postgres/doc/src/sgml/tablesample-method.sgml

<!-- doc/src/sgml/tablesample-method.sgml -->

<chapter id="tablesample-method">
 <title>Writing a Table Sampling Method</title>

 <indexterm zone="tablesample-method">
  <primary>table sampling method</primary>
 </indexterm>

 <indexterm zone="tablesample-method">
  <primary><literal>TABLESAMPLE</literal> method</primary>
 </indexterm>

 <para>
  <productname>PostgreSQL</productname>'s implementation of the <literal>TABLESAMPLE</literal>
  clause supports custom table sampling methods, in addition to
  the <literal>BERNOULLI</literal> and <literal>SYSTEM</literal> methods that are required
  by the SQL standard.  The sampling method determines which rows of the
  table will be selected when the <literal>TABLESAMPLE</literal> clause is used.
 </para>

 <para>
  At the SQL level, a table sampling method is represented by a single SQL
  function, typically implemented in C, having the signature
<programlisting>
method_name(internal) RETURNS tsm_handler
</programlisting>
  The name of the function is the same method name appearing in the
  <literal>TABLESAMPLE</literal> clause.  The <type>internal</type> argument is a dummy
  (always having value zero) that simply serves to prevent this function from
  being called directly from a SQL command.
  The result of the function must be a palloc'd struct of
  type <type>TsmRoutine</type>, which contains pointers to support functions for
  the sampling method.  These support functions are plain C functions and
  are not visible or callable at the SQL level.  The support functions are
  described in <xref linkend="tablesample-support-functions"/>.
 </para>

 <para>
  In addition to function pointers, the <type>TsmRoutine</type> struct must
  provide these additional fields:
 </para>

 <variablelist>
  <varlistentry>
   <term><literal>List *parameterTypes</literal></term>
   <listitem>
    <para>
     This is an OID list containing the data type OIDs of the parameter(s)
     that will be accepted by the <literal>TABLESAMPLE</literal> clause when this
     sampling method is used.  For example, for the built-in methods, this
     list contains a single item with value <literal>FLOAT4OID</literal>, which
     represents the sampling percentage.  Custom sampling methods can have
     more or different parameters.
    </para>
   </listitem>
  </varlistentry>

  <varlistentry>
   <term><literal>bool repeatable_across_queries</literal></term>
   <listitem>
    <para>
     If <literal>true</literal>, the sampling method can deliver identical samples
     across successive queries, if the same parameters
     and <literal>REPEATABLE</literal> seed value are supplied each time and the
     table contents have not changed.  When this is <literal>false</literal>,
     the <literal>REPEATABLE</literal> clause is not accepted for use with the
     sampling method.
    </para>
   </listitem>
  </varlistentry>

  <varlistentry>
   <term><literal>bool repeatable_across_scans</literal></term>
   <listitem>
    <para>
     If <literal>true</literal>, the sampling method can deliver identical samples
     across successive scans in the same query (assuming unchanging
     parameters, seed value, and snapshot).
     When this is <literal>false</literal>, the planner will not select plans that
     would require scanning the sampled table more than once, since that
     might result in inconsistent query output.
    </para>
   </listitem>
  </varlistentry>
 </variablelist>

 <para>
  The <type>TsmRoutine</type> struct type is declared
  in <filename>src/include/access/tsmapi.h</filename>, which see for additional
  details.
 </para>

 <para>
  The table sampling methods included in the standard distribution are good
  references when trying to write your own.  Look into
  the <filename>src/backend/access/tablesample</filename> subdirectory of the source
  tree for the built-in sampling methods, and into the <filename>contrib</filename>
  subdirectory for add-on methods.
 </para>

 <sect1 id="tablesample-support-functions">
  <title>Sampling Method Support Functions</title>

  <para>
   The TSM handler function returns a palloc'd <type>TsmRoutine</type> struct
   containing pointers to the support functions described below.  Most of
   the functions are required, but some are optional, and those pointers can
   be NULL.
  </para>

  <para>
<programlisting>
void
SampleScanGetSampleSize (PlannerInfo *root,
                         RelOptInfo *baserel,
                         List *paramexprs,
                         BlockNumber *pages,
                         double *tuples);
</programlisting>

   This function is called during planning.  It must estimate the number of
   relation pages that will be read during a sample scan, and the number of
   tuples that will be selected by the scan.  (For example, these might be
   determined by estimating the sampling fraction, and then multiplying
   the <literal>baserel-&gt;pages</literal> and <literal>baserel-&gt;tuples</literal>
   numbers by that, being sure to round the results to integral values.)
   The <literal>paramexprs</literal> list holds the expression(s) that are
   parameters to the <literal>TABLESAMPLE</literal> clause.  It is recommended to
   use <function>estimate_expression_value()</function> to try to reduce these
   expressions to constants, if their values are needed for estimation
   purposes; but the function must provide size estimates even if they cannot
   be reduced, and it should not fail even if the values appear invalid
   (remember that they're only estimates of what the run-time values will be).
   The <literal>pages</literal> and <literal>tuples</literal> parameters are outputs.
  </para>

  <para>
<programlisting>
void
InitSampleScan (SampleScanState *node,
                int eflags);
</programlisting>

