Baoding-FenShuaJiang
/
Green_GoodERP18_FSJ_Standard
Следить 1
В избранное 0
Вики 捐赠
gooderp18绿色标准版
Вы не можете выбрать более 25 тем Темы должны начинаться с буквы или цифры, могут содержать дефисы(-) и должны содержать не более 35 символов.
ветка: master
Ветки Теги
${ item.name }
Создать ветку ${ searchTerm }
из 'master'
${ noResults }
Green_GoodERP18_FSJ_Standard/PostgreSQL/doc/postgresql/html/fdw-callbacks.html

1066 lines
66KB
Исходник Постоянная ссылка Blame История 

  1. <?xml version="1.0" encoding="UTF-8" standalone="no"?>

  2. <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>56.2. Foreign Data Wrapper Callback Routines</title><link rel="stylesheet" type="text/css" href="stylesheet.css" /><link rev="made" href="pgsql-docs@lists.postgresql.org" /><meta name="generator" content="DocBook XSL Stylesheets V1.79.1" /><link rel="prev" href="fdw-functions.html" title="56.1. Foreign Data Wrapper Functions" /><link rel="next" href="fdw-helpers.html" title="56.3. Foreign Data Wrapper Helper Functions" /></head><body><div xmlns="http://www.w3.org/TR/xhtml1/transitional" class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="5" align="center">56.2. Foreign Data Wrapper Callback Routines</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="fdw-functions.html" title="56.1. Foreign Data Wrapper Functions">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="fdwhandler.html" title="Chapter 56. Writing a Foreign Data Wrapper">Up</a></td><th width="60%" align="center">Chapter 56. Writing a Foreign Data Wrapper</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 12.4 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="fdw-helpers.html" title="56.3. Foreign Data Wrapper Helper Functions">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="FDW-CALLBACKS"><div class="titlepage"><div><div><h2 class="title" style="clear: both">56.2. Foreign Data Wrapper Callback Routines</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-SCAN">56.2.1. FDW Routines for Scanning Foreign Tables</a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-JOIN-SCAN">56.2.2. FDW Routines for Scanning Foreign Joins</a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-UPPER-PLANNING">56.2.3. FDW Routines for Planning Post-Scan/Join Processing</a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-UPDATE">56.2.4. FDW Routines for Updating Foreign Tables</a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-ROW-LOCKING">56.2.5. FDW Routines for Row Locking</a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-EXPLAIN">56.2.6. FDW Routines for <code class="command">EXPLAIN</code></a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-ANALYZE">56.2.7. FDW Routines for <code class="command">ANALYZE</code></a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-IMPORT">56.2.8. FDW Routines for <code class="command">IMPORT FOREIGN SCHEMA</code></a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-PARALLEL">56.2.9. FDW Routines for Parallel Execution</a></span></dt><dt><span class="sect2"><a href="fdw-callbacks.html#FDW-CALLBACKS-REPARAMETERIZE-PATHS">56.2.10. FDW Routines for Reparameterization of Paths</a></span></dt></dl></div><p>

  3.      The FDW handler function returns a palloc'd <code class="structname">FdwRoutine</code>

  4.      struct containing pointers to the callback functions described below.

  5.      The scan-related functions are required, the rest are optional.

  6.     </p><p>

  7.      The <code class="structname">FdwRoutine</code> struct type is declared in

  8.      <code class="filename">src/include/foreign/fdwapi.h</code>, which see for additional

  9.      details.

  10.     </p><div class="sect2" id="FDW-CALLBACKS-SCAN"><div class="titlepage"><div><div><h3 class="title">56.2.1. FDW Routines for Scanning Foreign Tables</h3></div></div></div><p>

  11. </p><pre class="programlisting">

  12. void

  13. GetForeignRelSize(PlannerInfo *root,

  14.                   RelOptInfo *baserel,

  15.                   Oid foreigntableid);

  16. </pre><p>

  17. 

  18.      Obtain relation size estimates for a foreign table.  This is called

  19.      at the beginning of planning for a query that scans a foreign table.

  20.      <code class="literal">root</code> is the planner's global information about the query;

  21.      <code class="literal">baserel</code> is the planner's information about this table; and

  22.      <code class="literal">foreigntableid</code> is the <code class="structname">pg_class</code> OID of the

  23.      foreign table.  (<code class="literal">foreigntableid</code> could be obtained from the

  24.      planner data structures, but it's passed explicitly to save effort.)

  25.     </p><p>

  26.      This function should update <code class="literal">baserel-&gt;rows</code> to be the

  27.      expected number of rows returned by the table scan, after accounting for

  28.      the filtering done by the restriction quals.  The initial value of

  29.      <code class="literal">baserel-&gt;rows</code> is just a constant default estimate, which

  30.      should be replaced if at all possible.  The function may also choose to

  31.      update <code class="literal">baserel-&gt;width</code> if it can compute a better estimate

  32.      of the average result row width.

