Managing data flow

inputs -> Black Box -> outputs

Try to follow Base-R behavior (even if sometimes R function are not consistent).

9.0.1 Exerice 9.1

sample.int("Spanish Inquisition")
# Error: attempt to use zero-length variable name
seq_along(list(1, 2, 3))
# [1] 1 2 3
rep("bob", -1)
# 
rep("bob", -1)
# Error: attempt to use zero-length variable name
rep("bob", TRUE)
rep("bob", FALSE)
strrep("bob", NA)
# NA
train <- rbind(iris3[1:25,,1], iris3[1:25,,2], iris3[1:25,,3])
test <- rbind(iris3[26:50,,1], iris3[26:50,,2], iris3[26:50,,3])
cl <- factor(c(rep("s",25), rep("c",25), rep("v",25)))
class::knn(train, test, cl, k = numeric(0), prob=TRUE)
# Error: attempt to use zero-length variable name

Even if “some” functions check data integrity for us what happens when we multiple functions:

round_rand <- function(n, d)
{
    x <- runif(n)  # runif will check if `n` makes sense
    round(x, d)    # round will determine the appropriateness of `d`
}

Solution:

round_rand2 <- function(n, d)
{
    stopifnot(
        is.numeric(n), length(n) == 1,
        is.finite(n), n > 0, n == floor(n),
        is.numeric(d), length(d) == 1,
        is.finite(d), d > 0, d == floor(d)
    )
    x <- runif(n)
    round(x, d)
}
round_rand2(3, -1)

In some cases, maybe you want to be more permissif:

# [..]
if (length(n) > 1) {
    warning("only the first element will be used")
    n <- n[1]
    }
# [..]

9.0.2 Exercice 9.6

• A vectorised mathematical function:

  • empty vector? non-numeric input? -> No
  • what if it is equipped with the names attribute? -> should not mind them

• an aggregation function

  • what about missing values? -> warnings and na.rm = TRUE
  • empty vector -> No, Error

• a function vectorised with regard to two arguments:

  • vectorized, unsure elemtwise
  • vector vs vector of the same length allowed? -> yes but warnings

• a function vectorised with respect to all arguments: No see na.rm=TRUE

• a function vectorised with respect to the first argument but not the second: plenty of cases

9.0.3 Putting outputs into context

we ought to generate outputs of a predictable kind

IE: return same kind of object, do one task

9.0.4 Exercise 9.8

No, but unsure

fred <- list(happy = 1:10, name = "squash")
rep(fred, 5)
seq(NULL)
x <- 1:10
sample(x[x > 10])

9.0.5 Using invisible()

return something but not printed!