Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Skip to content

Commit 725d52d

Browse files
tglsfdctglsfdc
committed
Create the system catalog infrastructure needed for KNNGIST.
This commit adds columns amoppurpose and amopsortfamily to pg_amop, and column amcanorderbyop to pg_am. For the moment all the entries in amcanorderbyop are "false", since the underlying support isn't there yet. Also, extend the CREATE OPERATOR CLASS/ALTER OPERATOR FAMILY commands with [ FOR SEARCH | FOR ORDER BY sort_operator_family ] clauses to allow the new columns of pg_amop to be populated, and create pg_dump support for dumping that information. I also added some documentation, although it's perhaps a bit premature given that the feature doesn't do anything useful yet. Teodor Sigaev, Robert Haas, Tom Lane
1 parent 4fc09ad commit 725d52d

File tree

21 files changed

+913
-475
lines changed

21 files changed

+913
-475
lines changed

‎doc/src/sgml/catalogs.sgml

Lines changed: 66 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -423,7 +423,16 @@
423423
<entry><structfield>amcanorder</structfield></entry>
424424
<entry><type>bool</type></entry>
425425
<entry></entry>
426-
<entry>Does the access method support ordered scans?</entry>
426+
<entry>Does the access method support ordered scans sorted by the
427+
indexed column's value?</entry>
428+
</row>
429+
430+
<row>
431+
<entry><structfield>amcanorderbyop</structfield></entry>
432+
<entry><type>bool</type></entry>
433+
<entry></entry>
434+
<entry>Does the access method support ordered scans sorted by the result
435+
of an operator on the indexed column?</entry>
427436
</row>
428437

429438
<row>
@@ -598,9 +607,13 @@
598607
<para>
599608
The catalog <structname>pg_amop</structname> stores information about
600609
operators associated with access method operator families. There is one
601-
row for each operator that is a member of an operator family. An operator
610+
row for each operator that is a member of an operator family. A family
611+
member can be either a <firstterm>search</> operator or an
612+
<firstterm>ordering</> operator. An operator
602613
can appear in more than one family, but cannot appear in more than one
603-
position within a family.
614+
search position nor more than one ordering position within a family.
615+
(It is allowed, though unlikely, for an operator to be used for both
616+
search and ordering purposes.)
604617
</para>
605618

606619
<table>
@@ -645,6 +658,14 @@
645658
<entry>Operator strategy number</entry>
646659
</row>
647660

661+
<row>
662+
<entry><structfield>amoppurpose</structfield></entry>
663+
<entry><type>char</type></entry>
664+
<entry></entry>
665+
<entry>Operator purpose, either <literal>s</> for search or
666+
<literal>o</> for ordering</entry>
667+
</row>
668+
648669
<row>
649670
<entry><structfield>amopopr</structfield></entry>
650671
<entry><type>oid</type></entry>
@@ -659,10 +680,52 @@
659680
<entry>Index access method operator family is for</entry>
660681
</row>
661682

683+
<row>
684+
<entry><structfield>amopsortfamily</structfield></entry>
685+
<entry><type>oid</type></entry>
686+
<entry><literal><link linkend="catalog-pg-opfamily"><structname>pg_opfamily</structname></link>.oid</literal></entry>
687+
<entry>The btree operator family this entry sorts according to, if an
688+
ordering operator; zero if a search operator</entry>
689+
</row>
690+
662691
</tbody>
663692
</tgroup>
664693
</table>
665694

695+
<para>
696+
A <quote>search</> operator entry indicates that an index of this operator
697+
family can be searched to find all rows satisfying
698+
<literal>WHERE</>
699+
<replaceable>indexed_column</>
700+
<replaceable>operator</>
701+
<replaceable>constant</>.
702+
Obviously, such an operator must return boolean, and its left-hand input
703+
type must match the index's column data type.
704+
</para>
705+
706+
<para>
707+
An <quote>ordering</> operator entry indicates that an index of this
708+
operator family can be scanned to return rows in the order represented by
709+
<literal>ORDER BY</>
710+
<replaceable>indexed_column</>
711+
<replaceable>operator</>
712+
<replaceable>constant</>.
713+
Such an operator could return any sortable data type, though again
714+
its left-hand input type must match the index's column data type.
715+
The exact semantics of the <literal>ORDER BY</> are specified by the
716+
<structfield>amopsortfamily</structfield> column, which must reference
717+
a btree operator family for the operator's result type.
718+
</para>
719+
720+
<note>
721+
<para>
722+
At present, it's assumed that the sort order for an ordering operator
723+
is the default for the referenced opfamily, i.e., <literal>ASC NULLS
724+
LAST</>. This might someday be relaxed by adding additional columns
725+
to specify sort options explicitly.
726+
</para>
727+
</note>
728+
666729
<para>
667730
An entry's <structfield>amopmethod</> must match the
668731
<structname>opfmethod</> of its containing operator family (including

