1999-07-22 17:09:15 +02:00
|
|
|
<!--
|
2010-04-03 09:23:02 +02:00
|
|
|
$PostgreSQL: pgsql/doc/src/sgml/ref/declare.sgml,v 1.48 2010/04/03 07:23:00 petere Exp $
|
2001-12-08 04:24:40 +01:00
|
|
|
PostgreSQL documentation
|
1999-07-22 17:09:15 +02:00
|
|
|
-->
|
|
|
|
|
1999-07-06 19:16:42 +02:00
|
|
|
<refentry id="SQL-DECLARE">
|
|
|
|
<refmeta>
|
2010-04-03 09:23:02 +02:00
|
|
|
<refentrytitle>DECLARE</refentrytitle>
|
2008-11-14 11:22:48 +01:00
|
|
|
<manvolnum>7</manvolnum>
|
1999-07-06 19:16:42 +02:00
|
|
|
<refmiscinfo>SQL - Language Statements</refmiscinfo>
|
|
|
|
</refmeta>
|
2003-05-04 04:23:16 +02:00
|
|
|
|
1999-07-06 19:16:42 +02:00
|
|
|
<refnamediv>
|
2003-05-04 04:23:16 +02:00
|
|
|
<refname>DECLARE</refname>
|
|
|
|
<refpurpose>define a cursor</refpurpose>
|
1998-12-29 03:24:47 +01:00
|
|
|
</refnamediv>
|
2003-05-04 04:23:16 +02:00
|
|
|
|
2003-08-31 19:32:24 +02:00
|
|
|
<indexterm zone="sql-declare">
|
|
|
|
<primary>DECLARE</primary>
|
|
|
|
</indexterm>
|
|
|
|
|
2004-06-17 14:41:02 +02:00
|
|
|
<indexterm zone="sql-declare">
|
|
|
|
<primary>cursor</primary>
|
|
|
|
<secondary>DECLARE</secondary>
|
|
|
|
</indexterm>
|
|
|
|
|
1999-07-06 19:16:42 +02:00
|
|
|
<refsynopsisdiv>
|
2003-05-04 04:23:16 +02:00
|
|
|
<synopsis>
|
2003-09-22 02:16:58 +02:00
|
|
|
DECLARE <replaceable class="parameter">name</replaceable> [ BINARY ] [ INSENSITIVE ] [ [ NO ] SCROLL ]
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
CURSOR [ { WITH | WITHOUT } HOLD ] FOR <replaceable class="parameter">query</replaceable>
|
2003-05-04 04:23:16 +02:00
|
|
|
</synopsis>
|
1999-07-06 19:16:42 +02:00
|
|
|
</refsynopsisdiv>
|
|
|
|
|
2003-05-04 04:23:16 +02:00
|
|
|
<refsect1>
|
|
|
|
<title>Description</title>
|
|
|
|
|
1999-07-06 19:16:42 +02:00
|
|
|
<para>
|
1999-07-22 17:09:15 +02:00
|
|
|
<command>DECLARE</command> allows a user to create cursors, which
|
|
|
|
can be used to retrieve
|
2007-06-11 03:16:30 +02:00
|
|
|
a small number of rows at a time out of a larger query.
|
|
|
|
After the cursor is created, rows are fetched from it using
|
2010-04-03 09:23:02 +02:00
|
|
|
<xref linkend="sql-fetch">.
|
1999-07-06 19:16:42 +02:00
|
|
|
</para>
|
2009-04-10 19:56:21 +02:00
|
|
|
|
|
|
|
<note>
|
|
|
|
<para>
|
|
|
|
This page describes usage of cursors at the SQL command level.
|
|
|
|
If you are trying to use cursors inside a <application>PL/pgSQL</>
|
|
|
|
function, the rules are different —
|
|
|
|
see <xref linkend="plpgsql-cursors">.
