4 * Check for bug in GiNaC::ex::match() described here:
5 * http://www.ginac.de/pipermail/ginac-devel/2006-April/000942.html
8 #include "error_report.hpp"
10 using namespace GiNaC;
13 * basic::match(lst&) used to have an obscure side effect: repl_lst
14 * could be modified even if the match failed! Although this "feature"
15 * was documented it happend to be very confusing *even for GiNaC
17 * http://www.ginac.de/pipermail/ginac-devel/2006-April/000942.html
19 * It was fixed in 192ed7390b7b2b705ad100e3db0a92eedd2b20ad. Let's make
20 * sure it will be never re-added:
22 static void failed_match_have_side_effects()
26 ex pattern = pow(wild(0), -1);
27 // obviously e does NOT match the pattern
29 bool match_p = e.match(pattern, repls);
30 bug_on(match_p, "match(" << e << ", " << pattern << ") says \"Yes\"");
31 bug_on(repls.size() != 0,
32 "failed match have side effects: repls = " << repls);
36 * As a consequence of the bug described above pattern matching can wrongly
37 * fail. In particular, x^5*y^(-1) fails to match ($0)^(-1)*x^($2).
39 * The first thing that is attempted to match is x^5 with $0^(-1). This match
40 * will fail. However repl_lst will contain $0 == x as a side effect. This
41 * repl_lst will prevent the match of y^(-1) to ($0)^(-1) to succeed.
43 * This issue was worked around by 73f0ce4cf8d91f073f35a45443f5fbe886921c5c.
44 * Now we have a real fix (192ed7390b7b2b705ad100e3db0a92eedd2b20ad), but
47 static void match_false_negative()
49 symbol x("x"), y("y");
50 ex e = pow(x, 5)*pow(y, -1);
51 ex pattern = pow(wild(0), -1)*pow(x, wild(2));
53 bool match_p = e.match(pattern, repls);
54 bug_on(!match_p, "false negative: " << e << " did not match "
58 int main(int argc, char** argv)
60 std::cout << "checking for historical bugs in match()... " << std::flush;
61 failed_match_have_side_effects();
62 match_false_negative();
63 std::cout << "not found. ";