[geometry] revised solution for robustness, we don't go to integer which was shortly done. This is partly a revert of r85867, where we scaled to integer. Now we make side information consistent by iteratively selecting a larger epsilon. This will be moved into a separate class. All tests pass now, also the complete ticket_9081 test which failed before

[SVN r86368]
This commit is contained in:
Barend Gehrels
2013-10-20 15:30:21 +00:00
parent 8ea1f1000b
commit 3a6ae0bacb
12 changed files with 547 additions and 211 deletions

View File

@@ -258,24 +258,19 @@ void test_all()
1, 0, 13);
***/
#ifdef _MSC_VER
#ifdef TEST_ISOVIST
test_one<polygon, polygon, polygon>("isovist",
isovist1[0], isovist1[1],
if_typed_tt<ct>(4, 2), 0, 0.279121891701124,
if_typed_tt<ct>(4, 3), 0, if_typed_tt<ct>(224.889211358929, 223.777),
if_typed_tt<ct>(0.001, 0.2));
if_typed_tt<ct>(4, 2), -1, 0.279121,
4, -1, 224.8892,
if_typed_tt<ct>(0.001, 0.1));
// SQL Server gives: 0.279121891701124 and 224.889211358929
// PostGIS gives: 0.279121991127244 and 224.889205853156
#endif
test_one<polygon, polygon, polygon>("ggl_list_20110306_javier",
ggl_list_20110306_javier[0], ggl_list_20110306_javier[1],
1, -1, 71495.3331,
2, -1, 8960.49049);
#endif
test_one<polygon, polygon, polygon>("ggl_list_20110307_javier",
ggl_list_20110307_javier[0], ggl_list_20110307_javier[1],