How to effectively visualize a recursive function?
This is how I would go about explaining recursive functions in R
:
First, I agree with @AEBilgrau that the factorial is a good example for recursion. (Better than Fibonacci in my opionion.)
Then I would quickly go through the theoretical basis why the factorial can be defined as a recursive function, something simple like
4! = 4*3*2*1 = 4*3!
Then you could present them the respective recursive R
function
fact=function(x) if (x==0) return(1) else return(x*fact(x-1))
fact(3)
#6
but present them also the following output
#|fact(3) called
#|fact(3) is calculated via 3*fact(2)
#|fact(2) is unknown yet. Therefore calling fact(2) now
#|Waiting for result from fact(2)
#| fact(2) called
#| fact(2) is calculated via 2*fact(1)
#| fact(1) is unknown yet. Therefore calling fact(1) now
#| Waiting for result from fact(1)
#| | fact(1) called
#| | fact(1) is calculated via 1*fact(0)
#| | fact(0) is unknown yet. Therefore calling fact(0) now
#| | Waiting for result from fact(0)
#| | | fact(0) called
#| | | fact(0)=1 per definition. Nothing to calculate.
#| | | fact(0) returning 1 to waiting fact(1)
#| | fact(1) received 1 from fact(0)
#| | fact(1) can now calculate 1*fact(0)=1*1=1
#| | fact(1) returning 1 to waiting fact(2)
#| fact(2) received 1 from fact(1)
#| fact(2) can now calculate 2*fact(1)=2*1=2
#|fact(3) received 2 from fact(2)
#|fact(3) can now calculate 3*fact(2)=3*2=6
#[1] 6
as derived from
#helper function for formatting
tabs=function(n) paste0("|",rep("\t",n),collapse="")
fact=function(x) {
#determine length of call stack
sfl=length(sys.frames())-1
#we need to define tmp and tmp1 here because they are used in on.exit
tmp=NULL
tmp1=NULL
#on.exit will print the returned function value when we exit the function ...
#... i.e., when one function call is removed from the stack
on.exit({
if (sfl>1) {
cat(tabs(sfl),"fact(",x,") returning ",
tmp," to waiting fact(",x+1,")\n",sep="")
}
})
cat(tabs(sfl),"fact(",x,") called\n",sep="")
if (x==0) {
cat(tabs(sfl),"fact(0)=1 per definition. Nothing to calculate.\n",sep="")
#set tmp for printing in on.exit
tmp=1
return(1)
} else {
#print some info for students
cat(tabs(sfl),"fact(",x,") is calculated via ",x,"*fact(",x-1,")\n",sep="")
cat(tabs(sfl),"fact(",x-1,
") is unknown yet. Therefore calling fact(",x-1,") now\n",sep="")
cat(tabs(sfl),"Waiting for result from fact(",x-1,")\n",sep="")
#call fact again
tmp1=fact(x-1)
#more info for students
cat(tabs(sfl),"fact(",x,") received ",tmp1," from fact(",x-1,")\n",sep="")
tmp=x*tmp1
cat(tabs(sfl),"fact(",x,") can now calculate ",
x,"*fact(",x-1,")=",x,"*",tmp1,"=",tmp,"\n",sep="")
return(tmp)
}
}
fact(3)
Here's my example, probably used in quite a few textbooks:
recursive_sum <- function(n){
if(n == 1) {print("Remember 1, add everything together"); return(n)}
print(paste0("Remember ", n, ", pass ", n-1, " to recursive function"))
n + recursive_sum(n-1)
}
Output:
> recursive_sum(4)
[1] "Remember 4, pass 3 to recursive function"
[1] "Remember 3, pass 2 to recursive function"
[1] "Remember 2, pass 1 to recursive function"
[1] "Remember 1, add everything together"
[1] 10
I think the factorial function is a good example for recursion. Combining this with a printout (as others suggest) seem like a good way to describe what is going on:
factorial <- function(n) {
cat("factorial(", n, ") was called.\n", sep = "")
if (n == 0) {
return(1)
} else {
return(n * factorial(n - 1))
}
}
factorial(4)
#factorial(4) was called.
#factorial(3) was called.
#factorial(2) was called.
#factorial(1) was called.
#factorial(0) was called.
#[1] 24
You can also then implement a non-recursive factorial function and compare the computational efficiencies. Or maybe ask them what is problematic with the above implementation (e.g what happens with factorial(-4)
).
Regarding a more proper visualization (and not just easy examples), there are websites which illustrate the recursion tree.
Edit: Googling recursion is also a useful lesson.