CODEWITHSUNDEEP
r
user6335904
user6335904

Reputation:

Why is system.time only returning NA?

I'm working on another class project where I have to measure system.time for iterations of a loop to compare a function to cbind.

I have most of it working, but right now, the system.time part is only returning NA.

# Set memory size so it'll stop yelling at me
memory.size(max=FALSE)

# Set matrixes
A <- matrix(runif(n * n), n, n)
B <- matrix(runif(n * n), n, n)

# Set df
df <- data.frame(size=numeric(0),fast=numeric(0),slow=numeric(0))

# Set n
n <- c(200,400,600,800,1000,1200,1400,1600,1800,2000)

# Make a silly cbind-esque function
myCbind <- function(x,y) {
  flatA <- array(x)
  flatB <- array(y)
  comboAB <- data.frame(flatA,flatB)
  return(comboAB)
}

# Loop
for (i in n) {
  cbind(A,B)
  times <- system.time(cbind(A,B))
  fast_time = as.vector(times[n])
  myCbind(A,B)
  times <- system.time(myCbind(A,B))
  slow_time = as.vector(times[n])
  df <- rbind(df, data.frame(size=n, fast=fast_time, slow=slow_time))
}

# Print df to make sure it's working
df

And this is what it prints out (truncated for neat demonstration):

    size fast slow
1    200   NA   NA
2    400   NA   NA
3    600   NA   NA
4    800   NA   NA
5   1000   NA   NA
.      .    .    .
.      .    .    .
.      .    .    .

Any ideas?

Upvotes: 2

Views: 161

Answers (1)

Zheyuan Li
Zheyuan Li

Reputation: 73315

Change 

as.vector(times[n])

to

as.vector(times[3])

then everything works. system.time() returns a vector of length 5, the 3rd element of which is wall clock time. However, you asked for nth element while n is always greater than 5.

Try the following, then you will learn what system.time() returns:

x = system.time(runif(1000000, 0, 1)
x[1:10]

Update:

You know, actually, your code is far from being OK. Even after the correction for system.time(), your loop still looks really suspicious (see my comments):

for (i in n) {
  ## ? where is A, B ???
  cbind(A,B)  ## why print this out ???
  times <- system.time(cbind(A,B));
  fast_time = as.vector(times[3])
  myCbind(A,B)  ## why print this out ???
  times <- system.time(myCbind(A,B))
  slow_time = as.vector(times[3])
  ## size = n, or size = i ???
  df <- rbind(df, data.frame(size = n, fast = fast_time, slow = slow_time))
  }

So, putting it all together, this should be the right one:

n <- c(200,400,600,800,1000,1200,1400,1600,1800,2000)
df <- data.frame(size=numeric(0),fast=numeric(0),slow=numeric(0))

myCbind <- function(x,y) {
  flatA <- array(x)
  flatB <- array(y)
  comboAB <- data.frame(flatA,flatB)
  return(comboAB)
  }

for (i in n) {
  A <- matrix(runif(i * i), i, i)
  B <- matrix(runif(i * i), i, i)
  fast_time <- as.vector(system.time(cbind(A,B)))[3]
  slow_time <- as.vector(system.time(myCbind(A,B)))[3]
  df <- rbind(df, data.frame(size = i, fast = fast_time, slow = slow_time))
  }

> df
   size  fast  slow
1   200 0.000 0.001
2   400 0.003 0.003
3   600 0.007 0.007
4   800 0.013 0.013
5  1000 0.022 0.023
6  1200 0.032 0.032
7  1400 0.044 0.045
8  1600 0.059 0.060
9  1800 0.076 0.077
10 2000 0.094 0.094

Unfortunately, your version myCbind is no better than cbind!

Upvotes: 2

Related Questions