|
|
|
|
</para>
|
|
|
|
</note>
|
2003-05-04 04:23:16 +02:00
|
|
|
</refsect1>
|
|
|
|
|
|
|
|
<refsect1>
|
|
|
|
<title>Parameters</title>
|
|
|
|
|
|
|
|
<variablelist>
|
|
|
|
<varlistentry>
|
2003-09-22 02:16:58 +02:00
|
|
|
<term><replaceable class="parameter">name</replaceable></term>
|
2003-05-04 04:23:16 +02:00
|
|
|
<listitem>
|
|
|
|
<para>
|
2003-09-12 02:12:47 +02:00
|
|
|
The name of the cursor to be created.
|
2003-05-04 04:23:16 +02:00
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</varlistentry>
|
|
|
|
|
|
|
|
<varlistentry>
|
|
|
|
<term><literal>BINARY</literal></term>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
Causes the cursor to return data in binary rather than in text format.
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</varlistentry>
|
|
|
|
|
|
|
|
<varlistentry>
|
|
|
|
<term><literal>INSENSITIVE</literal></term>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
Indicates that data retrieved from the cursor should be
|
2007-06-11 03:16:30 +02:00
|
|
|
unaffected by updates to the table(s) underlying the cursor that occur
|
|
|
|
after the cursor is created. In <productname>PostgreSQL</productname>,
|
|
|
|
this is the default behavior; so this key word has no
|
|
|
|
effect and is only accepted for compatibility with the SQL standard.
|
2003-05-04 04:23:16 +02:00
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</varlistentry>
|
|
|
|
|
|
|
|
<varlistentry>
|
|
|
|
<term><literal>SCROLL</literal></term>
|
|
|
|
<term><literal>NO SCROLL</literal></term>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
Update reference documentation on may/can/might:
Standard English uses "may", "can", and "might" in different ways:
may - permission, "You may borrow my rake."
can - ability, "I can lift that log."
might - possibility, "It might rain today."
Unfortunately, in conversational English, their use is often mixed, as
in, "You may use this variable to do X", when in fact, "can" is a better
choice. Similarly, "It may crash" is better stated, "It might crash".
2007-02-01 00:26:05 +01:00
|
|
|
<literal>SCROLL</literal> specifies that the cursor can be used
|
2003-05-04 04:23:16 +02:00
|
|
|
to retrieve rows in a nonsequential fashion (e.g.,
|
|
|
|
backward). Depending upon the complexity of the query's
|
Update reference documentation on may/can/might:
Standard English uses "may", "can", and "might" in different ways:
may - permission, "You may borrow my rake."
can - ability, "I can lift that log."
might - possibility, "It might rain today."
Unfortunately, in conversational English, their use is often mixed, as
in, "You may use this variable to do X", when in fact, "can" is a better
choice. Similarly, "It may crash" is better stated, "It might crash".
2007-02-01 00:26:05 +01:00
|
|
|
execution plan, specifying <literal>SCROLL</literal> might impose
|
2003-05-04 04:23:16 +02:00
|
|
|
a performance penalty on the query's execution time.
|
|
|
|
<literal>NO SCROLL</literal> specifies that the cursor cannot be
|
2006-02-12 20:02:15 +01:00
|
|
|
used to retrieve rows in a nonsequential fashion. The default is to
|
2006-02-12 21:31:58 +01:00
|
|
|
allow scrolling in some cases; this is not the same as specifying
|
|
|
|
<literal>SCROLL</literal>. See <xref linkend="sql-declare-notes"
|
|
|
|
endterm="sql-declare-notes-title"> for details.
|
2003-05-04 04:23:16 +02:00
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</varlistentry>
|
|
|
|
|
|
|
|
<varlistentry>
|
|
|
|
<term><literal>WITH HOLD</literal></term>
|
|
|
|
<term><literal>WITHOUT HOLD</literal></term>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
Update reference documentation on may/can/might:
Standard English uses "may", "can", and "might" in different ways:
may - permission, "You may borrow my rake."
can - ability, "I can lift that log."
might - possibility, "It might rain today."
