I'm having trouble splitting and unsplitting a DataFrame, using the methods defined in IRanges. Here's an attempt at a minimal reprex.

library(IRanges)
df <- DataFrame(
    a = seq_len(4L),
    b = as.factor(rep(c("b", "a"), each = 2L)),
    row.names = LETTERS[seq_len(4L)]
)
print(df)

DataFrame with 4 rows and 2 columns
          a        b
  <integer> <factor>
A         1        b
B         2        b
C         3        a
D         4        a

split <- split(x = df, f = df[["b"]])
print(split)

SplitDataFrameList of length 2
$a
DataFrame with 2 rows and 2 columns
          a        b
  <integer> <factor>
C         3        a
D         4        a

$b
DataFrame with 2 rows and 2 columns
          a        b
  <integer> <factor>
A         1        b
B         2        b

This is all good and lets me manipulate the DataFrame by a grouping factor, similar to the approach in dplyr with group_by. However, I'm having trouble coercing the split back to a standard DataFrame via unsplit().

unlist() will coerce back to DataFrame but flips the row names, because we're not keeping track of our factor grouping:

unlist(split, use.names = FALSE)

DataFrame with 4 rows and 2 columns
          a        b
  <integer> <factor>
C         3        a
D         4        a
A         1        b
B         2        b

Neither one of these approaches with unsplit() seems to work:

unsplit(split, f = df[["b"]])

## Error in unsplit(split, f = df[["b"]]) : 
##   Length of 'unlist(value)' must equal length of 'f'

unsplit(split, f = split[, "b"])

## Error in `splitAsList<-`(`*tmp*`, f, drop = drop, value = value) : 
##   Length of 'value' must equal the length of a split on 'f'

See related S4 method definition:

getMethod(
    f = "unsplit",
    signature = "List",
    where = asNamespace("IRanges")
)

Thanks, fixed in version 2.18.2, to appear.