  33.      (The initial value is based on column data types and on column

  34.      average-width values measured by the last <code class="command">ANALYZE</code>.)

  35.      Also, this function may update <code class="literal">baserel-&gt;tuples</code> if

  36.      it can compute a better estimate of the foreign table's total row count.

  37.      (The initial value is

  38.      from <code class="structname">pg_class</code>.<code class="structfield">reltuples</code>

  39.      which represents the total row count seen by the

  40.      last <code class="command">ANALYZE</code>.)

  41.     </p><p>

  42.      See <a class="xref" href="fdw-planning.html" title="56.4. Foreign Data Wrapper Query Planning">Section 56.4</a> for additional information.

  43.     </p><p>

  44. </p><pre class="programlisting">

  45. void

  46. GetForeignPaths(PlannerInfo *root,

  47.                 RelOptInfo *baserel,

  48.                 Oid foreigntableid);

  49. </pre><p>

  50. 

  51.      Create possible access paths for a scan on a foreign table.

  52.      This is called during query planning.

  53.      The parameters are the same as for <code class="function">GetForeignRelSize</code>,

  54.      which has already been called.

  55.     </p><p>

  56.      This function must generate at least one access path

  57.      (<code class="structname">ForeignPath</code> node) for a scan on the foreign table and

  58.      must call <code class="function">add_path</code> to add each such path to

  59.      <code class="literal">baserel-&gt;pathlist</code>.  It's recommended to use

  60.      <code class="function">create_foreignscan_path</code> to build the

  61.      <code class="structname">ForeignPath</code> nodes.  The function can generate multiple

  62.      access paths, e.g., a path which has valid <code class="literal">pathkeys</code> to

  63.      represent a pre-sorted result.  Each access path must contain cost

  64.      estimates, and can contain any FDW-private information that is needed to

  65.      identify the specific scan method intended.

  66.     </p><p>

  67.      See <a class="xref" href="fdw-planning.html" title="56.4. Foreign Data Wrapper Query Planning">Section 56.4</a> for additional information.

  68.     </p><p>

  69. </p><pre class="programlisting">

  70. ForeignScan *

  71. GetForeignPlan(PlannerInfo *root,

  72.                RelOptInfo *baserel,

  73.                Oid foreigntableid,

  74.                ForeignPath *best_path,

  75.                List *tlist,

  76.                List *scan_clauses,

  77.                Plan *outer_plan);

  78. </pre><p>

  79. 

  80.      Create a <code class="structname">ForeignScan</code> plan node from the selected foreign

  81.      access path.  This is called at the end of query planning.

  82.      The parameters are as for <code class="function">GetForeignRelSize</code>, plus

  83.      the selected <code class="structname">ForeignPath</code> (previously produced by

  84.      <code class="function">GetForeignPaths</code>, <code class="function">GetForeignJoinPaths</code>,

  85.      or <code class="function">GetForeignUpperPaths</code>),

  86.      the target list to be emitted by the plan node,

  87.      the restriction clauses to be enforced by the plan node,

  88.      and the outer subplan of the <code class="structname">ForeignScan</code>,

  89.      which is used for rechecks performed by <code class="function">RecheckForeignScan</code>.

  90.      (If the path is for a join rather than a base

  91.      relation, <code class="literal">foreigntableid</code> is <code class="literal">InvalidOid</code>.)

  92.     </p><p>

  93.      This function must create and return a <code class="structname">ForeignScan</code> plan

  94.      node; it's recommended to use <code class="function">make_foreignscan</code> to build the

  95.      <code class="structname">ForeignScan</code> node.

  96.     </p><p>

  97.      See <a class="xref" href="fdw-planning.html" title="56.4. Foreign Data Wrapper Query Planning">Section 56.4</a> for additional information.

  98.     </p><p>

  99. </p><pre class="programlisting">

  100. void

  101. BeginForeignScan(ForeignScanState *node,

  102.                  int eflags);

  103. </pre><p>

  104. 

  105.      Begin executing a foreign scan. This is called during executor startup.

  106.      It should perform any initialization needed before the scan can start,

  107.      but not start executing the actual scan (that should be done upon the

  108.      first call to <code class="function">IterateForeignScan</code>).

  109.      The <code class="structname">ForeignScanState</code> node has already been created, but

  110.      its <code class="structfield">fdw_state</code> field is still NULL.  Information about

  111.      the table to scan is accessible through the

  112.      <code class="structname">ForeignScanState</code> node (in particular, from the underlying

  113.      <code class="structname">ForeignScan</code> plan node, which contains any FDW-private

  114.      information provided by <code class="function">GetForeignPlan</code>).

