[R] (almost) rolling function or fill?

[Andras Farkas]

Hello,
please see if you have a thought on how to achieve the following:
we have:
 df<-data.frame(a=Sys.Date()+1:10,               b=Sys.Date()+c(NA,NA,NA,rep(3,4),NA,NA,3),               c=Sys.Date()+c(NA,NA,NA,rep(9,4),NA,NA,9))


the idea I have difficulty wrapping my head around is to do the following: I need the system to look at df$a by row (lets call it the index row) and look at df$b and df$c 1 row before the given row in df$a  (lets call it index row -1) and evaluate if the index row value in df$a falls into the range (>= and <=) of the index row -1 values in df$b and df$c. If it does, then copy over the index row -1 values in df$b and df$c into the index row in df$b and df$c, if not place an NA in both cells of the index row in df$b and df$c. 
 examples:
1. the date value in df$a[8] is between df$b[7] and df$c[7] so we can copy the values in df$b[7] and df$c[7] into df$b[8] and df$c[8]2.  the date value in df$a[9] is between df$b[8] and df$c[8] (as we copied it in in step 1)  so we can copy the values in df$b[8] and df$c[8] into df$b[9] and df$c[9]3.  the date value in df$a[10] is NOT between df$b[9] and df$c[9] (as we copied it in in step 2)  so we can place NA in df$b[10] and df$c[10] 

also would like to do this going up, too, similar to fill(...,"downup"). On the end we would want to have this:
 dfwanted<-data.frame(a=Sys.Date()+1:10,               b=Sys.Date()+c(NA,NA,rep(3,7),NA),               c=Sys.Date()+c(NA,NA,rep(9,7),NA))


much appreciate any help you could provide.
thanks,

Andras 
	[[alternative HTML version deleted]]