Bug in addVolumeFraction() for profiles with floating point horizon depths

[pierreroudier]

There is a bug affecting the addVolumeFraction function, which affects profile which have floating points depths:

sandSiltClay <- new("SoilProfileCollection"
    , idcol = "id"
    , depthcols = c("top", "bottom")
    , metadata = structure(list(depth_units = "cm"), .Names = "depth_units", row.names = c(NA, 
-1L), class = "data.frame")
    , horizons = structure(list(id = c("Clay", "Clay", "Clay", "Clay", "Clay", 
"Sand", "Sand", "Sand", "Sand", "Sand", "Silt", "Silt", "Silt", 
"Silt", "Silt"), name = structure(c(2L, 1L, 4L, 5L, 3L, 2L, 1L, 
4L, 5L, 3L, 2L, 1L, 4L, 5L, 3L), .Label = c("SLFs", "tSLw", "VAl", 
"VLc", "VLl"), class = "factor"), top = c(0, 18.5, 28.5, 46, 
53.5, 0, 18.5, 28.5, 46, 53.5, 0, 18.5, 28.5, 46, 53.5), bottom = c(18.5, 
28.5, 46, 53.5, 101, 18.5, 28.5, 46, 53.5, 101, 18.5, 28.5, 46, 
53.5, 101), percent = c(21.5, 21.5, 17.5, 10, 2.5, 22.5, 22.5, 
30, 55, 90, 56.9, 56, 52.5, 35, 7.5), stones = c(23L, 28L, 48L, 
60L, 63L, 23L, 28L, 48L, 60L, 63L, 23L, 28L, 48L, 60L, 63L)), .Names = c("id", 
"name", "top", "bottom", "percent", "stones"), row.names = c(6L, 
7L, 8L, 9L, 10L, 1L, 2L, 3L, 4L, 5L, 11L, 12L, 13L, 14L, 15L), class = "data.frame")
    , site = structure(list(id = c("Clay", "Sand", "Silt")), .Names = "id", row.names = c(NA, 
-3L), class = "data.frame")
    , sp = new("SpatialPoints"
    , coords = structure(c(0, 0), .Dim = 1:2, .Dimnames = list(NULL, c("coords.x1", 
"coords.x2")))
    , bbox = structure(c(0, 0, 0, 0), .Dim = c(2L, 2L), .Dimnames = list(c("coords.x1", 
"coords.x2"), c("min", "max")))
    , proj4string = new("CRS"
    , projargs = NA_character_
)
)
    , diagnostic = structure(list(), .Names = character(0), row.names = integer(0), class = "data.frame")
)

plot(sandSiltClay)
addVolumeFraction(sandSiltClay, 'stones')

The error message is:

Error in `$<-.data.frame`(`*tmp*`, "val", value = c(NA, NA, NA, NA, 1,  : 
  replacement has 185 rows, data has 180

This bug is due to the rounding of floating point horizon depths in .volume2df, specifically:

d$val <- m[1:cells]

This is due to the rounding of depths in d (line 23):

d <- expand.grid(x=(1:res), y=(1:depth))

where

Browse[1]> 1:depth
 [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18

for depth = 18.5, so nrow(d) = 180, while on the other hand, length(m) = 185.

[pierreroudier]

Following a quick chat with @dylanbeaudette - the support of floating point horizon depths might actually be a problem with other functions in the package (slab, slice, etc...).

A quick patch could enforce the horizon depths to be integers, while working on a more permanent form of support.

[pierreroudier]

Just confirming that this is also a problem for slice.

[dylanbeaudette]

Recently added truncation to integer via floor() in 7822c27.

[dylanbeaudette]

This is actually a much larger problem for all functions that expect integer depths (slice, slab, and any that build on those). I'm going to close this for now, and note for future enhancement in #106.