stat_contour and geom_contour throwing errors when attempting to use z axis

[lazarillo]

I also submitted this via StackOverflow, since I wasn't 100% sure if the problem were user error or an actual problem with the code.

---

Successful Run

It's possible I do not understand the expectations, however, if I follow the example at stat_contour's documentation, I successfully get:

# Generate data
library(reshape2) # for melt
volcano3d <- melt(volcano)
names(volcano3d) <- c("x", "y", "z")

# Basic plot
v <- ggplot(volcano3d, aes(x, y, z = z))
v + stat_contour()

leading to:

Unsuccessful Run

However, when I try to do what appears to be _exactly the same thing with exactly the same data types_, I have errors:

my_plt <- ggplot(diamonds, aes(depth, carat, z=price))
my_plt + stat_contour()

yields:

Warning message:
Computation failed in `stat_contour()`:
(list) object cannot be coerced to type 'double' 

Debugging Attempt

The only potential difference I could imagine, especially with the error I was receiving, was an object type difference. So, here are some details there:

• volcano3d

  head(volcano3d)
  
    x y   z
  1 1 1 100
  2 2 1 101
  3 3 1 102
  4 4 1 103
  5 5 1 104
  6 6 1 105
  

  And...

  volcano3d %>% sapply(class)
       x         y         z 
  "integer" "integer" "numeric" 
  

• diamonds

  diamonds %>% select(depth, carat, price) %>% head
  Source: local data frame [6 x 3]
  
    depth carat price
    (dbl) (dbl) (dbl)
  1  61.5  0.23   326
  2  59.8  0.21   326
  3  56.9  0.23   327
  4  62.4  0.29   334
  5  63.3  0.31   335
  6  62.8  0.24   336
  

  And...

  diamonds %>% select(depth, carat, price) %>% sapply(class)
      depth     carat     price 
  "numeric" "numeric" "numeric" 
  

I cannot see any reason why I should be getting these errors. Any ideas??

Thanks!

[hadley]

The problem is that you need to provide a surface - i.e. for each value of x and y you should only have one z value. I've tweaked the docs and added an message to detect this case, but I suspect you're looking for a different geom (or need to do some preprocessing first).

[lazarillo]

Understood, thanks!

It wasn't really much of a problem for me, because the geom_raster(), geom_tile(), and geom_rect() functions are so flexible & powerful that I could just use them to accomplish the nearly same thing, and then effectively make a "filled contour", like your last example on the docs:

v + geom_raster(aes(fill = density)) + geom_contour(colour = "white").

(Except without the + geom_contour(colour = "white")... removing the geom_countour() does not impact the quality of the plot too much.)