  115.      <code class="literal">eflags</code> contains flag bits describing the executor's

  116.      operating mode for this plan node.

  117.     </p><p>

  118.      Note that when <code class="literal">(eflags &amp; EXEC_FLAG_EXPLAIN_ONLY)</code> is

  119.      true, this function should not perform any externally-visible actions;

  120.      it should only do the minimum required to make the node state valid

  121.      for <code class="function">ExplainForeignScan</code> and <code class="function">EndForeignScan</code>.

  122.     </p><p>

  123. </p><pre class="programlisting">

  124. TupleTableSlot *

  125. IterateForeignScan(ForeignScanState *node);

  126. </pre><p>

  127. 

  128.      Fetch one row from the foreign source, returning it in a tuple table slot

  129.      (the node's <code class="structfield">ScanTupleSlot</code> should be used for this

  130.      purpose).  Return NULL if no more rows are available.  The tuple table

  131.      slot infrastructure allows either a physical or virtual tuple to be

  132.      returned; in most cases the latter choice is preferable from a

  133.      performance standpoint.  Note that this is called in a short-lived memory

  134.      context that will be reset between invocations.  Create a memory context

  135.      in <code class="function">BeginForeignScan</code> if you need longer-lived storage, or use

  136.      the <code class="structfield">es_query_cxt</code> of the node's <code class="structname">EState</code>.

  137.     </p><p>

  138.      The rows returned must match the <code class="structfield">fdw_scan_tlist</code> target

  139.      list if one was supplied, otherwise they must match the row type of the

  140.      foreign table being scanned.  If you choose to optimize away fetching

  141.      columns that are not needed, you should insert nulls in those column

  142.      positions, or else generate a <code class="structfield">fdw_scan_tlist</code> list with

  143.      those columns omitted.

  144.     </p><p>

  145.      Note that <span class="productname">PostgreSQL</span>'s executor doesn't care

  146.      whether the rows returned violate any constraints that were defined on

  147.      the foreign table — but the planner does care, and may optimize

  148.      queries incorrectly if there are rows visible in the foreign table that

  149.      do not satisfy a declared constraint.  If a constraint is violated when

  150.      the user has declared that the constraint should hold true, it may be

  151.      appropriate to raise an error (just as you would need to do in the case

  152.      of a data type mismatch).

  153.     </p><p>

  154. </p><pre class="programlisting">

  155. void

  156. ReScanForeignScan(ForeignScanState *node);

  157. </pre><p>

  158. 

  159.      Restart the scan from the beginning.  Note that any parameters the

  160.      scan depends on may have changed value, so the new scan does not

  161.      necessarily return exactly the same rows.

  162.     </p><p>

  163. </p><pre class="programlisting">

  164. void

  165. EndForeignScan(ForeignScanState *node);

  166. </pre><p>

  167. 

  168.      End the scan and release resources.  It is normally not important

  169.      to release palloc'd memory, but for example open files and connections

  170.      to remote servers should be cleaned up.

  171.     </p></div><div class="sect2" id="FDW-CALLBACKS-JOIN-SCAN"><div class="titlepage"><div><div><h3 class="title">56.2.2. FDW Routines for Scanning Foreign Joins</h3></div></div></div><p>

  172.      If an FDW supports performing foreign joins remotely (rather than

  173.      by fetching both tables' data and doing the join locally), it should

  174.      provide this callback function:

  175.     </p><p>

  176. </p><pre class="programlisting">

  177. void

  178. GetForeignJoinPaths(PlannerInfo *root,

  179.                     RelOptInfo *joinrel,

  180.                     RelOptInfo *outerrel,

  181.                     RelOptInfo *innerrel,

  182.                     JoinType jointype,

  183.                     JoinPathExtraData *extra);

  184. </pre><p>

  185.      Create possible access paths for a join of two (or more) foreign tables

  186.      that all belong to the same foreign server.  This optional

  187.      function is called during query planning.  As

  188.      with <code class="function">GetForeignPaths</code>, this function should

  189.      generate <code class="structname">ForeignPath</code> path(s) for the

  190.      supplied <code class="literal">joinrel</code>

  191.      (use <code class="function">create_foreign_join_path</code> to build them),

  192.      and call <code class="function">add_path</code> to add these

  193.      paths to the set of paths considered for the join.  But unlike

  194.      <code class="function">GetForeignPaths</code>, it is not necessary that this function

  195.      succeed in creating at least one path, since paths involving local

  196.      joining are always possible.