   Initialize for execution of a SampleScan plan node.
   This is called during executor startup.
   It should perform any initialization needed before processing can start.
   The <structname>SampleScanState</structname> node has already been created, but
   its <structfield>tsm_state</structfield> field is NULL.
   The <function>InitSampleScan</function> function can palloc whatever internal
   state data is needed by the sampling method, and store a pointer to
   it in <literal>node-&gt;tsm_state</literal>.
   Information about the table to scan is accessible through other fields
   of the <structname>SampleScanState</structname> node (but note that the
   <literal>node-&gt;ss.ss_currentScanDesc</literal> scan descriptor is not set
   up yet).
   <literal>eflags</literal> contains flag bits describing the executor's
   operating mode for this plan node.
  </para>

  <para>
   When <literal>(eflags &amp; EXEC_FLAG_EXPLAIN_ONLY)</literal> is true,
   the scan will not actually be performed, so this function should only do
   the minimum required to make the node state valid for <command>EXPLAIN</command>
   and <function>EndSampleScan</function>.
  </para>

  <para>
   This function can be omitted (set the pointer to NULL), in which case
   <function>BeginSampleScan</function> must perform all initialization needed
   by the sampling method.
  </para>

  <para>
<programlisting>
void
BeginSampleScan (SampleScanState *node,
                 Datum *params,
                 int nparams,
                 uint32 seed);
</programlisting>

   Begin execution of a sampling scan.
   This is called just before the first attempt to fetch a tuple, and
   may be called again if the scan needs to be restarted.
   Information about the table to scan is accessible through fields
   of the <structname>SampleScanState</structname> node (but note that the
   <literal>node-&gt;ss.ss_currentScanDesc</literal> scan descriptor is not set
   up yet).
   The <literal>params</literal> array, of length <literal>nparams</literal>, contains the
   values of the parameters supplied in the <literal>TABLESAMPLE</literal> clause.
   These will have the number and types specified in the sampling
   method's <literal>parameterTypes</literal> list, and have been checked
   to not be null.
   <literal>seed</literal> contains a seed to use for any random numbers generated
   within the sampling method; it is either a hash derived from the
   <literal>REPEATABLE</literal> value if one was given, or the result
   of <literal>random()</literal> if not.
  </para>

  <para>
   This function may adjust the fields <literal>node-&gt;use_bulkread</literal>
   and <literal>node-&gt;use_pagemode</literal>.
   If <literal>node-&gt;use_bulkread</literal> is <literal>true</literal>, which it is by
   default, the scan will use a buffer access strategy that encourages
   recycling buffers after use.  It might be reasonable to set this
   to <literal>false</literal> if the scan will visit only a small fraction of the
   table's pages.
   If <literal>node-&gt;use_pagemode</literal> is <literal>true</literal>, which it is by
   default, the scan will perform visibility checking in a single pass for
   all tuples on each visited page.  It might be reasonable to set this
   to <literal>false</literal> if the scan will select only a small fraction of the
   tuples on each visited page.  That will result in fewer tuple visibility
   checks being performed, though each one will be more expensive because it
   will require more locking.
  </para>

  <para>
   If the sampling method is
   marked <literal>repeatable_across_scans</literal>, it must be able to
   select the same set of tuples during a rescan as it did originally, that is
   a fresh call of <function>BeginSampleScan</function> must lead to selecting the
   same tuples as before (if the <literal>TABLESAMPLE</literal> parameters
   and seed don't change).
  </para>

  <para>
<programlisting>
BlockNumber
NextSampleBlock (SampleScanState *node, BlockNumber nblocks);
</programlisting>

   Returns the block number of the next page to be scanned, or
   <literal>InvalidBlockNumber</literal> if no pages remain to be scanned.
  </para>

  <para>
   This function can be omitted (set the pointer to NULL), in which case
   the core code will perform a sequential scan of the entire relation.
   Such a scan can use synchronized scanning, so that the sampling method
   cannot assume that the relation pages are visited in the same order on
   each scan.
  </para>

  <para>
<programlisting>
OffsetNumber
NextSampleTuple (SampleScanState *node,
                 BlockNumber blockno,
                 OffsetNumber maxoffset);
</programlisting>

   Returns the offset number of the next tuple to be sampled on the
   specified page, or <literal>InvalidOffsetNumber</literal> if no tuples remain to
   be sampled.  <literal>maxoffset</literal> is the largest offset number in use
   on the page.
  </para>

  <note>
   <para>
    <function>NextSampleTuple</function> is not explicitly told which of the offset
    numbers in the range <literal>1 .. maxoffset</literal> actually contain valid
    tuples.  This is not normally a problem since the core code ignores
    requests to sample missing or invisible tuples; that should not result in
    any bias in the sample.  However, if necessary, the function can use
    <literal>node-&gt;donetuples</literal> to examine how many of the tuples
    it returned were valid and visible.
   </para>
  </note>

  <note>
   <para>
    <function>NextSampleTuple</function> must <emphasis>not</emphasis> assume
    that <literal>blockno</literal> is the same page number returned by the most
    recent <function>NextSampleBlock</function> call.  It was returned by some
    previous <function>NextSampleBlock</function> call, but the core code is allowed
    to call <function>NextSampleBlock</function> in advance of actually scanning
    pages, so as to support prefetching.  It is OK to assume that once
    sampling of a given page begins, successive <function>NextSampleTuple</function>
    calls all refer to the same page until <literal>InvalidOffsetNumber</literal> is
    returned.
   </para>
  </note>

  <para>
<programlisting>
void
EndSampleScan (SampleScanState *node);
</programlisting>

   End the scan and release resources.  It is normally not important
   to release palloc'd memory, but any externally-visible resources
   should be cleaned up.
   This function can be omitted (set the pointer to NULL) in the common
   case where no such resources exist.
  </para>

 </sect1>

</chapter>