HypoPG
HypoPG is a PostgreSQL extension adding support for hypothetical indexes.
An hypothetical -- or virtual -- index is an index that doesn't really exists, and thus doesn't cost CPU, disk or any resource to create. They're useful to know if specific indexes can increase performance for problematic queries, since you can know if PostgreSQL will use these indexes or not without having to spend resources to create them.
For more thorough informations, please consult the official documentation.
For other general information, you can also consult this blog post.
Installation
- Compatible with PostgreSQL 9.2 and above
- Needs PostgreSQL header files
- Decompress the tarball
sudo make install- In every needed database:
CREATE EXTENSION hypopg;
Updating the extension
Note that hypopg doesn't provide extension upgrade scripts, as there's no data saved in any of the objects created. Therefore, you need to first drop the extension then create it again to get the new version.
Usage
NOTE: The hypothetical indexes are contained in a single backend. Therefore,
if you add multiple hypothetical indexes, concurrent connections doing
EXPLAIN won't be bothered by your hypothetical indexes.
Assuming a simple test case:
rjuju=# CREATE TABLE hypo AS SELECT id, 'line ' || id AS val FROM generate_series(1,10000) id;
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
-------------------------------------------------------
Seq Scan on hypo (cost=0.00..180.00 rows=1 width=13)
Filter: (id = 1)
(2 rows)
The easiest way to create an hypothetical index is to use the
hypopg_create_index functions with a regular CREATE INDEX statement as arg.
For instance:
rjuju=# SELECT * FROM hypopg_create_index('CREATE INDEX ON hypo (id)');
NOTE: Some information from the CREATE INDEX statement will be ignored, such as
the index name if provided. Some of the ignored information will be handled in
a future release.
You can check the available hypothetical indexes in your own backend:
rjuju=# SELECT * FROM hypopg_list_indexes ;
indexrelid | indexname | nspname | relname | amname
-----------+-------------------------------------------+---------+---------+--------
205101 | <41072>btree_hypo_id | public | hypo | btree
If you need more technical information on the hypothetical indexes, the
hypopg() function will return the hypothetical indexes in a similar way as
pg_index system catalog.
And now, let's see if your previous EXPLAIN statement would use such an index:
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
------------------------------------------------------------------------------------
Index Scan using <41072>hypo_btree_hypo_id on hypo (cost=0.29..8.30 rows=1 width=13)
Index Cond: (id = 1)
(2 rows)
Of course, only EXPLAIN without ANALYZE will use hypothetical indexes:
rjuju=# EXPLAIN ANALYZE SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
-------------------------------------------------------------------------------------------------
Seq Scan on hypo (cost=0.00..180.00 rows=1 width=13) (actual time=0.036..6.072 rows=1 loops=1)
Filter: (id = 1)
Rows Removed by Filter: 9999
Planning time: 0.109 ms
Execution time: 6.113 ms
(5 rows)
To remove your backend's hypothetical indexes, you can use the function
hypopg_drop_index(indexrelid) with the OID that the hypopg_list_indexes
view returns and call hypopg_reset() to remove all at once, or just close
your current connection.
Continuing with the above case, you can hide existing indexes,
but should be use hypopg_reset() to clear the previous effects of other indexes at first.
Create two real indexes and run EXPLAIN:
rjuju=# SELECT hypopg_reset();
rjuju=# CREATE INDEX ON hypo(id);
rjuju=# CREATE INDEX ON hypo(id, val);
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
----------------------------------------------------------------------------------
Index Only Scan using hypo_id_val_idx on hypo (cost=0.29..8.30 rows=1 width=13)
Index Cond: (id = 1)
(2 rows)
The query plan is using the hypo_id_val_idx index. Use hypopg_hide_index(oid) to hide one of the indexes:
rjuju=# SELECT hypopg_hide_index('hypo_id_val_idx'::REGCLASS);
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
-------------------------------------------------------------------------
Index Scan using hypo_id_idx on hypo (cost=0.29..8.30 rows=1 width=13)
Index Cond: (id = 1)
(2 rows)
The query plan is using the other index hypo_id_idx now. Use hypopg_hide_index(oid) to hide it:
rjuju=# SELECT hypopg_hide_index('hypo_id_idx'::REGCLASS);
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
-------------------------------------------------------
Seq Scan on hypo (cost=0.00..180.00 rows=1 width=13)
Filter: (id = 1)
(2 rows)
And now the query plan changes back to Seq Scan. Use hypopg_unhide_index(oid) to restore index:
rjuju=# SELECT hypopg_unhide_index('hypo_id_idx'::regclass);
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
-------------------------------------------------------------------------
Index Scan using hypo_id_idx on hypo (cost=0.29..8.30 rows=1 width=13)
Index Cond: (id = 1)
(2 rows)
Of course, you can also hide hypothetical indexes:
rjuju=# SELECT hypopg_create_index('CREATE INDEX ON hypo(id)');
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
------------------------------------------------------------------------------------
Index Scan using "<12659>btree_hypo_id" on hypo (cost=0.04..8.05 rows=1 width=13)
Index Cond: (id = 1)
(2 rows)
rjuju=# SELECT hypopg_hide_index(12659);
rjuju=# EXPLAIN SELECT * FROM hypo WHERE id = 1;
QUERY PLAN
-------------------------------------------------------
Seq Scan on hypo (cost=0.00..180.00 rows=1 width=13)
Filter: (id = 1)
(2 rows)
You can check which indexes are hidden using hypopg_hidden_indexes() or the hypopg_hidden_indexes view:
rjuju=# SELECT * FROM hypopg_hidden_indexes();
indexid
---------
526604
526603
12659
(3 rows)
rjuju=# SELECT * FROM hypopg_hidden_indexes;
indexrelid | index_name | schema_name | table_name | am_name | is_hypo
------------+----------------------+-------------+------------+---------+---------
12659 | <12659>btree_hypo_id | public | hypo | btree | t
526603 | hypo_id_idx | public | hypo | btree | f
526604 | hypo_id_val_idx | public | hypo | btree | f
(3 rows)
To restore all existing indexes, you can use the function hypopg_unhide_all_indexes().
Note that the functionality to hide existing indexes only applies to the EXPLAIN command in the current session
and will not affect other sessions.