  197.     </p><p>

  198.      Note that this function will be invoked repeatedly for the same join

  199.      relation, with different combinations of inner and outer relations; it is

  200.      the responsibility of the FDW to minimize duplicated work.

  201.     </p><p>

  202.      If a <code class="structname">ForeignPath</code> path is chosen for the join, it will

  203.      represent the entire join process; paths generated for the component

  204.      tables and subsidiary joins will not be used.  Subsequent processing of

  205.      the join path proceeds much as it does for a path scanning a single

  206.      foreign table.  One difference is that the <code class="structfield">scanrelid</code> of

  207.      the resulting <code class="structname">ForeignScan</code> plan node should be set to zero,

  208.      since there is no single relation that it represents; instead,

  209.      the <code class="structfield">fs_relids</code> field of the <code class="structname">ForeignScan</code>

  210.      node represents the set of relations that were joined.  (The latter field

  211.      is set up automatically by the core planner code, and need not be filled

  212.      by the FDW.)  Another difference is that, because the column list for a

  213.      remote join cannot be found from the system catalogs, the FDW must

  214.      fill <code class="structfield">fdw_scan_tlist</code> with an appropriate list

  215.      of <code class="structfield">TargetEntry</code> nodes, representing the set of columns

  216.      it will supply at run time in the tuples it returns.

  217.     </p><p>

  218.      See <a class="xref" href="fdw-planning.html" title="56.4. Foreign Data Wrapper Query Planning">Section 56.4</a> for additional information.

  219.     </p></div><div class="sect2" id="FDW-CALLBACKS-UPPER-PLANNING"><div class="titlepage"><div><div><h3 class="title">56.2.3. FDW Routines for Planning Post-Scan/Join Processing</h3></div></div></div><p>

  220.      If an FDW supports performing remote post-scan/join processing, such as

  221.      remote aggregation, it should provide this callback function:

  222.     </p><p>

  223. </p><pre class="programlisting">

  224. void

  225. GetForeignUpperPaths(PlannerInfo *root,

  226.                      UpperRelationKind stage,

  227.                      RelOptInfo *input_rel,

  228.                      RelOptInfo *output_rel,

  229.                      void *extra);

  230. </pre><p>

  231.      Create possible access paths for <em class="firstterm">upper relation</em> processing,

  232.      which is the planner's term for all post-scan/join query processing, such

  233.      as aggregation, window functions, sorting, and table updates.  This

  234.      optional function is called during query planning.  Currently, it is

  235.      called only if all base relation(s) involved in the query belong to the

  236.      same FDW.  This function should generate <code class="structname">ForeignPath</code>

  237.      path(s) for any post-scan/join processing that the FDW knows how to

  238.      perform remotely

  239.      (use <code class="function">create_foreign_upper_path</code> to build them),

  240.      and call <code class="function">add_path</code> to add these paths to

  241.      the indicated upper relation.  As with <code class="function">GetForeignJoinPaths</code>,

  242.      it is not necessary that this function succeed in creating any paths,

  243.      since paths involving local processing are always possible.

  244.     </p><p>

  245.      The <code class="literal">stage</code> parameter identifies which post-scan/join step is

  246.      currently being considered.  <code class="literal">output_rel</code> is the upper relation

  247.      that should receive paths representing computation of this step,

  248.      and <code class="literal">input_rel</code> is the relation representing the input to this

  249.      step.  The <code class="literal">extra</code> parameter provides additional details,

  250.      currently, it is set only for <code class="literal">UPPERREL_PARTIAL_GROUP_AGG</code>

  251.      or <code class="literal">UPPERREL_GROUP_AGG</code>, in which case it points to a

  252.      <code class="literal">GroupPathExtraData</code> structure;

  253.      or for <code class="literal">UPPERREL_FINAL</code>, in which case it points to a

  254.      <code class="literal">FinalPathExtraData</code> structure.

  255.      (Note that <code class="structname">ForeignPath</code> paths added

  256.      to <code class="literal">output_rel</code> would typically not have any direct dependency

  257.      on paths of the <code class="literal">input_rel</code>, since their processing is expected

  258.      to be done externally.  However, examining paths previously generated for

  259.      the previous processing step can be useful to avoid redundant planning

  260.      work.)

  261.     </p><p>

  262.      See <a class="xref" href="fdw-planning.html" title="56.4. Foreign Data Wrapper Query Planning">Section 56.4</a> for additional information.

  263.     </p></div><div class="sect2" id="FDW-CALLBACKS-UPDATE"><div class="titlepage"><div><div><h3 class="title">56.2.4. FDW Routines for Updating Foreign Tables</h3></div></div></div><p>

  264.      If an FDW supports writable foreign tables, it should provide

  265.      some or all of the following callback functions depending on

  266.      the needs and capabilities of the FDW:

  267.     </p><p>

  268. </p><pre class="programlisting">

  269. void

  270. AddForeignUpdateTargets(Query *parsetree,

  271.                         RangeTblEntry *target_rte,

  272.                         Relation target_relation);

  273. </pre><p>

  274. 