‎doc/src/sgml/ref/alter_opfamily.sgml

Lines changed: 17 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -22,7 +22,7 @@ PostgreSQL documentation
2222
<refsynopsisdiv>
2323
<synopsis>
2424
ALTER OPERATOR FAMILY <replaceable>name</replaceable> USING <replaceable class="parameter">index_method</replaceable> ADD
25-
{ OPERATOR <replaceable class="parameter">strategy_number</replaceable> <replaceable class="parameter">operator_name</replaceable> ( <replaceable class="parameter">op_type</replaceable>, <replaceable class="parameter">op_type</replaceable> )
25+
{ OPERATOR <replaceable class="parameter">strategy_number</replaceable> <replaceable class="parameter">operator_name</replaceable> ( <replaceable class="parameter">op_type</replaceable>, <replaceable class="parameter">op_type</replaceable> ) [ FOR SEARCH | FOR ORDER BY <replaceable class="parameter">sort_family_name</replaceable> ]
2626
| FUNCTION <replaceable class="parameter">support_number</replaceable> [ ( <replaceable class="parameter">op_type</replaceable> [ , <replaceable class="parameter">op_type</replaceable> ] ) ] <replaceable class="parameter">function_name</replaceable> ( <replaceable class="parameter">argument_type</replaceable> [, ...] )
2727
} [, ... ]
2828
ALTER OPERATOR FAMILY <replaceable>name</replaceable> USING <replaceable class="parameter">index_method</replaceable> DROP
@@ -154,6 +154,22 @@ ALTER OPERATOR FAMILY <replaceable>name</replaceable> USING <replaceable class="
154154
</listitem>
155155
</varlistentry>
156156

157+
<varlistentry>
158+
<term><replaceable class="parameter">sort_family_name</replaceable></term>
159+
<listitem>
160+
<para>
161+
The name (optionally schema-qualified) of an existing btree operator
162+
family that describes the sort ordering associated with an ordering
163+
operator.
164+
</para>
165+
166+
<para>
167+
If neither <literal>FOR SEARCH</> nor <literal>FOR ORDER BY</> is
168+
specified, <literal>FOR SEARCH</> is the default.
169+
</para>
170+
</listitem>
171+
</varlistentry>
172+
157173
<varlistentry>
158174
<term><replaceable class="parameter">support_number</replaceable></term>
159175
<listitem>

‎doc/src/sgml/ref/create_opclass.sgml

Lines changed: 17 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -23,7 +23,7 @@ PostgreSQL documentation
2323
<synopsis>
2424
CREATE OPERATOR CLASS <replaceable class="parameter">name</replaceable> [ DEFAULT ] FOR TYPE <replaceable class="parameter">data_type</replaceable>
2525
USING <replaceable class="parameter">index_method</replaceable> [ FAMILY <replaceable class="parameter">family_name</replaceable> ] AS
26-
{ OPERATOR <replaceable class="parameter">strategy_number</replaceable> <replaceable class="parameter">operator_name</replaceable> [ ( <replaceable class="parameter">op_type</replaceable>, <replaceable class="parameter">op_type</replaceable> ) ]
26+
{ OPERATOR <replaceable class="parameter">strategy_number</replaceable> <replaceable class="parameter">operator_name</replaceable> [ ( <replaceable class="parameter">op_type</replaceable>, <replaceable class="parameter">op_type</replaceable> ) ] [ FOR SEARCH | FOR ORDER BY <replaceable class="parameter">sort_family_name</replaceable> ]
2727
| FUNCTION <replaceable class="parameter">support_number</replaceable> [ ( <replaceable class="parameter">op_type</replaceable> [ , <replaceable class="parameter">op_type</replaceable> ] ) ] <replaceable class="parameter">function_name</replaceable> ( <replaceable class="parameter">argument_type</replaceable> [, ...] )
2828
| STORAGE <replaceable class="parameter">storage_type</replaceable>
2929
} [, ... ]
@@ -180,6 +180,22 @@ CREATE OPERATOR CLASS <replaceable class="parameter">name</replaceable> [ DEFAUL
180180
</listitem>
181181
</varlistentry>
182182

