6

Download PDF Quarto (.qmd)

R Packages II

The first two questions refer to the kknn package (https://github.com/KlausVigo/kknn), which implements a popular predictive modeling algorithm called \(k\)-nearest neighbors. Install it with install.packages("kknn") and load it with library(kknn) before you start.

  1. Use data() to list the data sets stored in kknn. Then try typing the name of one of them, glass, at the console: unlike most packages you’ve used, it isn’t there until you run data(glass). What is this property of a package called, and which field of DESCRIPTION turns it on? Is that field in kknn’s DESCRIPTION?
  1. What is the name of the file in the package that is edited to create the help files for these data sets?

The rest of the questions use greetr, the package you made in PS 5, rebuilt with usethis the way we rebuilt mypkg in lecture. Download greetr.zip, unzip it, and open the greetr folder in Positron (or RStudio). hello() is in R/hello.R, and DESCRIPTION is filled in.

  1. Add the following tibble to your package as a data set called greetings. Write the command you used, and say which file it created.
library(tibble)
greetings <- tibble(
  greeting = c("Hello", "你好", "Hola", "Hei", "السلام علیکم"),
  language = c("English", "Chinese", "Spanish", "Norwegian", "Urdu")
)
  1. Document this data set using roxygen2 so that when a user types ?greetings at the console they see a description of the data set, its format (the columns and their values), and its source (you can cite STAT 133). Please transcribe your documentation below.
  1. Paste this Roxygen block into R/hello.R, directly above hello(), and fill in the three blanks: the argument, the return value, and an example. Run devtools::document() and preview the result with ?hello. Write your three completed lines below.
#' Say hello
#'
#' Prints a greeting to whoever is named in `x`.
#'
#' @param ___
#' @returns ___
#' @examples
#' ___
#' @export
  1. Run devtools::check(). How many errors, warnings, and notes does it report, and what does the note say needs fixing? Fix it: change the License field in DESCRIPTION to MIT + file LICENSE, and add a plain-text file called LICENSE at the top of the package with two lines, YEAR: 2026 and COPYRIGHT HOLDER: followed by your name. Run check() again. What does it report now?

Testing

You’ll need the testthat package for these problems: install it with install.packages("testthat") if you haven’t already. To run tests at the console, outside a package, load it with library(testthat).

  1. Here is a function and three tests for it. Without running the code, predict the ? values in the report that is returned by test(): [ FAIL ? | WARN 0 | SKIP 0 | PASS ? ]. For any test that fails, is the function wrong or the test?
celsius <- function(f) {
  (f - 32) * 5 / 9
}

test_that("celsius() converts the boiling point", {
  expect_equal(celsius(212), 100)
})

test_that("celsius() converts the freezing point", {
  expect_equal(celsius(0), 32)
})

test_that("celsius() converts a vector of temperatures", {
  temps <- celsius(c(32, 212))
  expect_length(temps, 2)
  expect_equal(temps, c(0, 100))
})
  1. A classmate wrote this test for celsius() from the last problem, ran it, and it passed. Why did it pass? Change the test so it would catch the mistake.
test_that("celsius() rejects text", {
  expect_error(celsuis("hot"))
})
  1. Here is part of a package with the strsplit1() from lecture and a new function that counts the words in a sentence, followed by the report from test().
R/n_words.R
n_words <- function(x) {
  length(strsplit1(x, split = " "))
}
Failure ('test-n_words.R:6:3'): n_words() counts words with extra spaces
Expected `n_words("two  spaces")` to equal 2.
`actual`:   3
`expected`: 2
[ FAIL 1 | WARN 0 | SKIP 0 | PASS 4 ]

  1. Which file and line holds the expectation that failed, and what is the test’s description?

  2. How many expectations passed?

  3. Why does n_words() return 3? (Hint: look closely at the input, and run strsplit1() on it.)

  4. Is this input a typical input, an edge case, or an error case?

The next four problems use your greetr package from the R Packages II problems. Open it in Positron (or RStudio), and run library(devtools) at the console.