Unfortunately, in conversational English, their use is often mixed, as
in, "You may use this variable to do X", when in fact, "can" is a better
choice. Similarly, "It may crash" is better stated, "It might crash".
2007-02-01 00:26:05 +01:00
|
|
|
<literal>WITH HOLD</literal> specifies that the cursor can
|
2003-05-04 04:23:16 +02:00
|
|
|
continue to be used after the transaction that created it
|
|
|
|
successfully commits. <literal>WITHOUT HOLD</literal> specifies
|
|
|
|
that the cursor cannot be used outside of the transaction that
|
|
|
|
created it. If neither <literal>WITHOUT HOLD</literal> nor
|
|
|
|
<literal>WITH HOLD</literal> is specified, <literal>WITHOUT
|
|
|
|
HOLD</literal> is the default.
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</varlistentry>
|
|
|
|
|
|
|
|
<varlistentry>
|
|
|
|
<term><replaceable class="parameter">query</replaceable></term>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
2010-04-03 09:23:02 +02:00
|
|
|
A <xref linkend="sql-select"> or
|
|
|
|
<xref linkend="sql-values"> command
|
2006-09-18 21:54:01 +02:00
|
|
|
which will provide the rows to be returned by the cursor.
|
2003-05-04 04:23:16 +02:00
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</varlistentry>
|
|
|
|
</variablelist>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
The key words <literal>BINARY</literal>,
|
Update reference documentation on may/can/might:
Standard English uses "may", "can", and "might" in different ways:
may - permission, "You may borrow my rake."
can - ability, "I can lift that log."
might - possibility, "It might rain today."
Unfortunately, in conversational English, their use is often mixed, as
in, "You may use this variable to do X", when in fact, "can" is a better
choice. Similarly, "It may crash" is better stated, "It might crash".
2007-02-01 00:26:05 +01:00
|
|
|
<literal>INSENSITIVE</literal>, and <literal>SCROLL</literal> can
|
2003-05-04 04:23:16 +02:00
|
|
|
appear in any order.
|
1999-07-06 19:16:42 +02:00
|
|
|
</para>
|
2003-05-04 04:23:16 +02:00
|
|
|
</refsect1>
|
1999-07-06 19:16:42 +02:00
|
|
|
|
2006-02-12 21:31:58 +01:00
|
|
|
<refsect1 id="sql-declare-notes">
|
|
|
|
<title id="sql-declare-notes-title">Notes</title>
|
1999-07-22 17:09:15 +02:00
|
|
|
|
2007-06-11 03:16:30 +02:00
|
|
|
<para>
|
|
|
|
Normal cursors return data in text format, the same as a
|
|
|
|
<command>SELECT</> would produce. The <literal>BINARY</> option
|
|
|
|
specifies that the cursor should return data in binary format.
|
|
|
|
This reduces conversion effort for both the server and client,
|
|
|
|
at the cost of more programmer effort to deal with platform-dependent
|
|
|
|
binary data formats.
|
|
|
|
As an example, if a query returns a value of one from an integer column,
|
|
|
|
you would get a string of <literal>1</> with a default cursor,
|
|
|
|
whereas with a binary cursor you would get
|
|
|
|
a 4-byte field containing the internal representation of the value
|
|
|
|
(in big-endian byte order).
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
Binary cursors should be used carefully. Many applications,
|
|
|
|
including <application>psql</application>, are not prepared to
|
|
|
|
handle binary cursors and expect data to come back in the text
|
|
|
|
format.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<note>
|
|
|
|
<para>
|
|
|
|
When the client application uses the <quote>extended query</> protocol
|
|
|
|
to issue a <command>FETCH</> command, the Bind protocol message
|
|
|
|
specifies whether data is to be retrieved in text or binary format.
|
|
|
|
This choice overrides the way that the cursor is defined. The concept
|
|
|
|
of a binary cursor as such is thus obsolete when using extended query
|
|
|
|
protocol — any cursor can be treated as either text or binary.