  275.      <code class="command">UPDATE</code> and <code class="command">DELETE</code> operations are performed

  276.      against rows previously fetched by the table-scanning functions.  The

  277.      FDW may need extra information, such as a row ID or the values of

  278.      primary-key columns, to ensure that it can identify the exact row to

  279.      update or delete.  To support that, this function can add extra hidden,

  280.      or <span class="quote">“<span class="quote">junk</span>”</span>, target columns to the list of columns that are to be

  281.      retrieved from the foreign table during an <code class="command">UPDATE</code> or

  282.      <code class="command">DELETE</code>.

  283.     </p><p>

  284.      To do that, add <code class="structname">TargetEntry</code> items to

  285.      <code class="literal">parsetree-&gt;targetList</code>, containing expressions for the

  286.      extra values to be fetched.  Each such entry must be marked

  287.      <code class="structfield">resjunk</code> = <code class="literal">true</code>, and must have a distinct

  288.      <code class="structfield">resname</code> that will identify it at execution time.

  289.      Avoid using names matching <code class="literal">ctid<em class="replaceable"><code>N</code></em></code>,

  290.      <code class="literal">wholerow</code>, or

  291.      <code class="literal">wholerow<em class="replaceable"><code>N</code></em></code>, as the core system can

  292.      generate junk columns of these names.

  293.      If the extra expressions are more complex than simple Vars, they

  294.      must be run through <code class="function">eval_const_expressions</code>

  295.      before adding them to the targetlist.

  296.     </p><p>

  297.      Although this function is called during planning, the

  298.      information provided is a bit different from that available to other

  299.      planning routines.

  300.      <code class="literal">parsetree</code> is the parse tree for the <code class="command">UPDATE</code> or

  301.      <code class="command">DELETE</code> command, while <code class="literal">target_rte</code> and

  302.      <code class="literal">target_relation</code> describe the target foreign table.

  303.     </p><p>

  304.      If the <code class="function">AddForeignUpdateTargets</code> pointer is set to

  305.      <code class="literal">NULL</code>, no extra target expressions are added.

  306.      (This will make it impossible to implement <code class="command">DELETE</code>

  307.      operations, though <code class="command">UPDATE</code> may still be feasible if the FDW

  308.      relies on an unchanging primary key to identify rows.)

  309.     </p><p>

  310. </p><pre class="programlisting">

  311. List *

  312. PlanForeignModify(PlannerInfo *root,

  313.                   ModifyTable *plan,

  314.                   Index resultRelation,

  315.                   int subplan_index);

  316. </pre><p>

  317. 

  318.      Perform any additional planning actions needed for an insert, update, or

  319.      delete on a foreign table.  This function generates the FDW-private

  320.      information that will be attached to the <code class="structname">ModifyTable</code> plan

  321.      node that performs the update action.  This private information must

  322.      have the form of a <code class="literal">List</code>, and will be delivered to

  323.      <code class="function">BeginForeignModify</code> during the execution stage.

  324.     </p><p>

  325.      <code class="literal">root</code> is the planner's global information about the query.

  326.      <code class="literal">plan</code> is the <code class="structname">ModifyTable</code> plan node, which is

  327.      complete except for the <code class="structfield">fdwPrivLists</code> field.

  328.      <code class="literal">resultRelation</code> identifies the target foreign table by its

  329.      range table index.  <code class="literal">subplan_index</code> identifies which target of

  330.      the <code class="structname">ModifyTable</code> plan node this is, counting from zero;

  331.      use this if you want to index into <code class="literal">plan-&gt;plans</code> or other

  332.      substructure of the <code class="literal">plan</code> node.

  333.     </p><p>

  334.      See <a class="xref" href="fdw-planning.html" title="56.4. Foreign Data Wrapper Query Planning">Section 56.4</a> for additional information.

  335.     </p><p>

  336.      If the <code class="function">PlanForeignModify</code> pointer is set to

  337.      <code class="literal">NULL</code>, no additional plan-time actions are taken, and the

  338.      <code class="literal">fdw_private</code> list delivered to

  339.      <code class="function">BeginForeignModify</code> will be NIL.

  340.     </p><p>

  341. </p><pre class="programlisting">

  342. void

  343. BeginForeignModify(ModifyTableState *mtstate,

  344.                    ResultRelInfo *rinfo,

  345.                    List *fdw_private,

  346.                    int subplan_index,

  347.                    int eflags);

  348. </pre><p>

  349. 

