Change GEQO fitness comparisons to consider disabled_nodes. Prior to v18, the total cost of a plan was just one number, a floating-point value. Commit e22253467942fdb100087787c3e1e3a8620c54b2 introduced a second component, the number of disabled_nodes at or below that level of the plan tree. Unfortunately, I (rhaas) failed to realize that GEQO extracts the cost from each candidate plan and compares it directly as a measure of fitness.
This means that, at least in principle, it's possible for GEQO to prefer a plan with more disabled_nodes and lower cost over one with fewer disabled_nodes and higher cost, which is incorrect. It may be possible for that to become a practical issue in v18, but it's much more likely to become an issue in v19 due to pg_plan_advice. Hence, since the straightforward fix breaks ABI compatibility, back-patch only to v19. Backpatch-through: 19 Discussion: https://postgr.es/m/ca+tgmoymxy-jip5qdhqneiyfebzqsaro6o2d9e8sznzqi1b...@mail.gmail.com Branch ------ master Details ------- https://git.postgresql.org/pg/commitdiff/c9c660e6ae03027615adff90cf4b26da5f8f97c1 Modified Files -------------- src/backend/optimizer/geqo/geqo_eval.c | 18 ++++++++++------ src/backend/optimizer/geqo/geqo_main.c | 13 +++++++----- src/backend/optimizer/geqo/geqo_misc.c | 19 +++++++++++------ src/backend/optimizer/geqo/geqo_pool.c | 29 ++++++++++--------------- src/include/optimizer/geqo.h | 2 +- src/include/optimizer/geqo_gene.h | 39 +++++++++++++++++++++++++++++++++- src/tools/pgindent/typedefs.list | 1 + 7 files changed, 82 insertions(+), 39 deletions(-)
