I created an SQLlite schema as follows:
CREATE TABLE tab1 (
year INTEGER,
tar_id TEXT,
content BLOB,
UNIQUE (year, tar_id) ON CONFLICT REPLACE);
CREATE INDEX tab1_ix1 ON bcas (year, tar_id);
Then I looked at a query plan:
sqlite> explain query plan select * from tab1 where tar_id = 1 and year = (select max(year) from tab1 where year < 2019 and tar_id = 1);
QUERY PLAN
|--SEARCH TABLE tab1 USING COVERING INDEX sqlite_autoindex_tab1_1 (year=? AND tar_id=?)
`--SCALAR SUBQUERY
`--SEARCH TABLE tab1 USING COVERING INDEX tab1_ix1 (year<?)
It seems to me that only one index would be sufficient to do this, but it uses both my explicit tab1_ix1 and and automatically generated sqlite_autoindex_tab1_1.
Is one of them redundant? If so, how do I get rid of one of them and get the same behaviour?
Yes, you have a redundant index. A unique constraint generates an index automatically. You do not need to create another one explicitly index for the same columns in the same order.
Note that an index on (tar_id, year) would be a different index, because the ordering of keys in the index matters.
Although both seam redundant but there is some detail to be focused. In first part of query "*" is used so the auto created candidate is best suited because it links all other columns to retrieve. While in second portion with "max" clause only two columns (year, tar_id) are under consideration and they are all present with needed sequence in "smaller" manually created index "tab1_ix1" and engine thinks smaller being efficient so it uses tab1_ix1 this time.
So the default auto created index will be utilized by engine if slight degradation in performance is affordable and second smaller tab1_ix1 seem burden maintenance wise.
Related
The problem
Using PostgreSQL 13, I ran into a performance issue selecting the highest id from a view that joins two tables, depending on the select statement I execute.
Here's a sample setup:
CREATE TABLE test1 (
id BIGSERIAL PRIMARY KEY,
joincol VARCHAR
);
CREATE TABLE test2 (
joincol VARCHAR
);
CREATE INDEX ON test1 (id);
CREATE INDEX ON test1 (joincol);
CREATE INDEX ON test2 (joincol);
CREATE VIEW testview AS (
SELECT test1.id,
test1.joincol AS t1charcol,
test2.joincol AS t2charcol
FROM test1, test2
WHERE test1.joincol = test2.joincol
);
What I found out
I'm executing two statements which result in completely different execution plans and runtimes. The following statement executes in less than 100ms. As far as I understand the execution plan, the runtime is independent of the rowcount, since Postgres iterates the rows one by one (starting at the highest id, using the index) until a join on a row is possible and immediately returns.
SELECT id FROM testview ORDER BY ID DESC LIMIT 1;
However, this one takes over 1 second on average (depending on rowcount), since the two tables are "joined completely", before Postgres uses the index to select the highest id.
SELECT MAX(id) FROM testview;
Please refer to this sample on dbfiddle to check the explain plans:
https://www.db-fiddle.com/f/bkMNeY6zXqBAYUsprJ5eWZ/1
My real environment
On my real environment test1 contains only a hand full of rows (< 100), having unique values in joincol. test2 contains up to ~10M rows, where joincol always matches a value of test1's joincol. test2's joincol is not nullable.
The actual question
Why does Postgres not recognize that it could use an Index Scan Backward on row basis for the second select? Is there anything I could improve on the tables/indexes?
Queries not strictly equivalent
why does Postgres not recognize that it could use a Index Scan Backward on row basis for the second select?
To make the context clear:
max(id) excludes NULL values. But ORDER BY ... LIMIT 1 does not.
NULL values sort last in ascending sort order, and first in descending. So an Index Scan Backward might not find the greatest value (according to max()) first, but any number of NULL values.
The formal equivalent of:
SELECT max(id) FROM testview;
is not:
SELECT id FROM testview ORDER BY id DESC LIMIT 1;
but:
SELECT id FROM testview ORDER BY id DESC NULLS LAST LIMIT 1;
The latter query doesn't get the fast query plan. But it would with an index with matching sort order: (id DESC NULLS LAST).