|
|
|
|
</para>
|
|
|
|
</note>
|
|
|
|
|
1999-07-06 19:16:42 +02:00
|
|
|
<para>
|
2003-09-12 02:12:47 +02:00
|
|
|
Unless <literal>WITH HOLD</literal> is specified, the cursor
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
created by this command can only be used within the current
|
2003-09-12 02:12:47 +02:00
|
|
|
transaction. Thus, <command>DECLARE</> without <literal>WITH
|
|
|
|
HOLD</literal> is useless outside a transaction block: the cursor would
|
|
|
|
survive only to the completion of the statement. Therefore
|
2007-10-25 01:27:08 +02:00
|
|
|
<productname>PostgreSQL</productname> reports an error if such a
|
2003-09-12 02:12:47 +02:00
|
|
|
command is used outside a transaction block.
|
|
|
|
Use
|
2010-04-03 09:23:02 +02:00
|
|
|
<xref linkend="sql-begin"> and
|
|
|
|
<xref linkend="sql-commit">
|
|
|
|
(or <xref linkend="sql-rollback">)
|
1999-07-22 17:09:15 +02:00
|
|
|
to define a transaction block.
|
1999-07-06 19:16:42 +02:00
|
|
|
</para>
|
1999-07-22 17:09:15 +02:00
|
|
|
|
2003-03-10 04:53:52 +01:00
|
|
|
<para>
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
If <literal>WITH HOLD</literal> is specified and the transaction
|
2003-05-04 04:23:16 +02:00
|
|
|
that created the cursor successfully commits, the cursor can
|
|
|
|
continue to be accessed by subsequent transactions in the same
|
|
|
|
session. (But if the creating transaction is aborted, the cursor
|
|
|
|
is removed.) A cursor created with <literal>WITH HOLD</literal>
|
|
|
|
is closed when an explicit <command>CLOSE</command> command is
|
|
|
|
issued on it, or the session ends. In the current implementation,
|
|
|
|
the rows represented by a held cursor are copied into a temporary
|
|
|
|
file or memory area so that they remain available for subsequent
|
|
|
|
transactions.
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
</para>
|
|
|
|
|
2007-06-11 03:16:30 +02:00
|
|
|
<para>
|
|
|
|
<literal>WITH HOLD</literal> may not be specified when the query
|
|
|
|
includes <literal>FOR UPDATE</> or <literal>FOR SHARE</>.
|
|
|
|
</para>
|
|
|
|
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
<para>
|
|
|
|
The <literal>SCROLL</> option should be specified when defining a
|
|
|
|
cursor that will be used to fetch backwards. This is required by
|
2003-05-04 04:23:16 +02:00
|
|
|
the SQL standard. However, for compatibility with earlier
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
versions, <productname>PostgreSQL</productname> will allow
|
|
|
|
backward fetches without <literal>SCROLL</>, if the cursor's query
|
|
|
|
plan is simple enough that no extra overhead is needed to support
|
|
|
|
it. However, application developers are advised not to rely on
|
|
|
|
using backward fetches from a cursor that has not been created
|
2003-05-04 04:23:16 +02:00
|
|
|
with <literal>SCROLL</literal>. If <literal>NO SCROLL</> is
|
|
|
|
specified, then backward fetches are disallowed in any case.
|
2003-03-10 04:53:52 +01:00
|
|
|
</para>
|
|
|
|
|
2008-11-16 18:34:28 +01:00
|
|
|
<para>
|
|
|
|
Backward fetches are also disallowed when the query
|
|
|
|
includes <literal>FOR UPDATE</> or <literal>FOR SHARE</>; therefore
|
|
|
|
<literal>SCROLL</literal> may not be specified in this case.