  350.      Begin executing a foreign table modification operation.  This routine is

  351.      called during executor startup.  It should perform any initialization

  352.      needed prior to the actual table modifications.  Subsequently,

  353.      <code class="function">ExecForeignInsert</code>, <code class="function">ExecForeignUpdate</code> or

  354.      <code class="function">ExecForeignDelete</code> will be called for each tuple to be

  355.      inserted, updated, or deleted.

  356.     </p><p>

  357.      <code class="literal">mtstate</code> is the overall state of the

  358.      <code class="structname">ModifyTable</code> plan node being executed; global data about

  359.      the plan and execution state is available via this structure.

  360.      <code class="literal">rinfo</code> is the <code class="structname">ResultRelInfo</code> struct describing

  361.      the target foreign table.  (The <code class="structfield">ri_FdwState</code> field of

  362.      <code class="structname">ResultRelInfo</code> is available for the FDW to store any

  363.      private state it needs for this operation.)

  364.      <code class="literal">fdw_private</code> contains the private data generated by

  365.      <code class="function">PlanForeignModify</code>, if any.

  366.      <code class="literal">subplan_index</code> identifies which target of

  367.      the <code class="structname">ModifyTable</code> plan node this is.

  368.      <code class="literal">eflags</code> contains flag bits describing the executor's

  369.      operating mode for this plan node.

  370.     </p><p>

  371.      Note that when <code class="literal">(eflags &amp; EXEC_FLAG_EXPLAIN_ONLY)</code> is

  372.      true, this function should not perform any externally-visible actions;

  373.      it should only do the minimum required to make the node state valid

  374.      for <code class="function">ExplainForeignModify</code> and <code class="function">EndForeignModify</code>.

  375.     </p><p>

  376.      If the <code class="function">BeginForeignModify</code> pointer is set to

  377.      <code class="literal">NULL</code>, no action is taken during executor startup.

  378.     </p><p>

  379. </p><pre class="programlisting">

  380. TupleTableSlot *

  381. ExecForeignInsert(EState *estate,

  382.                   ResultRelInfo *rinfo,

  383.                   TupleTableSlot *slot,

  384.                   TupleTableSlot *planSlot);

  385. </pre><p>

  386. 

  387.      Insert one tuple into the foreign table.

  388.      <code class="literal">estate</code> is global execution state for the query.

  389.      <code class="literal">rinfo</code> is the <code class="structname">ResultRelInfo</code> struct describing

  390.      the target foreign table.

  391.      <code class="literal">slot</code> contains the tuple to be inserted; it will match the

  392.      row-type definition of the foreign table.

  393.      <code class="literal">planSlot</code> contains the tuple that was generated by the

  394.      <code class="structname">ModifyTable</code> plan node's subplan; it differs from

  395.      <code class="literal">slot</code> in possibly containing additional <span class="quote">“<span class="quote">junk</span>”</span>

  396.      columns.  (The <code class="literal">planSlot</code> is typically of little interest

  397.      for <code class="command">INSERT</code> cases, but is provided for completeness.)

  398.     </p><p>

  399.      The return value is either a slot containing the data that was actually

  400.      inserted (this might differ from the data supplied, for example as a

  401.      result of trigger actions), or NULL if no row was actually inserted

  402.      (again, typically as a result of triggers).  The passed-in

  403.      <code class="literal">slot</code> can be re-used for this purpose.

  404.     </p><p>

  405.      The data in the returned slot is used only if the <code class="command">INSERT</code>

  406.      statement has a <code class="literal">RETURNING</code> clause or involves a view

  407.      <code class="literal">WITH CHECK OPTION</code>; or if the foreign table has

  408.      an <code class="literal">AFTER ROW</code> trigger.  Triggers require all columns,

  409.      but the FDW could choose to optimize away returning some or all columns

  410.      depending on the contents of the <code class="literal">RETURNING</code> clause or

  411.      <code class="literal">WITH CHECK OPTION</code> constraints.  Regardless, some slot

  412.      must be returned to indicate success, or the query's reported row count

  413.      will be wrong.

  414.     </p><p>

  415.      If the <code class="function">ExecForeignInsert</code> pointer is set to

  416.      <code class="literal">NULL</code>, attempts to insert into the foreign table will fail

  417.      with an error message.

  418.     </p><p>

  419.      Note that this function is also called when inserting routed tuples into

  420.      a foreign-table partition or executing <code class="command">COPY FROM</code> on

  421.      a foreign table, in which case it is called in a different way than it

  422.      is in the <code class="command">INSERT</code> case.  See the callback functions

  423.      described below that allow the FDW to support that.

