Handle interaction of regexp's makesearch and MATCHALL more honestly.

tglsfdc · tglsfdc · commit 3068e4579932 · 2021-08-27T12:18:58.000-04:00
Second thoughts about commit 824bf71: we apply makesearch() to an NFA after having determined whether it is a MATCHALL pattern. Prepending ".*" doesn't make it non-MATCHALL, but it does change the maximum possible match length, and makesearch() failed to update that. This has no ill effects given the stylized usage of search NFAs, but it seems like it's better to keep the data structure consistent. In particular, fixing this allows more honest handling of the MATCHALL check in matchuntil(): we can now assert that maxmatchall is infinity, instead of lamely assuming that it should act that way. In passing, improve the code in dump[c]nfa so that infinite maxmatchall is printed as "inf" not a magic number.
diff --git a/src/backend/regex/regc_nfa.c b/src/backend/regex/regc_nfa.c
@@ -3602,8 +3602,13 @@ dumpnfa(struct nfa *nfa,
 	if (nfa->flags & HASLACONS)
 		fprintf(f, ", haslacons");
 	if (nfa->flags & MATCHALL)
-		fprintf(f, ", minmatchall %d, maxmatchall %d",
-				nfa->minmatchall, nfa->maxmatchall);
+	{
+		fprintf(f, ", minmatchall %d", nfa->minmatchall);
+		if (nfa->maxmatchall == DUPINF)
+			fprintf(f, ", maxmatchall inf");
+		else
+			fprintf(f, ", maxmatchall %d", nfa->maxmatchall);
+	}
 	fprintf(f, "\n");
 	for (s = nfa->states; s != NULL; s = s->next)
 	{
@@ -3766,8 +3771,13 @@ dumpcnfa(struct cnfa *cnfa,
 	if (cnfa->flags & HASLACONS)
 		fprintf(f, ", haslacons");
 	if (cnfa->flags & MATCHALL)
-		fprintf(f, ", minmatchall %d, maxmatchall %d",
-				cnfa->minmatchall, cnfa->maxmatchall);
+	{
+		fprintf(f, ", minmatchall %d", cnfa->minmatchall);
+		if (cnfa->maxmatchall == DUPINF)
+			fprintf(f, ", maxmatchall inf");
+		else
+			fprintf(f, ", maxmatchall %d", cnfa->maxmatchall);
+	}
 	fprintf(f, "\n");
 	for (st = 0; st < cnfa->nstates; st++)
 		dumpcstate(st, cnfa, f);
diff --git a/src/backend/regex/regcomp.c b/src/backend/regex/regcomp.c
@@ -598,6 +598,13 @@ makesearch(struct vars *v,
 		/* and ^* and \A* too -- not always necessary, but harmless */
 		newarc(nfa, PLAIN, nfa->bos[0], pre, pre);
 		newarc(nfa, PLAIN, nfa->bos[1], pre, pre);
+
+		/*
+		 * The pattern is still MATCHALL if it was before, but the max match
+		 * length is now infinity.
+		 */
+		if (nfa->flags & MATCHALL)
+			nfa->maxmatchall = DUPINF;
 	}
 
 	/*
diff --git a/src/backend/regex/rege_dfa.c b/src/backend/regex/rege_dfa.c
@@ -385,14 +385,10 @@ matchuntil(struct vars *v,
 	{
 		size_t		nchr = probe - v->start;
 
-		/*
-		 * It might seem that we should check maxmatchall too, but the .* at
-		 * the front of the pattern absorbs any extra characters (and it was
-		 * tacked on *after* computing minmatchall/maxmatchall).  Thus, we
-		 * should match if there are at least minmatchall characters.
-		 */
 		if (nchr < d->cnfa->minmatchall)
 			return 0;
+		/* maxmatchall will always be infinity, cf. makesearch() */
+		assert(d->cnfa->maxmatchall == DUPINF);
 		return 1;
 	}
 
