Complex algorithm in R with data.tables using previous rows values

Question

I have my data in the form of a data.table given below

structure(list(atp = c(1, 0, 1, 0, 0, 1), len = c(2, NA, 3, NA, 
NA, 1), inv = c(593, 823, 668, 640, 5         


        

Answer 1

Adding the desired outcome to your example would be very helpful, as I'm having trouble following the if/then logic. But I took a stab at it anyway:

library(data.table)

# Example data:
dt <- structure(list(atp = c(1, 0, 1, 0, 0, 1), len = c(2, NA, 3, NA, NA, 1), inv = c(593, 823, 668, 640, 593, 745), GU = c(36, 94, 57, 105, 48, 67), RUTL = c(100, NA, 173, NA, NA, 7)), .Names = c("atp", "len", "inv", "GU", "RUTL"), row.names = c(NA, -6L), class = c("data.table", "data.frame"), .internal.selfref = "")

# Add a row number:
dt[,rn:=.I]

# Use this function to get the value from a previous (shiftLen is negative) or future (shiftLen is positive) row:
rowShift <- function(x, shiftLen = 1L) {
  r <- (1L + shiftLen):(length(x) + shiftLen)
  r[r<1] <- NA
  return(x[r])
}

# My attempt to follow the seemingly circular if/then rules:
lostsales2 <- function(transit) {
  # If atp==1, set csi_begin to inv and csi_end to csi_begin - GU:
  transit[atp==1, `:=`(csi_begin=inv, csi_end=inv-GU)]

  # Set csi_order to the value of csi_begin from two rows prior:
  transit[, csi_order:=rowShift(csi_begin,-2)]

  # Set csi_order to 0 if csi_begin from two rows prior was NA
  transit[is.na(csi_order), csi_order:=0]

  # Initialize IRQ to 0
  transit[, IRQ:=0]

  # If ATP==1, set IRQ to csi_order - RUTL
  transit[atp==1, IRQ:=csi_order-RUTL]

  # If ATP!=1, set csi_begin to inv + IRQ value from previous row, and csi_end to csi_begin - GU
  transit[atp!=1, `:=`(csi_begin=inv+rowShift(IRQ,-1), csi_end=inv+rowShift(IRQ,-1)-GU)]
  return(transit)
}

lostsales2(dt)
##    atp len inv  GU RUTL rn csi_begin csi_end csi_order  IRQ
## 1:   1   2 593  36  100  1       593     557         0 -100
## 2:   0  NA 823  94   NA  2        NA      NA         0    0
## 3:   1   3 668  57  173  3       668     611       593  420
## 4:   0  NA 640 105   NA  4       640     535         0    0
## 5:   0  NA 593  48   NA  5       593     545       668    0
## 6:   1   1 745  67    7  6       745     678       640  633

Is this output close to what you were expecting?