Fixes rare non-equi join segfault, closes #4388. (#4389)

NEWS.md

@@ -107,6 +107,8 @@ unit = "s")
 
 12. `rbindlist` no longer errors when coercing complex vectors to character vectors, [#4202](https://github.com/Rdatatable/data.table/issues/4202). Thanks to @sritchie73 for reporting and the PR.
 
+13. A relatively rare case of segfault when combining non-equi joins with `by=.EACHI` is now fixed, closes [#4388](https://github.com/Rdatatable/data.table/issues/4388).
+
 ## NOTES
 
 0. Retrospective license change permission was sought from and granted by 4 contributors who were missed in [PR#2456](https://github.com/Rdatatable/data.table/pull/2456), [#4140](https://github.com/Rdatatable/data.table/pull/4140). We had used [GitHub's contributor page](https://github.com/Rdatatable/data.table/graphs/contributors) which omits 3 of these due to invalid email addresses, unlike GitLab's contributor page which includes the ids. The 4th omission was a PR to a script which should not have been excluded; a script is code too. We are sorry these contributors were not properly credited before. They have now been added to the contributors list as displayed on CRAN. All the contributors of code to data.table hold its copyright jointly; your contributions belong to you. You contributed to data.table when it had a particular license at that time, and you contributed on that basis. This is why in the last license change, all contributors of code were consulted and each had a veto.

R/data.table.R

@@ -497,7 +497,7 @@ replace_dot_alias = function(e) {
         if (nqbyjoin) {
           irows = if (length(xo)) xo[irows] else irows
           xo = setorder(setDT(list(indices=rep.int(indices__, len__), irows=irows)))[["irows"]]
-          ans = .Call(CnqRecreateIndices, xo, len__, indices__, max(indices__))
+          ans = .Call(CnqRecreateIndices, xo, len__, indices__, max(indices__), nomatch) # issue#4388 fix
           f__ = ans[[1L]]; len__ = ans[[2L]]
           allLen1 = FALSE # TODO; should this always be FALSE?
           irows = NULL # important to reset

inst/tests/tests.Rraw

@@ -11937,7 +11937,14 @@ DT2[, DT2_ID := .I][, (cols) := lapply(.SD, as.Date), .SDcols=cols]
 ans1 = DT2[DT1, on=.(RANDOM_STRING, START_DATE <= DATE, EXPIRY_DATE >= DATE), .N, by=.EACHI ]$N > 0L
 tmp  = DT1[DT2, on=.(RANDOM_STRING, DATE >= START_DATE, DATE <= EXPIRY_DATE), which=TRUE, nomatch=0L]
 ans2 = DT1[, DT1_ID %in% tmp]
-test(1848, ans1, ans2)
+test(1848.1, ans1, ans2)
+
+# Fix for #4388; related to #2275 fix
+x <- data.table(id = "a", t = as.ITime(c(31140L, 31920L, 31860L, 31680L, 31200L, 31380L, 31020L, 31260L, 31320L, 31560L, 31080L, 31800L, 31500L, 31440L, 31740L, 31620L)), s = c(37.19, 37.10, 37.10, 37.10, 37.1, 24.81, 61.99, 37.1, 37.1, 37.38, 49.56, 73.89, 37.38, 24.81, 37.01, 37.38), val = c(40L, 53L, 52L, 49L, 41L, 44L, 38L, 42L, 43L, 47L, 39L, 51L, 46L, 45L, 50L, 48L))
+y <- data.table(id = c("a", "b"), t1 = as.ITime(c(31020L, 42240L)), t2 = as.ITime(c(31920L, 43140L)), s1 = c(0L, 0L),
+  s2 = c(200, 200))
+# testing that it doesn't segfault
+test(1848.2, x[y, on=.(id, s >= s1, s <= s2, t >= t1, t <= t2), .(val), by=.EACHI, nomatch=0L, allow.cartesian=TRUE]$val, x$val)
 
 # when last field is quoted contains sep and select= is used too, #2464
 test(1849.1, fread('Date,Description,Amount,Balance\n20150725,abcd,"$3,004","$5,006"', select=c("Date", "Description", "Amount")),

src/nqrecreateindices.c

@@ -2,9 +2,9 @@
 
 // TODO: Add oxygen style comments and cleanup var names.
 // See other TODOs inside the function.
-SEXP nqRecreateIndices(SEXP xo, SEXP len, SEXP indices, SEXP nArg) {
+SEXP nqRecreateIndices(SEXP xo, SEXP len, SEXP indices, SEXP nArg, SEXP nomatch) {
 
-  R_len_t n=INTEGER(nArg)[0];
+  R_len_t n=INTEGER(nArg)[0], xn=length(xo);
   SEXP ans, newstarts, newlen;
   ans = PROTECT(allocVector(VECSXP, 2));
   SET_VECTOR_ELT(ans, 0, (newstarts = allocVector(INTSXP, n)));
@@ -14,6 +14,7 @@ SEXP nqRecreateIndices(SEXP xo, SEXP len, SEXP indices, SEXP nArg) {
   const int *iindices = INTEGER(indices);
   const int *ilen = INTEGER(len);
   const int *ixo = INTEGER(xo);
+  const int *inomatch = INTEGER(nomatch);
   int *inewstarts = INTEGER(newstarts);
 
   for (int i=0; i<n; ++i) inewlen[i] = 0;
@@ -25,8 +26,10 @@ SEXP nqRecreateIndices(SEXP xo, SEXP len, SEXP indices, SEXP nArg) {
   // TODO: revisit to see if this be simplified further when I've some time.
   R_len_t j=0, tmp=0;
   for (int i=0; i<n; ++i) {
-    if (ixo[j] <= 0) { // NA_integer_ = INT_MIN is checked in init.c
-      inewstarts[i] = ixo[j];
+    if (ixo[j] <= 0 || j >= xn) {
+      // NA_integer_ = INT_MIN is checked in init.c
+      // j >= xn needed for special nomatch=0L case, see issue#4388 (due to xo[irows] from R removing '0' value in xo)
+      inewstarts[i] = inomatch[0];
       j++; // newlen will be 1 for xo=NA and 0 for xo=0 .. but we need to increment by 1 for both
     } else {
       inewstarts[i] = tmp+1;