Closes #2930 -- bugfix to as.matrix.data.table() #2938
Conversation
Merge branch 'master' into as_matrix_rownames # Conflicts: # NEWS.md # inst/tests/tests.Rraw
(hopefully no typos; on browser)
Rolling branch forward to master to apply bugfixes
Passing a column to use as the rownames to as.matrix.data.table() now works when the input data.table has a single row.
|
The branch i've submitted the pull request from is the same branch which I used to submit the original PR #2702 - the final commit in the history is the one pertinent to this PR. Let me know if I should create a new branch to submit this PR from instead. |
R/data.table.R
Outdated
| # rownames argument is a vector of row names, no column in x to drop. | ||
| rn <- rownames | ||
| rnc <- NULL | ||
| } else if (length(rownames) == nrow(x) && nrow(x) == 1 && is.character(rownames) && !(rownames %in% colnames(x))) { |
There was a problem hiding this comment.
I think an extra argument for the length-1 case is going to be necessary. Something like rownames.literal = TRUE.
as.matrix(data.table(x = 1, z = 2), "z")
#> x z
#> z 1 2
as.matrix(data.table(x = 1, z = 2), "z", rownames.literal = FALSE)
#> x z
#> 2 1 2There was a problem hiding this comment.
Maybe. I'm not sure having an extra argument is really necessary, since it will only apply to data.tables with 1 row, unless we want to allow vector recycling for the rownames when the data.table has > 1 rows.
There was a problem hiding this comment.
But this PR arose because of confusion about 1-row data.tables. And there may be other cases where the overloaded argument of rownames might cause problems.
rownames.literal=TRUE allows the user to specify a vector of rownames (of length 1) when the input data.table has only 1 row.
Argument added to method section of man page
| @@ -1887,14 +1887,19 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) { | |||
| # x | |||
| #} | |||
|
|
|||
| as.matrix.data.table <- function(x, rownames, ...) { | |||
| as.matrix.data.table <- function(x, rownames, rownames.literal=FALSE, ...) { | |||
There was a problem hiding this comment.
In such unwanted edge case we don't have to be consistent to data.frame by default, it is important to give option for consistency to data.frame, which new arg already provides, but we could tune its default.
Maybe new arg could have smart default like nrow(x)==1L? probably more correct would be nrow(x)==1L && !missing(rownames) && is.character(rownames) but it is a little bit messy, maybe nrow(x)==1L is sufficient in this case.
| # rownames argument is a vector of row names, no column in x to drop. | ||
| rn <- rownames | ||
| rnc <- NULL | ||
| } else if (nrow(x) == 1 && isTRUE(rownames.literal)) { |
There was a problem hiding this comment.
don't we need length(rownames) == 1L here?
| @@ -17,6 +17,11 @@ the \code{rownames} in the returned \code{matrix}. If \code{TRUE} the | |||
| single column, otherwise the first column in the \code{data.table} will | |||
| be used. Alternative a vector of length \code{nrow(x)} to assign as the | |||
| row names of the returned \code{matrix}.} | |||
| \item{rownames.literal}{optional, when x is a \code{data.table} that has | |||
There was a problem hiding this comment.
before word "optional" I would put "logical" to clearly state argument type in front, so {logical, optional, when ....
x alone should be wrapped in \code{x}
... when x is ...
to
... when \code{x} is ...
|
there are definitely too many commits in this branch, |
|
Thanks @jangorecki , i tried the above series of commands but git wouldnt let me push the changes. I've created a new branch instead containing just the bugfix and submitted a new PR #2939 |
Supplying a column to use as the rownames in as.matrix.data.table() now also works when the input data.table has only one row. Previously the rownames argument was treated as a vector of user supplied rownames because the length of rownames was the same as the number of rows in x.
The user can now still supply there own length 1 vector of rownames to use when the input data.table has one row, provided it is not a column in the input data.table.
This makes the case when the data.table has one row consistent with the functionality where the data.table has > 1 rows.