cbind fails if input is a 0-column data frame.

[kzenstratus]

Credit goes to @austin3dickey for diagnosing this and @hcui10 for digging further.

When one of the arguments is a 0-column data frame, the overloaded function, cbind() fails with
Error in data.table::data.table(...) : Item 2 has no length. Provide at least one item (such as NA, NA_integer_ etc) to be repeated to match the 5 rows in the longest column. Or, all columns can be 0 length, for insert()ing rows into.

Reproducible Example

> dt1 <- data.table::data.table(a = 1:5, b = 6:10)
> df1 <- as.data.frame(dt)
> df2 <- data.frame()[1:5, ]
> dt1
   a  b
1: 1  6
2: 2  7
3: 3  8
4: 4  9
5: 5 10
> df1
  a  b
1 1  6
2 2  7
3 3  8
4 4  9
5 5 10
> df2
data frame with 0 columns and 5 rows
> cbind(df1, df2)
     a  b
NA   1  6
NA.1 2  7
NA.2 3  8
NA.3 4  9
NA.4 5 10
> cbind(dt1, df2)

Error in data.table::data.table(...) : 
  Item 2 has no length. Provide at least one item (such as NA, NA_integer_ etc) to be repeated to match the 5 rows in the longest column. Or, all columns can be 0 length, for insert()ing rows into.

Root Cause Diagnosis

This seems to occur when cbind.data.frame converts a data.frame into a data.table.

# line 29-33 in data.table::data.table
else if (is.matrix(xi) || is.data.frame(xi)) {
            xi = as.data.table(xi, keep.rownames = keep.rownames)
            x[[i]] = xi
            numcols[i] = length(xi)
        }

This causes the number of rows to change.

> df <- data.frame()[1:5, ]
data frame with 0 columns and 5 rows
> data.table::data.table(df)
Empty data.table (0 rows) of 1 col: df

Which fails below since the number of rows is now 0.

# line 72-75 in data.table::data.table
if (nrows[i] == 0L) 
            stop("Item ", i, " has no length. Provide at least one item (such as NA, NA_integer_ etc) to be repeated to match the ", 
                nr, " row", if (nr > 1L) 
                  "s", " in the longest column. Or, all columns can be 0 length, for insert()ing rows into.")

Original Point of Failure

This was originally discovered when attempting to run the following.

library(ggfortify)
library(ggplot2)
# Works!
autoplot(stats::prcomp(iris[, c(1,2,3,4)]), data = iris, colour = 'Species')

# Doesn't work
iris <- data.table::as.data.table(iris)
autoplot(stats::prcomp(iris[, c(1,2,3,4)]), data = iris, colour = 'Species')

# Works Again!
iris <- as.data.frame(iris)
> autoplot(stats::prcomp(iris[, c(1,2,3,4)]), data = iris, colour = 'Species')

[jangorecki]

This could be simplified to

library(data.table)
data.table(data.table(a=1), data.table())
#Error in data.table::data.table(...) : 
#  Item 2 has no length. Provide at least one item (such as NA, NA_integer_ etc) to be repeated to match the 1 row in the longest column. Or, all columns can be 0 length, for insert()ing rows into.

in data.table we do not allow to have non-0 rows for 0 columns, but of course such usage could be handled nicely.