Jeff said:
https://www.postgresql.org/message-id/CAMkU%3D1zBJNVo2DGYBgLJqpu8fyjCE_ys%2Bmsr6pOEoiwA7y5jrA%40mail.gmail.com
|What would I find very useful is a verbosity option to get the cost
|estimates expressed as a multiplier of each *_cost parameter, rather than
|just as a scalar.
I guess the goal is something like
EXPLAIN(COSTS, VERBOSE) -- or some new option?
..would show something like
Seq Scan on public.sites (cost=0.00..2.90 rows=160 width=107)
Total tosts: Seq page: 1.01 Random page: 1.23 CPU tuple: .05 CPU oper: .01
Startup cost: [...]
It seems to me that's "just" a matter of redefining Cost and fixing everything
that breaks:
struct Cost {
double seq, rand;
double cpu_tuple, cpu_index_tuple, cpu_oper;
double parallel_setup; // This is probably always in startup_cost and
never in run_cost
double parallel_tuple; // This is probably always in run_cost and never
in startup_cost
double disable;
};
I'm perhaps 50% done with that - is there some agreement that's a desirable
goal and a good way to do it ?
To give an idea what I'm doing, there's a bunch of stuff like this:
- if (path1->startup_cost < path2->startup_cost)
+ if (cost_asscalar(&path1->startup_cost) <
cost_asscalar(&path2->startup_cost))
- qual_arg_cost += index_qual_cost.startup +
index_qual_cost.per_tuple;
+ cost_add2(&qual_arg_cost, &index_qual_cost.startup,
&index_qual_cost.per_tuple);
- if (cost.per_tuple > 10 * cpu_operator_cost)
+ if (cost_isgt_scalar(&cost.per_tuple, 10 *
cpu_operator_cost))
And a great deal of stuff like this:
- run_cost += cpu_run_cost;
+ cost_add(&run_cost, &cpu_run_cost);
/* tlist eval costs are paid per output row, not per tuple scanned */
- startup_cost += path->pathtarget->cost.startup;
- run_cost += path->pathtarget->cost.per_tuple * path->rows;
+ cost_add(&startup_cost, &path->pathtarget->cost.startup);
+ cost_add_mul(&run_cost, &path->pathtarget->cost.per_tuple, path->rows);
path->startup_cost = startup_cost;
- path->total_cost = startup_cost + run_cost;
+ cost_set_sum2(&path->total_cost, &startup_cost, &run_cost);
As I've written it, that's somewhat different from Jeff's suggestion, as all
the entries in my struct are in units of cost. That seems easier due to (for
example) per-tablespace IO costs.
I'd rather know sooner than later if there's a better way.
Justin