That's different for the aggregate functions min() and max(). Those get a fast plan when targeting table test1 directly using the plain PK index on (id). But not when based on the view (or the underlying join-query directly - the view is not the blocker). An index sorting NULL values in the right place has hardly any effect.
We know that id in this query can never be NULL. The column is defined NOT NULL. And the join in the view is effectively an INNER JOIN which cannot introduce NULL values for id.
We also know that the index on test.id cannot contain NULL values.
But the Postgres query planner is not an AI. (Nor does it try to be, that could get out of hands quickly.) I see two shortcomings:
min() and max() get the fast plan only when targeting the table, regardless of index sort order, an index condition is added: Index Cond: (id IS NOT NULL)
ORDER BY ... LIMIT 1 gets the fast plan only with the exactly matching index sort order.
Not sure, whether that might be improved (easily).
db<>fiddle here - demonstrating all of the above
Indexes
Is there anything I could improve on the tables/indexes?
This index is completely useless:
CREATE INDEX ON "test" ("id");
The PK on test.id is implemented with a unique index on the column, that already covers everything the additional index might do for you.
There may be more, waiting for the question to clear up.
Distorted test case
The test case is too far away from actual use case to be meaningful.
In the test setup, each table has 100k rows, there is no guarantee that every value in joincol has a match on the other side, and both columns can be NULL
Your real case has 10M rows in table1 and < 100 rows in table2, every value in table1.joincol has a match in table2.joincol, both are defined NOT NULL, and table2.joincol is unique. A classical one-to-many relationship. There should be a UNIQUE constraint on table2.joincol and a FK constraint t1.joincol --> t2.joincol.
But that's currently all twisted in the question. Standing by till that's cleaned up.
This is a very good problem, and good testcase.
I tested it in postgres 9.3 perhaps 13 is can it more more fast.
I used Occam's Razor and i excluded some possiblities
View (without view is slow to)
JOIN can filter some rows (unfortunatly in your test not, but more length md5 5-6 yes)
Other basic equivalent select statements not solve yout problem (inner query or exists)
I achieved to use just index, but because the tables isn't bigger than indexes it was not the solution.
I think
CREATE INDEX on "test" ("id");
is useless, because PK!
If you change this
CREATE INDEX on "test" ("joincol");
to this
CREATE INDEX ON TEST (joincol, id);
Than the second query use just indexes.
After you run this
REINDEX table test;
REINDEX table test2;
VACUUM ANALYZE test;
VACUUM ANALYZE test2;
you can achive some performance tuning. Because you created indexes before inserts.
I think the reason is the two aim of DB.
First aim optimalize just some row. So run Nested Loop. You can force it with limit x.
Second aim optimalize whole table. Run this query fast for whole table.
In this situation postgres optimalizer didn't notice that simple MAX can run with NESTED LOOP. Or perhaps postgres cannot use limit in aggregate clause (can run on whole partial select, what is filtered with query).
And this is very expensive. But you have possiblities to write there other aggregates, like SUM, MIN, AVG stb.
Perhaps can help you the Window functions too.
On the exact same table, if I was to put one index on it, either:
CREATE INDEX ix_single ON MyTable (uid asc) include (columnone)
or:
CREATE INDEX ix_multi ON MyTable (uid asc) include (
columnone,
columntwo,
columnthree,
....
columnX
)
Would the second index cause an even greater lag on how long it takes to write to the table than the first one? And why?
Included columns will need more diskspace as well as time on data manipulation...
If there is a clustered index on this table too (ideally on a implicitly sorted column like an IDENTITY column to avoid fragmentation) this will serve as fast lookup on all columns (but you must create the clustered index before the other one...)
To include columns into an index is a usefull approach in extremely performance related issues only...
I am not db guy. But I need to create tables and do CRUD operations on them. I get confused should I create the index on all columns by default
or not? Here is my understanding which I consider while creating index.
Index basically contains the memory location range ( starting memory location where first value is stored to end memory location where last value is
stored). So when we insert any value in table index for column needs to be updated as it has got one more value but update of column
value wont have any impact on index value. Right? So bottom line is when my column is used in join between two tables we should consider
creating index on column used in join but all other columns can be skipped because if we create index on them it will involve extra cost of
updating index value when new value is inserted in column.Right?
