The code for the satellite image:

library(terra)
library(here)
library(tidyterra)

fname <- file.path(
    here("slides/figures"), 
    "S2A_MSIL2A_20201014T154231_N0214_R011_T19TBG_20201014T200156_20m.tif")
img <- rast(fname)

ggplot() + 
    geom_spatraster_rgb(data = img, r = 4, g = 3, b = 2, stretch = "lin") +
    theme_void()

Today

  • Control structure
  • Loop functions *apply

Open your code cracker R markdown in your class note folder.

Warm up

  1. Identify all rows where pH greater than 8 and crop is “chickpea”.
  2. For only those rows, replace the current rainfall value with the maximum Nitrogen (N) value in this subset.
  3. From this subset, randomly select 5 rows with seed 10.
    • Compute: mean(rainfall) + mean(temperature) using those 5 rows.
    • Round the final result to 0.
    • Report the resulting value out loud once you finish.

Control structure

Commonly used control structure:

  • if and else, which test a condition and do things correspondingly.
  • switch, which branches
  • for, which execute a loop for a specific number of times.
  • while, which execute a loop when a condition is true.
  • repeat, which execute a repeat work until break out the loop.
  • break, which can control to exit a loop.
  • next, which can skip an iteration of a loop.

Note: we will learn later and then mainly use native R loop functions *apply for loops.

Control structure

if - else

# one line
if (4 > 5) print('4 is larger than 5.') else print("4 is smaller than 5.")
if (4 > 5) print('4 is larger than 5.')

# paragraph
if (4 > 5) {
  print('4 is larger than 5.')
} else print("4 is smaller than 5.")

# or could get longer
if (4 > 5) {
  print('4 is larger than 5.')
} else if (4 == 5){
  print("4 is equal to 5.")
} else {
  print("4 is smaller than 5.")
}

# ifelse
ifelse(4 > 5, print('4 is larger than 5.'), print("4 is smaller than 5."))
x <- 4
y <- 10
z <- ifelse(x > y, x + 1, y - 1)

Control structure

Your turn to use “if-else” on Crops dataset

  • Isolate the data for Row 70 of the dataset.
  • Logic Check:
    • If the label is “rice”:
      • Add 2 to its temperature.
      • If this new temperature is higher than 25, print: “Vulnerable to climate change”.
      • Otherwise, print: “Resistant to climate change”.
    • If the label is not “rice”, print: “Not rice”.

Control structure

switch

x <- 'b'

# If - else structure
if (x == "a") {
    "option 1"
  } else if (x == "b") {
    "option 2" 
  } else if (x == "c") {
    "option 3"
  } else {
    stop("Invalid `x` value")
  }

# switch
switch(x,
    a = "option 1",
    b = "option 2",
    c = "option 3",
    stop("Invalid `x` value")
)

The last component of a switch() should always throw an error, otherwise unmatched inputs will invisibly return NULL.

Control structure

Your turn to use “switch” on Crops dataset

  • Randomly select 1 rows from the dataset. Use seed 10.
  • Check:
    • If the label is “pigeonpeas”, print “Drought tolerant”
    • If the label is “kidneybeans”, print “Nutritious”
    • If the label is “muskmelon”, print “Juicy”
    • All others, “Invalid value”

Control structure

for

# One line
for (i in 1:10) print(i)

# code paragraph
for (i in 1:10){
  print(i)
  print(i + 1)
}

# nested for
for (i in 1:10){
  for (j in 2:11){
    print(x + y)
  }
}

Control structure

Your turn to use “for” on Crops dataset

  • Loop over the first 10 rows of the crops dataset
  • Identify “Acidic High-Nitrogen” samples:
    • If pH is less than 6.0 and N is greater than 80, print “Row [Number]: Action Required”. E.g. “Row 1: Action Required”. (Hint: use paste)
    • Otherwise, print “Row [Number]: Stable”

Control structure

while and repeat

# while
x <- 0 # important
while(x < 10){
  print(x)
  x <- x + 1 # Change the checking condition
}

# repeat
x <- 0
repeat{
  print(x)
  x <- x + 1
  if (x == 10){
    print('Finish repeat!')
    break
  }
}

Control structure

Do the same but with “while” and “repeat”

  • Loop over the first 10 rows of the crops dataset
  • Identify “Acidic High-Nitrogen” samples:
    • If pH is less than 6.0 and N is greater than 80, print “Row [Number]: Action Required”
    • Otherwise, print “Row [Number]: Stable”

Control structure

break and next

x <- 0
repeat{
  x <- x + 1 # what if we move this line after the first if?
  if (x < 5){
    print('Skip this interation!')
    next
  }
  if (x == 10){
    print('Finish repeat!')
    break
  }
  print(x)
}

Why *apply instead of for in R?

