2

I have been running into a computation time issue when working with large datasets and applying functions that use multiple columns of the data.table as inputs. Here is a simplified example:

require(data.table)
main_dt   <- data.table(Start=(1:2), End=c(2,2))
lookup_dt <- data.table(Year = 1:3, Amount = c(10,20,30))

So:

> main_dt
Start End
1:     1   2
2:     2   2

And

> lookup_dt
   Year Amount
1:    1     10
2:    2     20
3:    3     30

What I want is to add a column in data.table main_dt that includes the sum of Amounts of Years between Start and End. So something like this, but there is a referencing error.

main_dt[, Amount := lookup_dt[ Year >= start & Year <= end, sum(Amount)]]
Warning messages:
1: In Year >= Start :  longer object length is not a multiple of shorter object length
2: In Year <= End :  longer object length is not a multiple of shorter object length

If the Amount would depend only on one variable, I could do:

main_dt[, GreaterAmount := lapply(Start, function(x) return (lookup_dt[Year >= x, sum(Amount)]))]

to get:

> main_dt
Start End Amount GreaterAmount
1:     1   2     30            60
2:     2   2     30            50

Which would be ok, but

  • I need to apply a function that actually depends on both Start and End
  • Applying a function for every row individually slows down computation process rapidly. A vectorized solution would be highly appreciated.

Looking forward to any suggestions!

Thank you! Markus

Markus Ausserhofer
  • 23
  • 2

1 Answers1

0

First, the join columns should be the same class, so we can either convert main_dt$End to integer, or main_df$Start and lookup_dt$Year to numeric. I'll choose the first:

main_dt[, End := as.integer(End)]
main_dt
#    Start   End
#    <int> <int>
# 1:     1     2
# 2:     2     2

From here, we can do a joining-assignment:

main_dt[, Amount := lookup_dt[.SD, sum(Amount), on = .(Year >= Start, Year <= End), by = .EACHI]$V1 ]
main_dt
#    Start   End Amount
#    <int> <int>  <num>
# 1:     1     2     30
# 2:     2     2     20

If you're somewhat familiar with data.table, note that .SD referenced is actually the contents of main_dt, so lookup_dt[.SD,...] is effectively "main_dt left join lookup_dt". From there, the on= should be normal, and sum(Amount) is what you want to aggregate. The only new thing introduced here is the use of by=.EACHI, which can be confusing; some links for that:

r2evans
  • 141,215
  • 6
  • 77
  • 149