dcast和knitr的作用域问题

Question

I've created a simple R file called test.R :

library(knitr)

file <- "test.Rmd"

k <- function(input, output){
  knit(input, output)
}

k(file, "badtest.md")

knit(file, "goodtest.md")

k(file, "badtest2.md")

Accompanying it is a small Rmd file called test.Rmd :

```{r define, echo = FALSE, warning=FALSE} 
suppressPackageStartupMessages(library(data.table))
xsum <- function(x){
  sum(x, na.rm = TRUE)
}
```
```{r, echo = FALSE}
mt <- as.data.table(mtcars)
f <- function(d){
  dcast(d,  drat ~ cyl, fun.aggregate = list(xsum), value.var = "carb") 
}

f(mt)[1:5]

```

The goal is to run the xsum function on the data. In the first case where the knit call is inside a function ( badtest.md ), it fails with Error in eval(expr, envir, enclos): could not find function "xsum" .

The second case ( goodtest.md ) is outside of a function and it succeeds.

The third case ( badtest2.md ) is the same as the first, but it also succeeds now that the second case has run.

Why does this scoping issue occur and how can I resolve it so that I can use xsum or any custom function?

Answer 1

The solution is to pass through the function explicitly as I found out from this comment on the data.table issue page:

f <- function(d, fun.aggregate){
  dcast(d,  drat ~ cyl, fun.aggregate = fun.aggregate, value.var = "carb") 
}