
  # read in class survey
  survey = readRDS(url("https://github.com/Adv-R-Programming/Adv-R-Reader/raw/main/class_survey.rds"))
  # wait a moment
  Sys.sleep(1)
  # use comma_split
  comma_split(survey$pets,
              possible_columns = c("dog", "cat", "fish", "bird", "reptile", "rock", "none"))
})

Topic

Lab 8

To-Do

• Finish Worksheet