pgsql-bloat-estimation issues

Why `IndexTupleData` size is 2

5

Hi, I noticed when you compute the b-tree bloat, the IndexTupleData size is 2. But I couldn't understand why the size of `t_tid` + `t_info` is 2 after I read...

shangzixie

Use live_tuples instead of reltuples

7

See discussion: https://www.postgresql.org/message-id/flat/16db4468-edfa-830a-f921-39a50498e77e%402ndquadrant.com#[email protected]

ioguix

bug

Ignore table-wide NULL ratio for partial indexes where NULL is excluded

Technically speaking, this now ignores `pg_stats.null_frac` for indexed column `x` when index definition has `WHERE x IS NOT NULL`. This can fix some very skewed bloat estimates, for instance on...

rafbm

Optimize btree query

4

Requires CTE and thus Postgres 8.4. Here is a naïve comparison : ``` console $ time psql btree_bloat-8.2-8.3.sql > old.out real 0m2,087s user 0m0,035s sys 0m0,011s $ time psql btree_bloat.sql...

bersace

Toasted text

1

As discussed today on daliconsultants: In `pg_stats`, toasted data is exactly 18 bytes when data is totally toasted. So the query computes that the bloat is higher than real. Perhaps...

Krysztophe