|
|
|
|
</para>
|
|
|
|
|
2009-06-10 21:21:37 +02:00
|
|
|
<caution>
|
|
|
|
<para>
|
|
|
|
Scrollable and <literal>WITH HOLD</literal> cursors may give unexpected
|
|
|
|
results if they invoke any volatile functions (see <xref
|
|
|
|
linkend="xfunc-volatility">). When a previously fetched row is
|
|
|
|
re-fetched, the functions might be re-executed, perhaps leading to
|
|
|
|
results different from the first time. One workaround for such cases
|
|
|
|
is to declare the cursor <literal>WITH HOLD</literal> and commit the
|
|
|
|
transaction before reading any rows from it. This will force the
|
|
|
|
entire output of the cursor to be materialized in temporary storage,
|
|
|
|
so that volatile functions are executed exactly once for each row.
|
|
|
|
</para>
|
|
|
|
</caution>
|
|
|
|
|
2007-06-11 03:16:30 +02:00
|
|
|
<para>
|
|
|
|
If the cursor's query includes <literal>FOR UPDATE</> or <literal>FOR
|
|
|
|
SHARE</>, then returned rows are locked at the time they are first
|
|
|
|
fetched, in the same way as for a regular
|
2010-04-03 09:23:02 +02:00
|
|
|
<xref linkend="sql-select"> command with
|
2007-06-11 03:16:30 +02:00
|
|
|
these options.
|
|
|
|
In addition, the returned rows will be the most up-to-date versions;
|
|
|
|
therefore these options provide the equivalent of what the SQL standard
|
2008-11-16 18:34:28 +01:00
|
|
|
calls a <quote>sensitive cursor</>. (Specifying <literal>INSENSITIVE</>
|
|
|
|
together with <literal>FOR UPDATE</> or <literal>FOR SHARE</> is an error.)
|
2007-06-11 03:16:30 +02:00
|
|
|
</para>
|
|
|
|
|
2008-11-16 18:34:28 +01:00
|
|
|
<caution>
|
|
|
|
<para>
|
|
|
|
It is generally recommended to use <literal>FOR UPDATE</> if the cursor
|
|
|
|
is intended to be used with <command>UPDATE ... WHERE CURRENT OF</> or
|
|
|
|
<command>DELETE ... WHERE CURRENT OF</>. Using <literal>FOR UPDATE</>
|
|
|
|
prevents other sessions from changing the rows between the time they are
|
|
|
|
fetched and the time they are updated. Without <literal>FOR UPDATE</>,
|
|
|
|
a subsequent <literal>WHERE CURRENT OF</> command will have no effect if
|
|
|
|
the row was changed since the cursor was created.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
Another reason to use <literal>FOR UPDATE</> is that without it, a
|
|
|
|
subsequent <literal>WHERE CURRENT OF</> might fail if the cursor query
|
|
|
|
does not meet the SQL standard's rules for being <quote>simply
|
|
|
|
updatable</> (in particular, the cursor must reference just one table
|
|
|
|
and not use grouping or <literal>ORDER BY</>). Cursors
|
|
|
|
that are not simply updatable might work, or might not, depending on plan
|
|
|
|
choice details; so in the worst case, an application might work in testing
|
|
|
|
and then fail in production.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
The main reason not to use <literal>FOR UPDATE</> with <literal>WHERE
|
|
|
|
CURRENT OF</> is if you need the cursor to be scrollable, or to be
|
|
|
|
insensitive to the subsequent updates (that is, continue to show the old
|
|
|
|
data). If this is a requirement, pay close heed to the caveats shown
|
|
|
|
above.
|
|
|
|
</para>
|
|
|
|
</caution>
|
2007-10-25 01:27:08 +02:00
|
|
|
|
1999-07-06 19:16:42 +02:00
|
|
|
<para>
|
2003-05-04 04:23:16 +02:00
|
|
|
The SQL standard only makes provisions for cursors in embedded
|
|
|
|
<acronym>SQL</acronym>. The <productname>PostgreSQL</productname>
|
|
|
|
server does not implement an <command>OPEN</command> statement for
|
|
|
|
cursors; a cursor is considered to be open when it is declared.