  • Performance
  • Code robustness and readability

*apply family

  • apply()
  • lapply()
  • sapply()
  • tapply()
  • mapply()

These functions allow you process the data in batches looply. The primary difference among these functions is the object type of the input and output.

apply()

apply(x, MARGIN, FUN, …)

  • x: an array. other types (e.g. data.frame) will convert to matrix.
  • MARGIN: we could take it as the dimension to take batches, 1 indicates rows, 2 indicates columns, etc. etc. It can take more than one values, e.g. c(1, 2) means across both rows and columns.
  • FUN: the function applied to the batch.
  • returns a vector or array, sometimes a list (simplify = FALSE).

apply()

m <- matrix(c(1:5, 11:15, 21:25), nrow = 5, ncol = 3)
apply(m, 1, sum)
[1] 33 36 39 42 45

apply()

m <- matrix(c(1:5, 11:15, 21:25), nrow = 5, ncol = 3)
m
apply(m, 2, function(x) x + 2)

apply(m, 2, function(x) x + 2, simplify = FALSE)

apply(m, 1, function(x) x + 2) ## what will happen?

lapply()

lapply(x, FUN, …)

  • x: a vector (atomic and list).
  • FUN: the function applied to each element of x.
  • returns a a list of the same length as x.

lapply() is arguably the most widely used function in apply family, because it is super well organized. We can always convert the returned list to other types later (e.g. unlist()).

lapply()

set.seed(123)
l <- list(A = c(1:9), B = rnorm(5), 
          C = data.frame(x = sample(1:100, 3), y = runif(3)))
l
$A
[1] 1 2 3 4 5 6 7 8 9

$B
[1] -0.56047565 -0.23017749  1.55870831  0.07050839  0.12928774

$C
   x          y
1 14 0.89982497
2 25 0.24608773
3 90 0.04205953
lapply(l, sum)
$A
[1] 45

$B
[1] 0.9678513

$C
[1] 130.188

lapply()

The magic of do.call with dplyr

Try by yourself:

  • do.call(c, lapply(l, sum))
  • do.call(rbind, lapply(l, sum))
  • do.call(cbind, lapply(l, sum))

lapply() vs sapply()

sapply()

Almost the same as lapply(), but tries to simplify the output to the possible simplest data structure by default. By setting simplify = FALSE, it will return a list as well.

set.seed(123)
l <- list(A = c(1:9), B = rnorm(5), 
          C = data.frame(x = sample(1:100, 3), y = runif(3)))
sapply(l, sum)


Q: How to modify the code to calculate the max − min price for each element in l ? (Hint: use anonymous function)

Others

tapply(X, INDEX, FUN)

df <- data.frame(price = sample(18:65, 4), product = c('mouse', 'keyboard'))
summary_df <- tapply(df$price, df$product, mean) # max or min, etc. Any function
df
  price  product
1    44    mouse
2    45 keyboard
3    26    mouse
4    46 keyboard
summary_df
keyboard    mouse 
    45.5     35.0

Q: How to modify the code to calculate the max − min price for each product group? (Hint: use anonymous function)

Others

mapply(FUN, …, MoreArgs = NULL, SIMPLIFY = TRUE, USE.NAMES = TRUE)

mapply(function(x, y){
  switch(y,
         mouse = x + 10,
         keyboard = x / 100
         )
}, df$price, df$product)
[1] 54.00  0.45 36.00  0.46

apply practice on crops dataset

  • Calculate the sum of N, P, K for all ROWs. (use apply)
  • Get the range of column of temperature, humidity and rainfall (use lapply, and function range). And collect the results into a matrix by row.
  • Calculate mean rainfall per crop types. (use tapply)
  • Calculate a custom score N + (P * 2) across all rows. (use mapply)
  • Subset only numeric columns of crops dataset
    • Try colSums and rowMeans.

Share your solution of the second task in Chat.

Homework