#2611 Fix week() calculation to start week 1 at days 1–7

NEWS.md

@@ -38,6 +38,8 @@
 
 1. `data.table(x=1, <expr>)`, where `<expr>` is an expression resulting in a 1-column matrix without column names, will eventually have names `x` and `V2`, not `x` and `V1`, consistent with `data.table(x=1, <expr>)` where `<expr>` results in an atomic vector, for example `data.table(x=1, cbind(1))` and `data.table(x=1, 1)` will both have columns named `x` and `V2`. In this release, the matrix case continues to be named `V1`, but the new behavior can be activated by setting `options(datatable.old.matrix.autoname)` to `FALSE`. See point 5 under Bug Fixes for more context; this change will provide more internal consistency as well as more consistency with `data.frame()`.
 
+2. The `week()` function will be fixed in a future release to correctly calculate weeks sequentially (days 1-7 as week 1), which is a potential breaking change. For now, the current (buggy) behavior remains the default. A one-time deprecation warning will be issued when the old and new behaviors differ; this can be suppressed with `options(datatable.warn.week.deprecation = FALSE)`. Users can opt-in to the correct behavior now with the temporary option `options(datatable.week.new = TRUE)`. See [#2611](https://github.com/Rdatatable/data.table/issues/2611) for details. Thanks @MichaelChirico for the report.
+
 ### NEW FEATURES
 
 1. New `sort_by()` method for data.tables, [#6662](https://github.com/Rdatatable/data.table/issues/6662). It uses `forder()` to improve upon the data.frame method and also matches `DT[order(...)]` behavior with respect to locale. Thanks @rikivillalba for the suggestion and PR.

inst/tests/tests.Rraw

@@ -18384,7 +18384,7 @@ x = c("1111-11-11", "2019-01-01", "2019-02-28", "2019-03-01", "2019-12-31", "202
 test(2236.1, yday(x), c(315L, 1L, 59L, 60L, 365L, 60L, 61L, 366L, 1L, 366L, 60L, NA))
 test(2236.2, mday(x), c(11L, 1L, 28L, 1L, 31L, 29L, 1L, 31L, 1L, 31L, 1L, NA))
 test(2236.3, wday(x), c(7L, 3L, 5L, 6L, 3L, 7L, 1L, 5L, 1L, 2L, 2L, NA))
-test(2236.4, week(x), c(46L, 1L, 9L, 9L, 53L, 9L, 9L, 53L, 1L, 53L, 9L, NA))
+test(2236.4, week(x), c(46L, 1L, 9L, 9L, 53L, 9L, 9L, 53L, 1L, 53L, 9L, NA), warning="The current behavior of data.table::week() is deprecated")
 test(2236.5, month(x), c(11L, 1L, 2L, 3L, 12L, 2L, 3L, 12L, 1L, 12L, 3L, NA))
 test(2236.6, quarter(x), c(4L, 1L, 1L, 1L, 4L, 1L, 1L, 4L, 1L, 4L, 1L, NA))
 test(2236.7, year(x), c(1111L, 2019L, 2019L, 2019L, 2019L, 2020L, 2020L, 2020L, 2040L, 2040L, 2100L, NA))
@@ -21815,3 +21815,12 @@ test(2341.24, fread('a
   # leading cmnt
 b
 ', comment.char = '#', strip.white = FALSE, sep = ","), data.table(a=c("  ", "b")))
+
+# week() sequential numbering fix tests #2611
+options(datatable.week.new = TRUE)
+test(2342.1, week(as.IDate("1970-01-01") + 0:7), c(1L,1L,1L,1L,1L,1L,1L,2L))  # Jan 1–7 all week 1, Jan 8 week 2
+test(2342.2, week(as.IDate(c("2012-02-28","2012-02-29","2012-03-01"))), c(9L,9L,9L))  # leap day stays in same week
+test(2342.3, week(as.IDate(c("2019-12-31","2020-01-01"))), c(53L,1L))  # year boundary non-leap → reset to 1
+test(2342.4, week(as.IDate(c("2020-12-31","2021-01-01"))), c(53L,1L))  # year boundary leap → reset to 1
+test(2342.5, week(as.IDate("2021-01-06") + 0:6), c(1L,1L,2L,2L,2L,2L,2L))
+test(2342.6, week(as.IDate(c("2016-02-27","2016-02-28","2016-02-29","2016-03-01","2016-03-02"))), c(9L,9L,9L,9L,9L))

src/idatetime.c

@@ -1,5 +1,7 @@
 #include "data.table.h"
 
+static int week_deprecation_warning_issued = 0;
+
 static const int YEARS400 = 146097;
 static const int YEARS100 = 36524;
 static const int YEARS4 = 1461;
@@ -58,13 +60,11 @@ void convertSingleDate(int x, datetype type, void *out)
 
     int leap = !years1 && (years4 || !years100);
 
-    if (type == YDAY || type == WEEK) {
+    if (type == YDAY) {
         int yday = days + 31 + 28 + leap;
         if (yday >= YEARS1 + leap)
             yday -= YEARS1 + leap;
         *(int *)out = ++yday;
-        if (type == WEEK)
-            *(int *)out = (*(int *)out / 7) + 1;
         return;
     }
 
@@ -143,6 +143,46 @@ SEXP convertDate(SEXP x, SEXP type)
     else if (!strcmp(ctype_str, "yearqtr")) { ctype = YEARQTR; ansint = false; }
     else internal_error(__func__, "invalid type, should have been caught before"); // # nocov
 
+    if (ctype == WEEK) {
+        SEXP ans = PROTECT(allocVector(INTSXP, n));
+        int *ansp = INTEGER(ans);
+
+        SEXP opt_new = GetOption(install("datatable.week.new"), R_NilValue);
+        bool use_new_behavior = isLogical(opt_new) && LOGICAL(opt_new)[0] == TRUE;
+
+        bool can_warn = false;
+        if (!use_new_behavior && !week_deprecation_warning_issued) {
+            SEXP opt_warn_depr = GetOption(install("datatable.warn.week.deprecation"), R_NilValue);
+            can_warn = !isLogical(opt_warn_depr) || LOGICAL(opt_warn_depr)[0] != FALSE;
+        }
+
+        for (int i = 0; i < n; i++) {
+            if (ix[i] == NA_INTEGER) {
+                ansp[i] = NA_INTEGER;
+                continue;
+            }
+            int yday;
+            convertSingleDate(ix[i], YDAY, &yday);
+
+            if (use_new_behavior) {
+                ansp[i] = ((yday - 1) / 7) + 1;
+            } else {
+                int old_week = (yday / 7) + 1;
+                ansp[i] = old_week;
+                if (can_warn) {
+                    int new_week = ((yday - 1) / 7) + 1;
+                    if (new_week != old_week) {
+                        Rf_warning("The current behavior of data.table::week() is deprecated and will be fixed in a future version to be sequential (day 1-7 is week 1). To opt-in to the new behavior now, run: options(datatable.week.new = TRUE). To suppress this warning, run: options(datatable.warn.week.deprecation = FALSE)");
+                        week_deprecation_warning_issued = 1;
+                        can_warn = false;
+                    }
+                }
+            }
+        }
+        UNPROTECT(1);
+        return ans;
+    }
+
     if (ansint) {
         SEXP ans = PROTECT(allocVector(INTSXP, n));
         int *ansp = INTEGER(ans);