|
|
|
|
However, <application>ECPG</application>, the embedded SQL
|
|
|
|
preprocessor for <productname>PostgreSQL</productname>, supports
|
|
|
|
the standard SQL cursor conventions, including those involving
|
|
|
|
<command>DECLARE</command> and <command>OPEN</command> statements.
|
1999-07-06 19:16:42 +02:00
|
|
|
</para>
|
2006-01-18 07:49:30 +01:00
|
|
|
|
|
|
|
<para>
|
2006-02-26 04:20:46 +01:00
|
|
|
You can see all available cursors by querying the <link
|
|
|
|
linkend="view-pg-cursors"><structname>pg_cursors</structname></link>
|
|
|
|
system view.
|
2006-01-18 07:49:30 +01:00
|
|
|
</para>
|
1998-09-01 17:53:09 +02:00
|
|
|
</refsect1>
|
1999-07-06 19:16:42 +02:00
|
|
|
|
2003-05-04 04:23:16 +02:00
|
|
|
<refsect1>
|
|
|
|
<title>Examples</title>
|
|
|
|
|
1999-07-06 19:16:42 +02:00
|
|
|
<para>
|
1998-09-01 17:53:09 +02:00
|
|
|
To declare a cursor:
|
2003-05-04 04:23:16 +02:00
|
|
|
<programlisting>
|
|
|
|
DECLARE liahona CURSOR FOR SELECT * FROM films;
|
|
|
|
</programlisting>
|
2010-04-03 09:23:02 +02:00
|
|
|
See <xref linkend="sql-fetch"> for more
|
2003-05-04 04:23:16 +02:00
|
|
|
examples of cursor usage.
|
1999-07-06 19:16:42 +02:00
|
|
|
</para>
|
|
|
|
</refsect1>
|
1998-09-01 17:53:09 +02:00
|
|
|
|
2003-05-04 04:23:16 +02:00
|
|
|
<refsect1>
|
|
|
|
<title>Compatibility</title>
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
|
2003-05-04 04:23:16 +02:00
|
|
|
<para>
|
2007-10-25 01:27:08 +02:00
|
|
|
The SQL standard says that it is implementation-dependent whether cursors
|
|
|
|
are sensitive to concurrent updates of the underlying data by default. In
|
2007-06-11 03:16:30 +02:00
|
|
|
<productname>PostgreSQL</productname>, cursors are insensitive by default,
|
2007-10-25 01:27:08 +02:00
|
|
|
and can be made sensitive by specifying <literal>FOR UPDATE</>. Other
|
|
|
|
products may work differently.
|
2003-05-04 04:23:16 +02:00
|
|
|
</para>
|
This patch implements holdable cursors, following the proposal
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
2003-03-27 17:51:29 +01:00
|
|
|
|
2003-05-04 04:23:16 +02:00
|
|
|
<para>
|
2007-06-11 03:16:30 +02:00
|
|
|
The SQL standard allows cursors only in embedded
|
|
|
|
<acronym>SQL</acronym> and in modules. <productname>PostgreSQL</>
|
|
|
|
permits cursors to be used interactively.
|
2003-05-04 04:23:16 +02:00
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
Binary cursors are a <productname>PostgreSQL</productname>
|
|
|
|
extension.
|
|
|
|
</para>
|
1998-12-29 03:24:47 +01:00
|
|
|
</refsect1>
|
2005-01-04 01:39:53 +01:00
|
|
|
|
|
|
|
<refsect1>
|
|
|
|
<title>See Also</title>
|
|
|
|
|
|
|
|
<simplelist type="inline">
|
2010-04-03 09:23:02 +02:00
|
|
|
<member><xref linkend="sql-close"></member>
|
|
|
|
<member><xref linkend="sql-fetch"></member>
|
|
|
|
<member><xref linkend="sql-move"></member>
|
2005-01-04 01:39:53 +01:00
|
|
|
</simplelist>
|
|
|
|
</refsect1>
|
1999-07-06 19:16:42 +02:00
|
|
|
</refentry>
|