Using xyplot within custom function with panel function in R - r

Question

I'm trying to generate a custom xyplot in the lattice graphics package. It is a plot of two lines with range limits plotted.
# Raw dataframe in long format
df <- data.frame(
Response = c(runif(100), rnorm(100)),
Trial = c(rep("A", 100), rep("C", 100)),
Year = 1:10,
Rep = rep(1:10, each = 10))
# Aggregate the data (take mean/min/max across "Rep" variable
gdf <- do.call(data.frame, aggregate(Response ~ Year + Trial, data = df,
FUN = function(x) c(avg = mean(x), mini = min(x), maxi = max(x))))
# Plot using xyplot (without making this a function)
my.panel.bands <- function(x, y, upper, lower, fill, col,
subscripts, ..., font, fontface){
upper <- upper[subscripts]
lower <- lower[subscripts]
panel.polygon(c(x, rev(x)), c(upper, rev(lower)),
col = fill, alpha = 0.2, border = FALSE, ...)
}
f1 <- formula(Response.avg ~ Year)
p <- xyplot(x = f1, data = gdf, groups = Trial,
col=c("red", "blue"), pch = 16,
scales = list(x = list(rot = 45)),
xlab = 'Year', ylab = 'Response',
layout = c(1, 1),
ylim = c(min(gdf$Response.mini), max(gdf$Response.maxi)),
upper = gdf$Response.maxi,
lower = gdf$Response.mini,
panel = function(x, y, ...){
panel.superpose(x, y, panel.groups = my.panel.bands,
type = 'l', fill = c("red", "blue"),...)
panel.xyplot(x, y, type = 'b', cex = 0.6, lty = 1, ...)
}
)
png("panel_plot.png")
print(p)
dev.off()
However, if I try to make a custom function out of this xyplot command then I get a very different plot to what I was expecting. I'm assuming I'm doing something incorrect with passing the grouping variable or in using the panel function.
panel_plot <- function(f, df, grouper, xlabel, ylabel,
ylim, upper_border, lower_border, mfcol){
p <- xyplot(x = f, data = df, groups = eval(grouper),
col = c("red", "blue"), pch = 16,
scales = list(x = list(rot = 45)),
xlab = xlabel, ylab = ylabel,
ylim = ylim,
layout = mfcol,
upper = upper_border,
lower = lower_border,
panel = function(x, y, ...){
panel.superpose(x, y, panel.groups = my.panel.bands,
type = 'l', fill = c("red", "blue"),...)
panel.xyplot(x, y, type = 'b', cex = 0.6, lty = 1, ...)
}
)
return(p)
}
f1 <- formula(Response.avg ~ Year)
p <- panel_plot(f1, gdf, grouper = Trial,
xlabel = "Year", ylabel = "Response",
ylim = c(min(gdf$Response.mini),max(gdf$Response.maxi)),
upper_border = gdf$Response.maxi,
lower_border = gdf$Response.mini,
mfcol = c(1, 1))
png("panel_plot_asfunction.png")
print(p)
dev.off()
Finally, if I pass the name of the variable to the group argument as a string and modify the panel_plot() function to redefine a new variable in the data.frame, then it works as expected but this seems like a strange way to do things.
panel_plot <- function(f, df, grouper, xlabel, ylabel,
ylim, upper_border, lower_border, mfcol){
df$grouper <- df[, grouper]
p <- xyplot(x = f, data = df, groups = grouper,
...
p <- panel_plot(f1, gdf, grouper = "Trial",
...
How do I define the panel_plot function so that I don't have to create a dummy column and can pass the variable name (Trial) to this function so that it is not passed as a string?
I've tried using the suggestion here but using eval() on the variable name provided the unexpected figure above.

Answer 1

Actually, consider combining both solutions of your linked SO post using match.call() list and eval(call(), ...). By themselves alone neither worked on my end.
panel_plot <- function(f, df, grouper, xlabel, ylabel,
ylim, upper_border, lower_border, mfcol){
ll <- as.list(match.call(expand.dots = FALSE)[-1])
my_panel <- function(x, y, ...){
panel.superpose(x, y, panel.groups = my.panel.bands,
type = 'l', fill = c("red", "blue"),...)
panel.xyplot(x, y, type = 'b', cex = 0.6, lty = 1, ...)
}
p <- eval(call("xyplot",
x = ll$f,
data = ll$df,
groups = ll$grouper,
xlab = ll$xlabel, ylab = ll$ylabel,
ylim = ll$ylim,
layout = ll$mfcol,
upper = ll$upper_border,
lower = ll$lower_border,
panel = my_panel
)
)
return(p)
}
f1 <- formula(Response.avg ~ Year)
p <- panel_plot(f1, gdf, grouper = Trial,
xlabel = "Year", ylabel = "Response",
ylim = c(min(gdf$Response.mini), max(gdf$Response.maxi)),
upper_border = gdf$Response.maxi,
lower_border = gdf$Response.mini,
mfcol = c(1, 1))
p
However, aesthetics are not exactly the same such as red and blue lines and points. Possibly this is because eval(call(...)) requires all arguments to be specified? Try adjusting arguments or my_panel.
Alternatively, use your dummy column but still pass unquoted name and evaluate column name with eval. Here, aesthetics render as desired non-function version.
panel_plot <- function(f, df, grouper, xlabel, ylabel,
ylim, upper_border, lower_border, mfcol){
df$grouper <- eval(as.name(deparse(substitute(grouper))), df, .GlobalEnv)
p <- xyplot(x = f, data = df, groups = grouper,
...
}