Closes #1446. data.table() gains stringsAsFactors argument.

R/data.table.R

@@ -206,7 +206,7 @@ null.data.table <-function() {
     alloc.col(ans)
 }
 
-data.table <-function(..., keep.rownames=FALSE, check.names=FALSE, key=NULL)
+data.table <-function(..., keep.rownames=FALSE, check.names=FALSE, key=NULL, stringsAsFactors=FALSE)
 {
     # NOTE: It may be faster in some circumstances to create a data.table by creating a list l first, and then setattr(l,"class",c("data.table","data.frame")) at the expense of checking.
     # TO DO: rewrite data.table(), one of the oldest functions here. Many people use data.table() to convert data.frame rather than
@@ -346,6 +346,8 @@ data.table <-function(..., keep.rownames=FALSE, check.names=FALSE, key=NULL)
            && !any(duplicated(names(value)[names(value) %in% ckey])))
            setattr(value, "sorted", ckey)
     }
+    # FR #1446, setfactor is an internal function in fread.R
+    if (isTRUE(stringsAsFactors)) setfactor(value, which(vapply(value, is.character, TRUE)), FALSE)
     alloc.col(value)  # returns a NAMED==0 object, unlike data.frame()
 }
 
@@ -2520,6 +2522,12 @@ gsd <- function(x, na.rm=FALSE) .Call(Cgsd, x, na.rm)
 gstart <- function(o, f, l, rows) .Call(Cgstart, o, f, l, rows)
 gend <- function() .Call(Cgend)
 
+# rowwise summary functions
+rowmeans <- function(x, na.rm=FALSE) .Call("Crowmeans", x, na.rm)
+rowsums <- function(x, na.rm=FALSE) .Call("Crowsums", x, na.rm)
+rowmins <- function(x, na.rm=FALSE) .Call("Crowmins", x, na.rm)
+rowmaxs <- function(x, na.rm=FALSE) .Call("Crowmaxs", x, na.rm)
+
 isReallyReal <- function(x) {
     .Call(CisReallyReal, x)
 }

R/fread.R

@@ -104,15 +104,6 @@ fread <- function(input="",sep="auto",sep2="auto",nrows=-1L,header="auto",na.str
     if (isTRUE(as.logical(check.names))) {
         setattr(ans, 'names', make.names(names(ans), unique=TRUE))
     }
-    as_factor <- function(x) {
-        lev = forderv(x, retGrp = TRUE, na.last = NA)
-        # get levels, also take care of all sorted condition
-        if (length(lev)) lev = x[lev[attributes(lev)$starts]]
-        else lev = x[attributes(lev)$starts]
-        ans = chmatch(x, lev)
-        setattr(ans, 'levels', lev)
-        setattr(ans, 'class', 'factor')
-    }
     cols = NULL
     if (isTRUE(as.logical(stringsAsFactors)))
         cols = which(vapply(ans, is.character, TRUE))
@@ -122,11 +113,7 @@ fread <- function(input="",sep="auto",sep2="auto",nrows=-1L,header="auto",na.str
         else if (is.atomic(colClasses) && "factor" %chin% colClasses)
             cols = which(vapply(ans, is.character, TRUE))
     }
-    if (length(cols)) {
-        if (verbose) cat("Converting column(s) [", paste(names(ans)[cols], collapse = ", "), "] from 'char' to 'factor'\n", sep = "")
-        for (j in cols) 
-            set(ans, j = j, value = as_factor(.subset2(ans, j)))
-    }
+    setfactor(ans, cols, verbose)
     # FR #768
     if (!missing(col.names))
         setnames(ans, col.names) # setnames checks and errors automatically
@@ -140,3 +127,21 @@ fread <- function(input="",sep="auto",sep2="auto",nrows=-1L,header="auto",na.str
     }
     ans
 }
+
+# for internal use only. Used in `fread` and `data.table` for 'stringsAsFactors' argument
+setfactor <- function(x, cols, verbose) {
+    # simplified but faster version of `factor()` for internal use.
+    as_factor <- function(x) {
+        lev = forderv(x, retGrp = TRUE, na.last = NA)
+        # get levels, also take care of all sorted condition
+        lev = if (length(lev)) x[lev[attributes(lev)$starts]] else x[attributes(lev)$starts]
+        ans = chmatch(x, lev)
+        setattr(ans, 'levels', lev)
+        setattr(ans, 'class', 'factor')
+    }
+    if (length(cols)) {
+        if (verbose) cat("Converting column(s) [", paste(names(x)[cols], collapse = ", "), "] from 'char' to 'factor'\n", sep = "")
+        for (j in cols) set(x, j = j, value = as_factor(.subset2(x, j)))
+    }
+    invisible(x)
+}

README.md

@@ -53,6 +53,8 @@
 
   19. `dcast.data.table` now allows `drop = c(FALSE, TRUE)` and `drop = c(TRUE, FALSE)`. The former only fills all missing combinations of formula LHS, where as the latter fills only all missing combinations of formula RHS. Thanks to Ananda Mahto for [this SO post](http://stackoverflow.com/q/34830908/559784) and to Jaap for filing [#1512](https://github.com/Rdatatable/data.table/issues/1512).
 
+  20. `data.table()` function gains `stringsAsFactors` argument with default `FALSE`, [#643](https://github.com/Rdatatable/data.table/issues/643). Thanks to @Jan for reviving this issue.
+
 #### BUG FIXES
 
   1. Now compiles and runs on IBM AIX gcc. Thanks to Vinh Nguyen for investigation and testing, [#1351](https://github.com/Rdatatable/data.table/issues/1351).

man/data.table.Rd

@@ -16,7 +16,7 @@
    Tip: one of the quickest ways to learn the features is to type \code{example(data.table)} and study the output at the prompt.
 }
 \usage{
-data.table(..., keep.rownames=FALSE, check.names=FALSE, key=NULL)
+data.table(..., keep.rownames=FALSE, check.names=FALSE, key=NULL, stringsAsFactors=FALSE)
 
 \method{[}{data.table}(x, i, j, by, keyby, with = TRUE,
   nomatch = getOption("datatable.nomatch"),                   # default: NA_integer_
@@ -45,6 +45,8 @@ data.table(..., keep.rownames=FALSE, check.names=FALSE, key=NULL)
 }
   \item{key}{ Character vector of one or more column names which is passed to \code{\link{setkey}}. It may be a single comma separated string such as \code{key="x,y,z"}, or a vector of names such as \code{key=c("x","y","z")}.
 
+}
+  \item{stringsAsFactors}{Logical (default is \code{FALSE}). Convert all \code{character} columns to \code{factor}s?
 }
   \item{x}{ A \code{data.table}.