fix the bug when keys contain non UTF8 strings

NEWS.md

@@ -124,6 +124,9 @@
 
 30. "Negative length vectors not allowed" error when grouping `median` and `var` fixed, [#2046](https://github.com/Rdatatable/data.table/issues/2046) and [#2111](https://github.com/Rdatatable/data.table/issues/2111). Thanks to @caneff and @osofr for reporting and to @kmillar for debugging and explaining the cause.
 
+31. Fixed a bug on Windows where `data.table`s containing non-UTF8 strings in `key`s were not properly sorted, [#2462](https://github.com/Rdatatable/data.table/issues/2462), [#1826](https://github.com/Rdatatable/data.table/issues/1826)  and [StackOverflow](https://stackoverflow.com/questions/47599934/why-doesnt-r-data-table-support-well-for-non-ascii-keys-on-windows). Thanks to @shrektan for reporting and fixing.
+
+
 #### NOTES
 
 0. The license has been changed from GPL to MPL (Mozilla Public License). All contributors were consulted and approved. [PR#2456](https://github.com/Rdatatable/data.table/pull/2456) details the reasons for the change.

inst/tests/issue_2566.csv

@@ -0,0 +1,6 @@
+x,y,z
+公允价值变动损益,公允价值变动损益,1
+红利收入,红利收入,2
+价差收入,价差收入,3
+其他业务支出,其他业务支出,4
+资产减值损失,资产减值损失,5

inst/tests/tests.Rraw

@@ -11258,6 +11258,18 @@ test(1863.1, DT[, median(y), by=x], data.table(x=1, V1=0, key="x"))
 DT = data.table(col1 = c(1,1,1, 2,2,2), col2 = c(2,2,2,1,1,1), ID = c(rep(1,3), rep(2,3)), key="ID")
 test(1863.2, DT[, lapply(.SD, var), by=ID], data.table(ID=c(1,2), col1=0, col2=0, key="ID"))
 
+# Fix the bug when keys contain non UTF8 strings #2566 #2462 #1826
+# Only on Windows platform it might fail, because other platforms use UTF8 as the native encoding.
+DT <- fread(file = testDir("issue_2566.csv"), encoding = "UTF-8")
+# `fread` return a utf-8 encoded data, we should convert x to native encoding.
+# However, we need this condition to ensure the native encoding can be used for Chinese characters.
+# Otherwise, the test will fail because the strings have been damaged.
+if (identical(enc2utf8(enc2native(DT$x)), DT$x)) DT[, x:= enc2native(x)]
+setkey(DT, x)
+test(1864.1, DT[J("\u516c\u5141\u4ef7\u503c\u53d8\u52a8\u635f\u76ca"), z], 1L)
+setkey(DT, y)
+test(1864.2, DT[J("\u516c\u5141\u4ef7\u503c\u53d8\u52a8\u635f\u76ca"), z], 1L)
+
 
 ##########################
 

src/forder.c

@@ -866,7 +866,7 @@ static void csort(SEXP *x, int *o, int n)
   /* can't use otmp, since iradix might be called here and that uses otmp (and xtmp).
      alloc_csort_otmp(n) is called from forder for either n=nrow if 1st column,
      or n=maxgrpn if onwards columns */
-  for(i=0; i<n; i++) csort_otmp[i] = (x[i] == NA_STRING) ? NA_INTEGER : -TRUELENGTH(x[i]);
+  for(i=0; i<n; i++) csort_otmp[i] = (x[i] == NA_STRING) ? NA_INTEGER : -TRUELENGTH(ENC2UTF8(x[i]));
   if (nalast == 0 && n == 2) {                        // special case for nalast==0. n==1 is handled inside forder. at least 1 will be NA here
     if (o[0] == -1) for (i=0; i<n; i++) o[i] = i+1;    // else use o from caller directly (not 1st column)
     for (int i=0; i<n; i++) if (csort_otmp[i] == NA_INTEGER) o[i] = 0;
@@ -899,7 +899,7 @@ static void csort_pre(SEXP *x, int n)
   // savetl_init() is called once at the start of forder
   old_un = ustr_n;
   for(i=0; i<n; i++) {
-    s = x[i];
+    s = ENC2UTF8(x[i]);
     if (TRUELENGTH(s)<0) continue;   // this case first as it's the most frequent. Already in ustr, this negative is its ordering.
     if (TRUELENGTH(s)>0) {  // Save any of R's own usage of tl (assumed positive, so we can both count and save in one scan), to restore
       savetl(s);          // afterwards. From R 2.14.0, tl is initialized to 0, prior to that it was random so this step saved too much.