Consider this scenario where table mytable contains two three columns i.e col1,col2,col3. Now we fire this query
select col1,col2 from mytable
Now there are two cases here. In first case we create the index on col1 and col2. In second case we don't create any index.** As per my understanding
case 1 will be faster than case2 because in case 1 we oracle can quickly find column memory location. So here I have not used any join columns but
still index is helping here. So should I consider creating index here or not?**
What if in the same scenario above if we fire
select * from mytable
instead of
select col1,col2 from mytable
Will index help here?
Don't create Indexes in every column! It will slow things down on insert/delete/update operations.
As a simple reminder, you can create an index in columns that are common in WHERE, ORDER BY and GROUP BY clauses. You may consider adding an index in colums that are used to relate other tables (through a JOIN, for example)
Example:
SELECT col1,col2,col3 FROM my_table WHERE col2=1
Here, creating an index on col2 would help this query a lot.
Also, consider index selectivity. Simply put, create index on values that has a "big domain", i.e. Ids, names, etc. Don't create them on Male/Female columns.
but update of column value wont have any impact on index value. Right?
No. Updating an indexed column will have an impact. The Oracle 11g performance manual states that:
UPDATE statements that modify indexed columns and INSERT and DELETE
statements that modify indexed tables take longer than if there were
no index. Such SQL statements must modify data in indexes and data in
tables. They also create additional undo and redo.
So bottom line is when my column is used in join between two tables we should consider creating index on column used in join but all other columns can be skipped because if we create index on them it will involve extra cost of updating index value when new value is inserted in column. Right?
Not just Inserts but any other Data Manipulation Language statement.
Consider this scenario . . . Will index help here?
With regards to this last paragraph, why not build some test cases with representative data volumes so that you prove or disprove your assumptions about which columns you should index?
In the specific scenario you give, there is no WHERE clause, so a table scan is going to be used or the index scan will be used, but you're only dropping one column, so the performance might not be that different. In the second scenario, the index shouldn't be used, since it isn't covering and there is no WHERE clause. If there were a WHERE clause, the index could allow the filtering to reduce the number of rows which need to be looked up to get the missing column.
Oracle has a number of different tables, including heap or index organized tables.
If an index is covering, it is more likely to be used, especially when selective. But note that an index organized table is not better than a covering index on a heap when there are constraints in the WHERE clause and far fewer columns in the covering index than in the base table.
Creating indexes with more columns than are actually used only helps if they are more likely to make the index covering, but adding all the columns would be similar to an index organized table. Note that Oracle does not have the equivalent of SQL Server's INCLUDE (COLUMN) which can be used to make indexes more covering (it's effectively making an additional clustered index of only a subset of the columns - useful if you want an index to be unique but also add some data which you don't want to be considered in the uniqueness but helps to make it covering for more queries)
You need to look at your plans and then determine if indexes will help things. And then look at the plans afterwards to see if they made a difference.
I need help understanding how to create indexes. I have a table that looks like this
Id
Name
Age
Location
Education,
PhoneNumber
My query looks like this:
SELECT *
FROM table1
WHERE name = 'sam'
What's the correct way to create an index for this with included columns?
What if the query has a order by statement?
SELECT *
FROM table1
WHERE name = 'sam'
ORDER BY id DESC
What if I have 2 parameters in my where statement?
SELECT *
FROM table1
WHERE name = 'sam'
AND age > 12
The correct way to create an index with included columns? Either via Management Studio/Toad/etc, or SQL (documentation):
CREATE INDEX idx_table_1 ON db.table_1 (name) INCLUDE (id)
What if the Query has an ORDER BY
The ORDER BY can use indexes, if the optimizer sees fit to (determined by table statistics & query). It's up to you to test if a composite index or an index with INCLUDE columns works best by reviewing the query cost.
If id is the clustered key (not always the primary key though), I probably wouldn't INCLUDE the column...
What if I have 2 parameters in my where statement?
Same as above - you need to test what works best for your query. Might be composite, or include, or separate indexes.
But keep in mind that:
tweaking for one query won't necessarily benefit every other query
indexes do slow down INSERT/UPDATE/DELETE statements, and require maintenance
You can use the Database Tuning Advisor (DTA) for index recommendations, including when some are redundant
Recommended reading
I highly recommend reading Kimberly Tripp's "The Tipping Point" for a better understanding of index decisions and impacts.