  1. Set greetr up for testing, with a test file for hello(). Write the two commands you used. What files did they create, and what changed in DESCRIPTION?
  1. Replace the placeholder in test-hello.R with two tests: one that hello("Oski") returns "Hello Oski!", and one that hello() given two names returns a character vector of length 2. Run test(). What does its last line say?
  1. hello() prints its greeting as well as returning it, which clutters the output of test(). Refactor hello(): remove the print() so it just returns the string. Run test() again. Do the tests still pass, and what does that tell you? Does hello("Oski") still show its greeting when you run it at the console?
  1. A user files a bug report: “hello(NA) says Hello NA!.” Fix it with test-driven development:

  1. Red: add a test to test-hello.R that hello(NA) throws an error whose message mentions "missing". Run test() and confirm that it fails.

  2. Green: change hello() so it throws that error, using stop(), whenever any element of x is NA. Run test() until it passes.

  1. Testing isn’t only for packages. Suppose you’re about to analyze storms (from dplyr) and are relying on two facts about it:
  • no row appears twice;
  • every row with hurricane-force winds (wind of 64 knots or more) has a category.

Write each as its own stopifnot() call. Both fail. For each, look at the rows that break it and decide whether the data or the expectation is wrong.

  1. Optional Challenge: Write the grade calculator from lecture, test first.
  1. Create a test file for a function letter_grade() that holds every promise from the lecture’s activity:
Score 95 85 42 98 91 100 58 62 NA 0.85 0.91 86.5 89.5 0.865
Grade A B F A+ A- A+ F D- F B A- B+ A- B+

Add a test for a score of 1, deciding what it should get.

  1. Write letter_grade() so that every test passes, for a vector of scores. Two tools you haven’t seen will help: x %% 10 is the last digit of a whole number x, and floor(x + 0.5) rounds halves up (round() rounds halves to the even number, so round(86.5) is 86).

APIs I

The BART API serves live data, so some answers in this section (how many trains are about to depart, how long the wait is) depend on when you run your code. For those there is no single right answer.

Revisit the homepage for the BART Legacy API discussed in Lecture.

https://www.bart.gov/schedules/developers/api

  1. What is the public key that can be used to access data through the legacy API?

Next visit the documentation site: https://api.bart.gov/docs/overview/index.aspx. Note that each category of data has its own endpoint / base URL, and all take a similar format (e.g. for estimated departure data use https://api.bart.gov/api/etd.aspx.)

  1. Consider the following URL. Which part is the endpoint, and what are the parameters and their values? Using the documentation, what does this request ask for?
https://api.bart.gov/api/stn.aspx?cmd=stninfo&orig=12TH&key=MW9S-E7SL-26DU-VV8V&json=y
  1. Provide the URL that will query the API for a list of all BART stations in JSON format.
  1. Provide the corresponding shell/terminal command.
  1. Provide the corresponding R code that will perform the query and parse the JSON response into an R list object. Use simplifyVector = TRUE in resp_body_json(), so that JSON arrays of records become data frames.
  1. Subset the list to return a data frame that contains the station abbreviation, name, latitude, longitude, address, city, county, state, and zipcode for each station. Which county has the greatest number of stations?
  1. Provide the R code that will query the API for estimated departure times (ETDs) from the Downtown Berkeley Station and parse the JSON response into an R list object.
  1. Extract the element called root and then the station element within it. View the object using the data viewer. What type of object is it and what are its dimensions?
  1. The unnest() function from the tidyr package expands a list-column into ordinary rows and columns. Run the code below, then explain what each unnest() step did to the rows and the columns of the data frame.
library(tidyr)
etd_unnested <- etd |>
  unnest(etd)
etd_fully_unnested <- etd_unnested |>
  unnest(estimate)
  1. How long (in minutes) is the shortest estimated wait time for a train departing from Downtown Berkeley Station? Which destination does this train go to? For reference, what time is it now?
  1. Optional: Query the BSA (BART System Advisories) endpoint to retrieve the latest system advisory information. You’re welcome to use any method: web browser, command line, or R. What are the different bits of information that they provide through this endpoint?
  1. Optional Challenge: Use the data provided by the BART API to construct a map of the BART system showing the locations of all stations. You could make it interactive with leaflet or static like the one below.