Should never recycle the zero-length vector

NEWS.md

@@ -12,6 +12,37 @@
 
 3. `fwrite()` now writes UTF-8 or native csv files by specifying the `encoding=` argument, [#1770](https://github.com/Rdatatable/data.table/pull/1770). Thanks to @shrektan for the request and the PR.
 
+4. `data.table()` no longer fills empty vectors with `NA` with warning. Instead a 0-row `data.table` is returned, [#3727](https://github.com/Rdatatable/data.table/issues/3727). Since `data.table()` is used internally by `.()`, this brings the following examples in line with expectations in most cases. Thanks to @shrektan for the suggestion and PR.
+
+    ```R
+    DT = data.table(A=1:3, B=letters[1:3])
+    DT[A>3,   .(ITEM='A>3', A, B)]  # (1)
+    DT[A>3][, .(ITEM='A>3', A, B)]  # (2)
+    # the above are now equivalent as expected and return:  
+    Empty data.table (0 rows and 3 cols): ITEM,A,B
+    # Previously, (2) returned :
+          ITEM     A      B
+       <char> <int> <char>
+    1:    A>3    NA   <NA>
+    Warning messages:
+    1: In as.data.table.list(jval, .named = NULL) :
+      Item 2 has 0 rows but longest item has 1; filled with NA
+    2: In as.data.table.list(jval, .named = NULL) :
+      Item 3 has 0 rows but longest item has 1; filled with NA
+    ```
+
+    ```R
+    DT = data.table(A=1:3, B=letters[1:3], key="A")
+    DT[.(1:3, double()), B]
+    # new result :
+    character(0)   
+    # old result :
+    [1] "a" "b" "c"
+    Warning message:
+    In as.data.table.list(i) :
+      Item 2 has 0 rows but longest item has 3; filled with NA
+    ```
+
 ## BUG FIXES
 
 1. `by=.EACHI` when `i` is keyed but `on=` different columns than `i`'s key could create an invalidly keyed result, [#4603](https://github.com/Rdatatable/data.table/issues/4603) [#4911](https://github.com/Rdatatable/data.table/issues/4911). Thanks to @myoung3 and @adamaltmejd for reporting, and @ColeMiller1 for the PR. An invalid key is where a `data.table` is marked as sorted by the key columns but the data is not sorted by those columns, leading to incorrect results from subsequent queries.

R/as.data.table.R

@@ -129,6 +129,7 @@ as.data.table.list = function(x,
   eachncol = integer(n)
   missing.check.names = missing(check.names)
   origListNames = if (missing(.named)) names(x) else NULL  # as.data.table called directly, not from inside data.table() which provides .named, #3854
+  empty_atomic = FALSE  
   for (i in seq_len(n)) {
     xi = x[[i]]
     if (is.null(xi)) next    # eachncol already initialized to 0 by integer() above
@@ -148,10 +149,13 @@ as.data.table.list = function(x,
     }
     eachnrow[i] = NROW(xi)    # for a vector (including list() columns) returns the length
     eachncol[i] = NCOL(xi)    # for a vector returns 1
+    if (is.atomic(xi) && length(xi)==0L && !is.null(xi)) {
+      empty_atomic = TRUE  # any empty atomic (not empty list()) should result in nrows=0L, #3727
+    }
   }
   ncol = sum(eachncol)  # hence removes NULL items silently (no error or warning), #842.
   if (ncol==0L) return(null.data.table())
-  nrow = max(eachnrow)
+  nrow = if (empty_atomic) 0L else max(eachnrow)
   ans = vector("list",ncol)  # always return a new VECSXP
   recycle = function(x, nrow) {
     if (length(x)==nrow) {
@@ -173,8 +177,6 @@ as.data.table.list = function(x,
     if (is.null(xi)) { n_null = n_null+1L; next }
     if (eachnrow[i]>1L && nrow%%eachnrow[i]!=0L)   # in future: eachnrow[i]!=nrow
       warning("Item ", i, " has ", eachnrow[i], " rows but longest item has ", nrow, "; recycled with remainder.")
-    if (eachnrow[i]==0L && nrow>0L && is.atomic(xi))   # is.atomic to ignore list() since list() is a common way to initialize; let's not insist on list(NULL)
-      warning("Item ", i, " has 0 rows but longest item has ", nrow, "; filled with NA")  # the rep() in recycle() above creates the NA vector
     if (is.data.table(xi)) {   # matrix and data.frame were coerced to data.table above
       prefix = if (!isFALSE(.named[i]) && isTRUE(nchar(names(x)[i])>0L)) paste0(names(x)[i],".") else ""  # test 2058.12
       for (j in seq_along(xi)) {

inst/tests/tests.Rraw

@@ -5839,7 +5839,7 @@ test(1380, DT[a==TRUE], DT[3:4])
 # Fix #847, as.data.table.list and character(0) issue
 x <- data.table(a=character(0), b=character(0), c=numeric(0))
 setkey(x, a, b)
-test(1381, x[J("foo", character(0)), nomatch=0L], x, warning="Item 2 has 0 rows but longest item has 1; filled with NA")
+test(1381, x[J("foo", character(0)), nomatch=0L], x)
 
 # Fix for #813 and #758
 DT = data.table(x = 1:2)
@@ -13754,7 +13754,7 @@ test(1967.34, data.table(1:5, NULL), data.table(V1=1:5))
 ###   if (novname[i]) vnames[[i]] = namesi
 ### but, on pause for now pending #3193
 ### test(1967.35, data.table(1:5, matrix(6:15, nrow = 5L))
-test(1967.35, data.table(1:5, integer(0L)), data.table(1:5, NA_integer_), warning="Item 2 has 0 rows but longest item has 5; filled with NA")
+test(1967.35, data.table(1:5, integer(0L)), data.table(integer(0L), integer(0L)))  # no longer NA-fill zero-length, PR#4262
 test(1967.36, data.table(1:5, key = 5L), error = 'must be character')
 
 x = data.table(a = 1:5)
@@ -17346,3 +17346,14 @@ test(2170.2, DT[A > -1, which = NA], 1L)
 test(2170.3, DT[A > -1 | is.na(A), which = NA], integer())
 test(2170.4, DT[A > 10, which = NA], seq_len(nrow(DT)))
 test(2170.5, DT[!(A > 1), which = NA], c(1:3,6L)) # matches DT[A <= 1, which = NA]
+
+# data.table() zero-nrow result if any non-null & atomic element is length 0, #3727
+test(2171.1, data.table(A=double(), B=1:2), data.table(A=double(), B=integer()))
+DT = data.table(CODE=c('a','b'), DATE=1:2, VALUE=c(1.3, 1.5), key=c('CODE','DATE'))
+test(2171.2, DT[J(character(), 1), VALUE], double()) # because "J" is a wrapper of list()
+test(2171.3, data.table(A=NULL, B=1.0), data.table(B=1.0)) # NULL is omited
+test(2171.4, NROW(data.table(A=list(), B=1.0)), 1L) # empty list() regarded as `list(list())` which is length 1, and recycled
+DT = data.table(A=1:3, B=letters[1:3])
+test(2171.5, ans <- DT[A>3,   .(ITEM='A>3', A, B)],  # now identical as expected
+                    DT[A>3][, .(ITEM='A>3', A, B)])
+test(2171.6, ans, data.table(ITEM=character(), A=integer(), B=character())) # not just identical to each other, but correct too