183+
<varlistentry>
184+
<term><replaceable class="parameter">sort_family_name</replaceable></term>
185+
<listitem>
186+
<para>
187+
The name (optionally schema-qualified) of an existing btree operator
188+
family that describes the sort ordering associated with an ordering
189+
operator.
190+
</para>
191+
192+
<para>
193+
If neither <literal>FOR SEARCH</> nor <literal>FOR ORDER BY</> is
194+
specified, <literal>FOR SEARCH</> is the default.
195+
</para>
196+
</listitem>
197+
</varlistentry>
198+
183199
<varlistentry>
184200
<term><replaceable class="parameter">support_number</replaceable></term>
185201
<listitem>

‎doc/src/sgml/xindex.sgml

Lines changed: 62 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -276,10 +276,13 @@
276276
</table>
277277

278278
<para>
279-
Notice that all strategy operators return Boolean values. In
280-
practice, all operators defined as index method strategies must
279+
Notice that all the operators listed above return Boolean values. In
280+
practice, all operators defined as index method search operators must
281281
return type <type>boolean</type>, since they must appear at the top
282282
level of a <literal>WHERE</> clause to be used with an index.
283+
(Some index access methods also support <firstterm>ordering operators</>,
284+
which typically don't return Boolean values; that feature is discussed
285+
in <xref linkend="xindex-ordering-ops">.)
283286
</para>
284287
</sect2>
285288

@@ -464,7 +467,7 @@
464467
</table>
465468

466469
<para>
467-
Unlike strategy operators, support functions return whichever data
470+
Unlike search operators, support functions return whichever data
468471
type the particular index method expects; for example in the case
469472
of the comparison function for B-trees, a signed integer. The number
470473
and types of the arguments to each support function are likewise
@@ -921,6 +924,62 @@ ALTER OPERATOR FAMILY integer_ops USING btree ADD
921924
</para>
922925
</sect2>
923926

927+
<sect2 id="xindex-ordering-ops">
928+
<title>Ordering Operators</title>
929+
930+
<para>
931+
Some index access methods (currently, only GiST) support the concept of
932+
<firstterm>ordering operators</>. What we have been discussing so far
933+
are <firstterm>search operators</>. A search operator is one for which
934+
the index can be searched to find all rows satisfying
935+
<literal>WHERE</>
936+
<replaceable>indexed_column</>
937+
<replaceable>operator</>
938+
<replaceable>constant</>.
939+
Note that nothing is promised about the order in which the matching rows
940+
will be returned. In contrast, an ordering operator does not restrict the
941+
set of rows that can be returned, but instead determines their order.
942+
An ordering operator is one for which the index can be scanned to return
943+
rows in the order represented by
944+
<literal>ORDER BY</>
945+
<replaceable>indexed_column</>
946+
<replaceable>operator</>
947+
<replaceable>constant</>.
948+
The reason for defining ordering operators that way is that it supports
949+
nearest-neighbor searches, if the operator is one that measures distance.
950+
For example, a query like
951+
<programlisting><![CDATA[
952+
SELECT * FROM places ORDER BY location <-> point '(101,456)' LIMIT 10;
953+
]]>
954+
</programlisting>
955+
finds the ten places closest to a given target point. A GiST index
956+
on the location column can do this efficiently because
957+
<literal>&lt;-&gt;</> is an ordering operator.
958+
</para>
959+
960+
<para>
961+
While search operators have to return Boolean results, ordering operators
962+
usually return some other type, such as float or numeric for distances.
963+
This type is normally not the same as the data type being indexed.
964+
To avoid hard-wiring assumptions about the behavior of different data
965+
types, the definition of an ordering operator is required to name
966+
a B-tree operator family that specifies the sort ordering of the result
967+
data type. As was stated in the previous section, B-tree operator families
968+
define <productname>PostgreSQL</productname>'s notion of ordering, so
969+
this is a natural representation. Since the point <literal>&lt;-&gt;</>
970+
operator returns <type>float8</>, it could be specified in an operator
971+
class creation command like this:
972+
<programlisting><![CDATA[
973+
OPERATOR 15 <-> (point, point) FOR ORDER BY float_ops
974+
]]>
975+
</programlisting>
976+
where <literal>float_ops</> is the built-in operator family that includes
977+
operations on <type>float8</>. This declaration states that the index
978+
is able to return rows in order of increasing values of the
979+
<literal>&lt;-&gt;</> operator.
980+
</para>
981+
</sect2>
982+
924983
<sect2 id="xindex-opclass-features">
925984
<title>Special Features of Operator Classes</title>
926985

0 commit comments

Comments
 (0)