  424.     </p><p>

  425. </p><pre class="programlisting">

  426. TupleTableSlot *

  427. ExecForeignUpdate(EState *estate,

  428.                   ResultRelInfo *rinfo,

  429.                   TupleTableSlot *slot,

  430.                   TupleTableSlot *planSlot);

  431. </pre><p>

  432. 

  433.      Update one tuple in the foreign table.

  434.      <code class="literal">estate</code> is global execution state for the query.

  435.      <code class="literal">rinfo</code> is the <code class="structname">ResultRelInfo</code> struct describing

  436.      the target foreign table.

  437.      <code class="literal">slot</code> contains the new data for the tuple; it will match the

  438.      row-type definition of the foreign table.

  439.      <code class="literal">planSlot</code> contains the tuple that was generated by the

  440.      <code class="structname">ModifyTable</code> plan node's subplan; it differs from

  441.      <code class="literal">slot</code> in possibly containing additional <span class="quote">“<span class="quote">junk</span>”</span>

  442.      columns.  In particular, any junk columns that were requested by

  443.      <code class="function">AddForeignUpdateTargets</code> will be available from this slot.

  444.     </p><p>

  445.      The return value is either a slot containing the row as it was actually

  446.      updated (this might differ from the data supplied, for example as a

  447.      result of trigger actions), or NULL if no row was actually updated

  448.      (again, typically as a result of triggers).  The passed-in

  449.      <code class="literal">slot</code> can be re-used for this purpose.

  450.     </p><p>

  451.      The data in the returned slot is used only if the <code class="command">UPDATE</code>

  452.      statement has a <code class="literal">RETURNING</code> clause or involves a view

  453.      <code class="literal">WITH CHECK OPTION</code>; or if the foreign table has

  454.      an <code class="literal">AFTER ROW</code> trigger.  Triggers require all columns,

  455.      but the FDW could choose to optimize away returning some or all columns

  456.      depending on the contents of the <code class="literal">RETURNING</code> clause or

  457.      <code class="literal">WITH CHECK OPTION</code> constraints.  Regardless, some slot

  458.      must be returned to indicate success, or the query's reported row count

  459.      will be wrong.

  460.     </p><p>

  461.      If the <code class="function">ExecForeignUpdate</code> pointer is set to

  462.      <code class="literal">NULL</code>, attempts to update the foreign table will fail

  463.      with an error message.

  464.     </p><p>

  465. </p><pre class="programlisting">

  466. TupleTableSlot *

  467. ExecForeignDelete(EState *estate,

  468.                   ResultRelInfo *rinfo,

  469.                   TupleTableSlot *slot,

  470.                   TupleTableSlot *planSlot);

  471. </pre><p>

  472. 

  473.      Delete one tuple from the foreign table.

  474.      <code class="literal">estate</code> is global execution state for the query.

  475.      <code class="literal">rinfo</code> is the <code class="structname">ResultRelInfo</code> struct describing

  476.      the target foreign table.

  477.      <code class="literal">slot</code> contains nothing useful upon call, but can be used to

  478.      hold the returned tuple.

  479.      <code class="literal">planSlot</code> contains the tuple that was generated by the

  480.      <code class="structname">ModifyTable</code> plan node's subplan; in particular, it will

  481.      carry any junk columns that were requested by

  482.      <code class="function">AddForeignUpdateTargets</code>.  The junk column(s) must be used

  483.      to identify the tuple to be deleted.

  484.     </p><p>

  485.      The return value is either a slot containing the row that was deleted,

  486.      or NULL if no row was deleted (typically as a result of triggers).  The

  487.      passed-in <code class="literal">slot</code> can be used to hold the tuple to be returned.

  488.     </p><p>

  489.      The data in the returned slot is used only if the <code class="command">DELETE</code>

  490.      query has a <code class="literal">RETURNING</code> clause or the foreign table has

  491.      an <code class="literal">AFTER ROW</code> trigger.  Triggers require all columns, but the

  492.      FDW could choose to optimize away returning some or all columns depending

  493.      on the contents of the <code class="literal">RETURNING</code> clause.  Regardless, some

  494.      slot must be returned to indicate success, or the query's reported row

  495.      count will be wrong.

  496.     </p><p>

  497.      If the <code class="function">ExecForeignDelete</code> pointer is set to

  498.      <code class="literal">NULL</code>, attempts to delete from the foreign table will fail

  499.      with an error message.

  500.     </p><p>

  501. </p><pre class="programlisting">

  502. void

  503. EndForeignModify(EState *estate,

  504.                  ResultRelInfo *rinfo);

  505. </pre><p>

  506. 

  507.      End the table update and release resources.  It is normally not important

  508.      to release palloc'd memory, but for example open files and connections

  509.      to remote servers should be cleaned up.

