R - Multiply a Dataframe by Another dataframe

Question

I have 2 data frames df1 and df2. df1 and df2 have the same size (rows and columns) and same factors. Say:

df1 <- data.frame(a=c('alpha','beta','gamma'), b=c(1,2,3), c=c('x','y','z'), d=c(4,5,6))

      a b c d
1 alpha 1 x 4
2  beta 2 y 5
3 gamma 3 z 6

and

df2 <- data.frame(a=c('alpha','beta','gamma'), b=c(7,8,9), c=c('x','y','z'), d=c(10,11,12))

      a b c  d
1 alpha 7 x 10
2  beta 8 y 11
3 gamma 9 z 12

I would like to multiply these 2 dataframes and get a result like tyhis:

      a b  c d
1 alpha 7  x 40
2  beta 16 y 55
3 gamma 27 z 72

I have done some search and attempted the following code:

M <- merge(df1,df2,by=c('a','c'))
S <- M[,grepl("*\\.x$",names(M))] * M[,grepl("*\\.y$",names(M))]
cbind(M[,c('a','c'),drop=FALSE],S)

this code works fine and gives the following:

      a c b.x d.x
1 alpha x   7  40
2  beta y  16  55
3 gamma z  27  72

Question: Is there a better way to achieve this multiplication ? Keep in mind that my dataframes have same number of rows, columns and factor names. My real life dataframes are much larger, both rows and columns.

Answer 1

Something like this maybe?:

data.frame(
 Map(function(x,y) if(all(is.numeric(x),is.numeric(y))) x * y else x, df1, df2)
)

#      a  b c  d
#1 alpha  7 x 40
#2  beta 16 y 55
#3 gamma 27 z 72

Some benchmarking:

smp <- sample(1:4,50000,replace=TRUE)
df1big <- df1[,smp]
df2big <- df2[,smp]

lmfun <- function() {
 out <- data.frame(
 Map(function(x,y) if(all(is.numeric(x),is.numeric(y))) x * y else x,
     df1big, df2big)
)
}
johnfun <- function() {
  sel <- sapply(df1big, is.numeric)
  df1big[,sel] <- df1big[,sel] * df2big[,sel]
}

system.time(lmfun())
#   user  system elapsed 
#   6.06    0.00    6.07 
system.time(johnfun())
#   user  system elapsed 
#  24.91    0.00   24.99

Answer 2

Assuming the columns in each DF match, you could simply select the numeric ones and then multiply them. This method minimizes the amount of non-vectorized R as much as possible.

sel <- sapply(df1, is.numeric)
df1[,sel] <- df1[,sel] * df2[,sel]

You could make a copy of df1 first so that you keep that.

If you have potential unmatched numeric columns it's relatively easy to adjust it.

sel <- sapply(df1, is.numeric) & sapply(df2, is.numeric)
df1[,sel] <- df1[,sel] * df2[,sel]