R help - Aug 2014 - dynamic runSum

Hello,
runSum calculates a running sum looking back a fixed distance n, e.g. 20.  
How do I calculate a dynamic runSum function for an xts object?
In 
otherwords, I want to calculate a running sum at each point in time 
looking back a variable distance. In this example, values governed by 
the vector VL.  
Here's a minimum reproducible example:
 
 library(quantstrat)
symbols = c('^GSPC') 
 
 start.date <- as.Date("2010-01-01")
 end.date <- as.Date("2013-12-31")
 
 getSymbols(symbols, from=as.character(start.date),
to=as.character(end.date),adjust=T)
 
 "acF1" <- function(x, n1=5, n2=10, n3=20, nacF1=25, n0=20, ...) {
    var1 <- x - lag(x,1,na.pad=T)
    var2 <- runSD(x, n=n1, sample=TRUE, cumulative=FALSE) 
    var3 <- runMean(var2, n=n2, cumulative=FALSE) 
    VL <- ifelse( trunc(n3/(var2/var3))>nacF1, nacF1,
trunc(n3/(var2/var3)))
    p_pos <- ifelse(var1>=0, var1, 0)
    out1 <- runSum(p_pos,  n=n0, cumulative=FALSE)
 
    res <- cbind(var1, var2, var3, VL, p_pos, out1)
    colnames(res) <-
c("var1","var2","var3","VL",
"p_pos", "out1")
 
    reclass(res)
 }
 
 
 acf1 <- acF1( GSPC[,c("GSPC.Close")], n1=5, n2=10, n3=20,
nacF1=25, n0=20)
 acf1
 
 
So on 
2010-02-02, I want runSum to be looking back 23 points as governed by VL , not
20 points
2010-02-03, I want runSum to be looking back 24 points as governed by VL,  not
20 points
 etc etc 
 2013-12-31, I want runSum to be looking back 25 points as governed by VL, not
20 points
 
 
Amarjit

On Thu, Aug 7, 2014 at 9:32 AM, amarjit chandhial
<a.chandhial at btinternet.com> wrote:
> Hello,
> runSum calculates a running sum looking back a fixed distance n, e.g. 20.
> How do I calculate a dynamic runSum function for an xts object?
> In
> otherwords, I want to calculate a running sum at each point in time
> looking back a variable distance. In this example, values governed by
> the vector VL.
The width argument in rollapplyr in the zoo package can be a vector.
It can't be NA though so we have used 1 in those cases here and at the
end used na.omit to get rid of the junk at the beginning:

na.omit(rollapplyr(acf1, ifelse(is.na(acf1$VL), 1, acf1$VL), sum))