</indexterm>

<para>

<literal>bloom</> is acontrib which implements index access method. It comes

as example of custom access methods and generic WAL records usage. But it

is also useful itself.

<literal>bloom</> is amodule which implements an index access method. It comes

asanexample of custom access methods and generic WAL records usage. But it

is also usefulinitself.

</para>

<sect2>

<title>Introduction</title>

<para>

<ulink url="http://en.wikipedia.org/wiki/Bloom_filter">Bloom filter</ulink>

allows fast exclusion of non-candidate tuples.

Since signature is a lossy representation of all indexed attributes,

search resultsshould be rechecked using heap information.

Usercan specify signature length (in uint16, default is 5) and the number of

bits, which can besetted, per attribute (1 < colN < 2048).

allows fast exclusion of non-candidate tuples via signatures.

Sinceasignature is a lossy representation of all indexed attributes,

search resultsmust be rechecked using heap information.

The usercan specify signature length (in uint16, default is 5) and the

number ofbits, which can beset per attribute (1 < colN < 2048).

</para>

<para>

This index is useful if table has many attributes and queries can include

their arbitarycombinations.Traditional<literal>btree</> index is faster

than bloom index, but it'drequiretoomany indexes to support all possible

queries, while oneneed onlyonebloom index. Bloom index supports only

equality comparison. Since it's a signature file, not a tree, it always

shouldbereaded fully, but sequentially, so index search performance is

constant and doesn't depend on a query.

This index is useful ifatable has many attributes and queries include

arbitrarycombinations of them.A traditional<literal>btree</> index is

fasterthanabloom index, but it canrequire many indexes to support all

possiblequeries where oneneeds onlya singlebloom index.ABloom index

supports onlyequality comparison. Since it's a signature file,andnot a

tree, it always mustberead fully, but sequentially, sothatindex search

performance isconstant and doesn't depend on a query.

</para>

</sect2>

<sect2>

<title>Parameters</title>

<para>

<literal>bloom</> indexes accept following parameters in <literal>WITH</>

<literal>bloom</> indexes accept the following parameters in the

<literal>WITH</>

clause.

</para>

<title>Examples</title>

<para>

Example of index definition is given below.

An example of an index definition is given below.

</para>

<programlisting>

</programlisting>

<para>

Here, wecreatebloom index with signature length 80 bits and attributes

i1,i2mapped to 2 bits, attribute i3 - to 4 bits.

Here, wecreated abloom index withasignature lengthof80 bits,

and attributes i1 andi2 mapped to 2 bits,andattribute i3 to 4 bits.

</para>

<para>

Exampleof index definition and usage is given below.

Here is a fuller exampleof index definition and usage:

</para>

<programlisting>

</programlisting>

<para>

Btree index will be not used for this query.

A btree index will be not used for this query.

</para>

<programlisting>

<title>Opclass interface</title>

<para>

Bloom opclass interface is simple. It requires 1 supporting function:

hash function for indexing datatype.And it provides 1 search operator:

equality operator. The example below shows <literal>opclass</> definition

for <literal>text</> datatype.

TheBloom opclass interface is simple. It requires 1 supporting function:

ahash function fortheindexing datatype.It provides 1 search operator:

theequality operator. The example below shows <literal>opclass</>

definitionfor <literal>text</> datatype.

</para>

<programlisting>

<itemizedlist>

<listitem>

<para>

For now, only opclasses for <literal>int4</>, <literal>text</>comes

withcontrib. However, users may define more of them.

For now, only opclasses for <literal>int4</>, <literal>text</>come

withthe module. However, users may define more of them.

</para>

</listitem>

<listitem>

<para>

Only <literal>=</literal> operator is supported for searchnow. But it's

possible to add supportof arrays with contains and intersection

operations in future.

Onlythe<literal>=</literal> operator is supported for searchat the

moment. But it'spossible to add supportfor arrays with contains and

intersectionoperations in the future.

</para>

</listitem>

</itemizedlist>

<title>Authors</title>

<para>

Teodor Sigaev <email>teodor@postgrespro.ru</email>, Postgres Professional, Moscow, Russia

Teodor Sigaev <email>teodor@postgrespro.ru</email>,

Postgres Professional, Moscow, Russia

</para>

<para>

Alexander Korotkov <email>a.korotkov@postgrespro.ru</email>, Postgres Professional, Moscow, Russia

Alexander Korotkov <email>a.korotkov@postgrespro.ru</email>,

Postgres Professional, Moscow, Russia

</para>

<para>

Oleg Bartunov <email>obartunov@postgrespro.ru</email>, Postgres Professional, Moscow, Russia

Oleg Bartunov <email>obartunov@postgrespro.ru</email>,

Postgres Professional, Moscow, Russia

</para>

</sect2>