doc/src/sgml/generic-wal.sgml

<!-- doc/src/sgml/generic-wal.sgml -->

<chapter id="generic-wal">
 <title>Generic WAL Records</title>

  <para>
   Although all built-in WAL-logged modules have their own types of WAL
   records, there is also a generic WAL record type, which describes changes
   to pages in a generic way. This is useful for extensions that provide
   custom access methods.
  </para>

  <para>
   In comparison with <link linkend="custom-rmgr">Custom WAL Resource
   Managers</link>, Generic WAL is simpler for an extension to implement and
   does not require the extension library to be loaded in order to apply the
   records.
  </para>

  <note>
   <para>
    Generic WAL records are ignored during <link
    linkend="logicaldecoding">Logical Decoding</link>. If logical decoding is
    required for your extension, consider a Custom WAL Resource Manager.
   </para>
  </note>

  <para>
   The API for constructing generic WAL records is defined in
   <filename>access/generic_xlog.h</filename> and implemented
   in <filename>access/transam/generic_xlog.c</filename>.
  </para>

  <para>
   To perform a WAL-logged data update using the generic WAL record
   facility, follow these steps:

   <orderedlist>
    <listitem>
     <para>
      <function>state = GenericXLogStart(relation)</function> &mdash; start
      construction of a generic WAL record for the given relation.
     </para>
    </listitem>

    <listitem>
     <para>
      <function>page = GenericXLogRegisterBuffer(state, buffer, flags)</function>
      &mdash; register a buffer to be modified within the current generic WAL
      record.  This function returns a pointer to a temporary copy of the
      buffer's page, where modifications should be made.  (Do not modify the
      buffer's contents directly.)  The third argument is a bit mask of flags
      applicable to the operation.  Currently the only such flag is
      <literal>GENERIC_XLOG_FULL_IMAGE</literal>, which indicates that a full-page
      image rather than a delta update should be included in the WAL record.
      Typically this flag would be set if the page is new or has been
      rewritten completely.
      <function>GenericXLogRegisterBuffer</function> can be repeated if the
      WAL-logged action needs to modify multiple pages.
     </para>
    </listitem>

    <listitem>
     <para>
      Apply modifications to the page images obtained in the previous step.
     </para>
    </listitem>

    <listitem>
     <para>
      <function>GenericXLogFinish(state)</function> &mdash; apply the changes to
      the buffers and emit the generic WAL record.
     </para>
    </listitem>
   </orderedlist>
  </para>

  <para>
   WAL record construction can be canceled between any of the above steps by
   calling <function>GenericXLogAbort(state)</function>.  This will discard all
   changes to the page image copies.
  </para>

  <para>
   Please note the following points when using the generic WAL record
   facility:

   <itemizedlist>
    <listitem>
     <para>
      No direct modifications of buffers are allowed!  All modifications must
      be done in copies acquired from <function>GenericXLogRegisterBuffer()</function>.
      In other words, code that makes generic WAL records should never call
      <function>BufferGetPage()</function> for itself.  However, it remains the
      caller's responsibility to pin/unpin and lock/unlock the buffers at
      appropriate times.  Exclusive lock must be held on each target buffer
      from before <function>GenericXLogRegisterBuffer()</function> until after
      <function>GenericXLogFinish()</function>.
     </para>
    </listitem>

    <listitem>
     <para>
      Registrations of buffers (step 2) and modifications of page images
      (step 3) can be mixed freely, i.e., both steps may be repeated in any
      sequence.  Keep in mind that buffers should be registered in the same
      order in which locks are to be obtained on them during replay.
     </para>
    </listitem>

    <listitem>
     <para>
      The maximum number of buffers that can be registered for a generic WAL
      record is <literal>MAX_GENERIC_XLOG_PAGES</literal>.  An error will be thrown
      if this limit is exceeded.
     </para>
    </listitem>

    <listitem>
     <para>
      Generic WAL assumes that the pages to be modified have standard
      layout, and in particular that there is no useful data between
      <structfield>pd_lower</structfield> and <structfield>pd_upper</structfield>.
     </para>
    </listitem>

    <listitem>
     <para>
      Since you are modifying copies of buffer
      pages, <function>GenericXLogStart()</function> does not start a critical
      section.  Thus, you can safely do memory allocation, error throwing,
      etc. between <function>GenericXLogStart()</function> and
      <function>GenericXLogFinish()</function>.  The only actual critical section is
      present inside <function>GenericXLogFinish()</function>.  There is no need to
      worry about calling  <function>GenericXLogAbort()</function> during an error
      exit, either.
     </para>
    </listitem>

    <listitem>
     <para>
      <function>GenericXLogFinish()</function> takes care of marking buffers dirty
      and setting their LSNs.  You do not need to do this explicitly.
     </para>
    </listitem>

    <listitem>
     <para>
      For unlogged relations, everything works the same except that no
      actual WAL record is emitted.  Thus, you typically do not need to do
      any explicit checks for unlogged relations.
     </para>
    </listitem>

    <listitem>
     <para>
      The generic WAL redo function will acquire exclusive locks to buffers
      in the same order as they were registered.  After redoing all changes,
      the locks will be released in the same order.
     </para>
    </listitem>

    <listitem>
     <para>
      If <literal>GENERIC_XLOG_FULL_IMAGE</literal> is not specified for a
      registered buffer, the generic WAL record contains a delta between
      the old and the new page images.  This delta is based on byte-by-byte
      comparison.  This is not very compact for the case of moving data
      within a page, and might be improved in the future.
     </para>
    </listitem>
   </itemizedlist>
  </para>
</chapter>