# nameclass[ratio]: the shapes of shapes.sql that scripts/plangate.sh --calibrate ranks the plans of (a competitor besides the scan: b-tree, bitmap, sequential, PostgreSQL's own aggregate); every shape of the file is here, the class other for those that no class names; known: a wrong choice that is open work, with the worst ratio of its instructions to the best plan's that bench/results/estimate-misses.tsv records (written by --calibrate --update): a run that grows it by over 25% fails one_parent_narrow point point_day_count point point_days_three point point_days_ten point distinct_one_parent point one_day point range_only_by_month range deleted_rows range global range top_groups order_limit top_rows_by_ts order_limit top_rows_one_parent order_limit top_rows_trailing_key order_limit partitions_top order_limit count_all count count_by_kind count count_live count partitions_count count distinct_global count one_key group parent_key group two_keys_month group having group sum_by_parent group join_tier join in_active_parents join join_two_lookups known 2.63 exists_in_table join float_metric agg min_max agg mean_of_counts agg two_distincts agg running_total agg arms_union agg arms_per_parent_metrics agg two_keys_parent_kind other three_keys other filter_clauses other expression_key other float_sum_unrounded other top_two_keys other top_months other top_offset other top_rows_trailing_key_desc_offset other top_rows_trailing_key_ties other join_active_filter other join_month other join_spans_values other left_join_active other join_other_fact_filter other join_filter_on_lookup_column other lateral_latest_fact other lateral_first_fact_per_parent other lateral_recordset other lateral_series other lateral_dim_limit_one other sum_of_maxima_by_month other mean_of_means other groups_over_threshold other two_level_join_outer other count_of_groups other spine_range_join known 2.07 spine_derived_cells known spine_expression_key known 2.81 spine_span_kind other spine_series other spine_dims_cross known 1.78 exists_element other not_exists_element other jsonpath_element other exists_by_parent_month other not_exists_in_table other elements_counted other partitions_group other partition_pruned known 15.20 partitions_join other sum_by_kind_month other distinct_by_kind other distinct_by_month other select_distinct other distinct_join other arms_scalar_subselects other arms_filter other rank_within_month other share_of_total other lag_of_months other top_parents_per_kind other search_group other search_count other search_one_parent known 5.92 search_latest_rows other search_or_by_month other search_join other scope_path_filter_in other scope_path_filter_join other scope_expression_filter other skew_heaviest_parents other skew_rarest_parents other flat_list other any_array_subselect other in_constants other text_not_equal other top_rows_per_kind_unindexed_column known 1.28 skew_join_rows_read join dirty_pages_set group # dimexpr_dim_c (tuned lane, 250,000 rows): PostgreSQL's own choice, not the scan's: seq_page_cost 100 makes it drop the sequential plan that takes 822 M instructions for the plan it runs at 1,111 M; the scan (1,173 M) is not chosen and is not the best. dimexpr_dim_c known 1.35 dimexpr_two join dimsum_two join # dimexpr_few_sets (stock lane, 30,000 rows): PostgreSQL's own choice: it prices its hash join at 188 against 274 for the index plan that takes 0.44 M instructions instead of 2.34 M; the scan takes 0.66 M. A loss of 1.9 M instructions, under half a millisecond. dimexpr_few_sets known 5.32 dimrate_one join dimrate_two join dimright_dim_c join dimfull_dim_b join dimright_filtered join linkjoin_200_parents join linksemi_200_parents join linkanti_200_parents join linkleft_200_parents join linkjoin_2000_parents join linksemi_2000_parents join linkanti_2000_parents join linkleft_2000_parents join linkjoin_all_parents join linksemi_all_parents join linkanti_all_parents join linkleft_all_parents join