Since I do not know which exactly tasks your DB is going to implement and how many records in it, I would suggest that you take a look at the Index Basics MSDN article. It will allow you to decide yourself which indexes to create.
If ID is your primary and/or clustered index key, just create an index on Name, Age. This will cover all three queries.
Included fields are best used to retrieve row-level values for columns that are not in the filter list, or to retrieve aggregate values where the sorted field is in the GROUP BY clause.
If inserts are rare, create as much indexes as You want.
For first query create index for name column.
Id column I think already is primary key...
Create 2nd index with name and age. You can keep only one index: 'name, ag'e and it will not be much slower for 1st query.
I've just heard the term covered index in some database discussion - what does it mean?
A covering index is an index that contains all of, and possibly more, the columns you need for your query.
For instance, this:
SELECT *
FROM tablename
WHERE criteria
will typically use indexes to speed up the resolution of which rows to retrieve using criteria, but then it will go to the full table to retrieve the rows.
However, if the index contained the columns column1, column2 and column3, then this sql:
SELECT column1, column2
FROM tablename
WHERE criteria
and, provided that particular index could be used to speed up the resolution of which rows to retrieve, the index already contains the values of the columns you're interested in, so it won't have to go to the table to retrieve the rows, but can produce the results directly from the index.
This can also be used if you see that a typical query uses 1-2 columns to resolve which rows, and then typically adds another 1-2 columns, it could be beneficial to append those extra columns (if they're the same all over) to the index, so that the query processor can get everything from the index itself.
Here's an article: Index Covering Boosts SQL Server Query Performance on the subject.
Covering index is just an ordinary index. It's called "covering" if it can satisfy query without necessity to analyze data.
example:
CREATE TABLE MyTable
(
ID INT IDENTITY PRIMARY KEY,
Foo INT
)
CREATE NONCLUSTERED INDEX index1 ON MyTable(ID, Foo)
SELECT ID, Foo FROM MyTable -- All requested data are covered by index
This is one of the fastest methods to retrieve data from SQL server.
Covering indexes are indexes which "cover" all columns needed from a specific table, removing the need to access the physical table at all for a given query/ operation.
Since the index contains the desired columns (or a superset of them), table access can be replaced with an index lookup or scan -- which is generally much faster.
Columns to cover:
parameterized or static conditions; columns restricted by a parameterized or constant condition.
join columns; columns dynamically used for joining
selected columns; to answer selected values.
While covering indexes can often provide good benefit for retrieval, they do add somewhat to insert/ update overhead; due to the need to write extra or larger index rows on every update.
Covering indexes for Joined Queries
Covering indexes are probably most valuable as a performance technique for joined queries. This is because joined queries are more costly & more likely then single-table retrievals to suffer high cost performance problems.
in a joined query, covering indexes should be considered per-table.
each 'covering index' removes a physical table access from the plan & replaces it with index-only access.
investigate the plan costs & experiment with which tables are most worthwhile to replace by a covering index.
by this means, the multiplicative cost of large join plans can be significantly reduced.
For example:
select oi.title, c.name, c.address
from porderitem poi
join porder po on po.id = poi.fk_order
join customer c on c.id = po.fk_customer
where po.orderdate > ? and po.status = 'SHIPPING';
create index porder_custitem on porder (orderdate, id, status, fk_customer);
See:
http://literatejava.com/sql/covering-indexes-query-optimization/
Lets say you have a simple table with the below columns, you have only indexed Id here:
Id (Int), Telephone_Number (Int), Name (VARCHAR), Address (VARCHAR)
Imagine you have to run the below query and check whether its using index, and whether performing efficiently without I/O calls or not. Remember, you have only created an index on Id.
SELECT Id FROM mytable WHERE Telephone_Number = '55442233';
When you check for performance on this query you will be dissappointed, since Telephone_Number is not indexed this needs to fetch rows from table using I/O calls. So, this is not a covering indexed since there is some column in query which is not indexed, which leads to frequent I/O calls.
To make it a covered index you need to create a composite index on (Id, Telephone_Number).
For more details, please refer to this blog:
https://www.percona.com/blog/2006/11/23/covering-index-and-prefix-indexes/