substitute, eval, bquote, do.call ... some guidance for substituting expressions

Question

I would like to program a time series class. The idea is that I instantiate an object with an expression and some other time series objects, for instance

(two time series)
x <- ts(rnorm(10), frequency = 4, start = c(1959, 2))
y <- ts(rnorm(10), frequency = 4, start = c(1959, 2))

(a time series, defined to be the sum of x and y)
z <- exprTs("x+y", parents=list(x=x, y=y))

(get some part of the series)
window(z, start=1960, end=1960.75)

The problem is, how can I evaluate the expression? I tried the following:

(constructor for class)
exprTs <- function(expr, parents) {
res = list(expr=expr, parents=parents)
class(res) <- "exprTs"
res
}

(window method)
window.exprTs <- function(z, ...) {
eval(substitute(z$expr, lapply(z$parents, window, ...)))
#do.call(z$expr, lapply(z$parents, window, ...))
}

I can not get the window method to work.

If you could guide me to how to use substitute, eval, do.call appropriately, that would be very helpful.

Answer 1

One problem is that you are specifying the expression as a string so it will be evaluated into a string. If you want to parse a string into an expression, you need to use the parse command:

> "x+y"
[1] "x+y"
> parse(text="x+y")
expression(x+y)
attr(,"srcfile")
<text> 

But yeah, why not use apply and functions instead?

Answer 2

You may be thinking about this in the wrong way and just over-complicate things. There is already an addition defined for ts objects:

R> set.seed(42)
R> x <- ts(rnorm(10), frequency = 4, start = c(1959, 2))
R> y <- ts(rnorm(10), frequency = 4, start = c(1959, 2))
R> z <- x + y
R> cbind(x,y,z)
               x       y       z
1959 Q2  1.37096  1.3049  2.6758
1959 Q3 -0.56470  2.2866  1.7219
1959 Q4  0.36313 -1.3889 -1.0257
1960 Q1  0.63286 -0.2788  0.3541
1960 Q2  0.40427 -0.1333  0.2709
1960 Q3 -0.10612  0.6360  0.5298
1960 Q4  1.51152 -0.2843  1.2273
1961 Q1 -0.09466 -2.6565 -2.7511
1961 Q2  2.01842 -2.4405 -0.4220
1961 Q3 -0.06271  1.3201  1.2574
R> 

You don't really need an expression parser to operate on R objects.

Answer 3

Now I found a solution:

window.exprTs <- function(z, ...) {
names(z$parents)<-NULL
do.call(z$expr, lapply(z$parents, window, ...))
}

plus <- function(x, ...) if (nargs() == 1) x else x + Recall(...)
z <- exprTs(plus, parents=list(x=x, y=y))

Now
window(z, start=1960, end=1960.75)
gives the desired result.