fix(7571): bug fix for narm issue on gforce in int64 case

NEWS.md

@@ -34,6 +34,8 @@
 
 3. `fread("file://...")` works for file URIs with spaces, [#7550](https://github.com/Rdatatable/data.table/issues/7550). Thanks @aitap for the report and @MichaelChirico for the PR.
 
+4. `sum(<int64 column>)` by group is correct with missing entries and GForce activated ([#7571](https://github.com/Rdatatable/data.table/issues/7571)). Thanks to @rweberc for the report and @manmita for the fix. The issue was caused by a faulty early `break` that spilled between groups, and resulted in silently incorrect results!
+
 ## data.table [v1.18.0](https://github.com/Rdatatable/data.table/milestone/37?closed=1)  23 December 2025
 
 ### BREAKING CHANGE

inst/tests/tests.Rraw

@@ -21978,3 +21978,55 @@ local({
   test(2357.1, fread(f), DT)
   test(2357.2, fread(paste0("file://", f)), DT)
 })
+
+#7571 issue for na.rm on int64
+if (test_bit64) local({
+  # integer64 + GForce grouped sum with na.rm = FALSE
+  # Example 1 from issue: ids 1:8, 9, 9; three leading NAs then 4:10
+  dt_short = data.table(
+    id = c(1:8, 9, 9),
+    value = c(rep(NA_integer64_, 3L), as.integer64(4:10))
+  )
+  test(2358.1, options=c(datatable.optimize=2L),
+    dt_short[, sum(value, na.rm = FALSE), by = id]$V1,
+    as.integer64(c(NA, NA, NA, 4:8, 19))
+  )
+
+  # Example 2 from issue: ids in pairs, same values; checks multi-row groups
+  dt_short2 = data.table(
+    id = rep(1:5, each = 2L),
+    value = c(rep(NA_integer64_, 3L), as.integer64(4:10))
+  )
+  test(2358.2, options=c(datatable.optimize=2L),
+    dt_short2[, sum(value, na.rm = FALSE), by = id]$V1,
+    as.integer64(c(NA, NA, 11, 15, 19))
+  )
+
+  # Test mean for integer64 with NA
+  dt_mean = data.table(
+    id = c(1,1,2,2,3,3),
+    value = as.integer64(c(NA, NA, NA, 20000000, 5, 3))
+  )
+  test(2358.3, options=c(datatable.optimize=2L),
+    dt_mean[, mean(value, na.rm=FALSE), by = id]$V1,
+    c(NA, NA, 4)
+  )
+
+  # GForce sum vs base::sum for integer64
+  DT = data.table(id = sample(letters, 1000, TRUE), value = as.integer64(sample(c(1:100, NA), 1000, TRUE)))
+  gforce = DT[, .(gforce_sum = sum(value)), by=id]
+  base = DT[, .(true_sum = base::sum(value)), by=id]
+  merged = merge(gforce, base, by="id", all=TRUE)
+  test(2358.4, options=c(datatable.optimize=2L),
+    merged$gforce_sum, merged$true_sum
+  )
+
+  # GForce mean vs base::mean for integer64
+  DTm = data.table(id = sample(letters, 1000, TRUE), value = as.integer64(sample(c(1:100, NA), 1000, TRUE)))
+  gforce_m = DTm[, .(gforce_mean = mean(value)), by=id]
+  base_m = DTm[, .(true_mean = base::mean(value)), by=id]
+  merged_m = merge(gforce_m, base_m, by="id", all=TRUE)
+  test(2358.5, options=c(datatable.optimize=2L),
+    merged$gforce_mean, merged$true_mean
+  )
+})

src/gsumm.c

@@ -502,13 +502,13 @@ SEXP gsum(SEXP x, SEXP narmArg)
               const int64_t *my_gx = gx + b*batchSize + pos;
               const uint16_t *my_low = low + b*batchSize + pos;
               for (int i=0; i<howMany; i++) {
-                const int64_t elem = my_gx[i];
-                if (elem!=INT64_MIN) {
-                  _ans[my_low[i]] += elem;
-                } else {
-                  _ans[my_low[i]] = INT64_MIN;
-                  break;
+                if (_ans[my_low[i]] == INT64_MIN) continue;
+                const int64_t b = my_gx[i];
+                if (b == INT64_MIN) {
+                  if (!narm) _ans[my_low[i]] = INT64_MIN;
+                  continue;
                 }
+                _ans[my_low[i]] += b;
               }
             }
           }