  510.     </p><p>

  511.      If the <code class="function">EndForeignModify</code> pointer is set to

  512.      <code class="literal">NULL</code>, no action is taken during executor shutdown.

  513.     </p><p>

  514.      Tuples inserted into a partitioned table by <code class="command">INSERT</code> or

  515.      <code class="command">COPY FROM</code> are routed to partitions.  If an FDW

  516.      supports routable foreign-table partitions, it should also provide the

  517.      following callback functions.  These functions are also called when

  518.      <code class="command">COPY FROM</code> is executed on a foreign table.

  519.     </p><p>

  520. </p><pre class="programlisting">

  521. void

  522. BeginForeignInsert(ModifyTableState *mtstate,

  523.                    ResultRelInfo *rinfo);

  524. </pre><p>

  525. 

  526.      Begin executing an insert operation on a foreign table.  This routine is

  527.      called right before the first tuple is inserted into the foreign table

  528.      in both cases when it is the partition chosen for tuple routing and the

  529.      target specified in a <code class="command">COPY FROM</code> command.  It should

  530.      perform any initialization needed prior to the actual insertion.

  531.      Subsequently, <code class="function">ExecForeignInsert</code> will be called for

  532.      each tuple to be inserted into the foreign table.

  533.     </p><p>

  534.      <code class="literal">mtstate</code> is the overall state of the

  535.      <code class="structname">ModifyTable</code> plan node being executed; global data about

  536.      the plan and execution state is available via this structure.

  537.      <code class="literal">rinfo</code> is the <code class="structname">ResultRelInfo</code> struct describing

  538.      the target foreign table.  (The <code class="structfield">ri_FdwState</code> field of

  539.      <code class="structname">ResultRelInfo</code> is available for the FDW to store any

  540.      private state it needs for this operation.)

  541.     </p><p>

  542.      When this is called by a <code class="command">COPY FROM</code> command, the

  543.      plan-related global data in <code class="literal">mtstate</code> is not provided

  544.      and the <code class="literal">planSlot</code> parameter of

  545.      <code class="function">ExecForeignInsert</code> subsequently called for each

  546.      inserted tuple is <code class="literal">NULL</code>, whether the foreign table is

  547.      the partition chosen for tuple routing or the target specified in the

  548.      command.

  549.     </p><p>

  550.      If the <code class="function">BeginForeignInsert</code> pointer is set to

  551.      <code class="literal">NULL</code>, no action is taken for the initialization.

  552.     </p><p>

  553.      Note that if the FDW does not support routable foreign-table partitions

  554.      and/or executing <code class="command">COPY FROM</code> on foreign tables, this

  555.      function or <code class="function">ExecForeignInsert</code> subsequently called

  556.      must throw error as needed.

  557.     </p><p>

  558. </p><pre class="programlisting">

  559. void

  560. EndForeignInsert(EState *estate,

  561.                  ResultRelInfo *rinfo);

  562. </pre><p>

  563. 

  564.      End the insert operation and release resources.  It is normally not important

  565.      to release palloc'd memory, but for example open files and connections

  566.      to remote servers should be cleaned up.

  567.     </p><p>

  568.      If the <code class="function">EndForeignInsert</code> pointer is set to

  569.      <code class="literal">NULL</code>, no action is taken for the termination.

  570.     </p><p>

  571. </p><pre class="programlisting">

  572. int

  573. IsForeignRelUpdatable(Relation rel);

  574. </pre><p>

  575. 

  576.      Report which update operations the specified foreign table supports.

  577.      The return value should be a bit mask of rule event numbers indicating

  578.      which operations are supported by the foreign table, using the

  579.      <code class="literal">CmdType</code> enumeration; that is,

  580.      <code class="literal">(1 &lt;&lt; CMD_UPDATE) = 4</code> for <code class="command">UPDATE</code>,

  581.      <code class="literal">(1 &lt;&lt; CMD_INSERT) = 8</code> for <code class="command">INSERT</code>, and

  582.      <code class="literal">(1 &lt;&lt; CMD_DELETE) = 16</code> for <code class="command">DELETE</code>.

  583.     </p><p>

  584.      If the <code class="function">IsForeignRelUpdatable</code> pointer is set to

  585.      <code class="literal">NULL</code>, foreign tables are assumed to be insertable, updatable,

  586.      or deletable if the FDW provides <code class="function">ExecForeignInsert</code>,

  587.      <code class="function">ExecForeignUpdate</code>, or <code class="function">ExecForeignDelete</code>

  588.      respectively.  This function is only needed if the FDW supports some

  589.      tables that are updatable and some that are not.  (Even then, it's

  590.      permissible to throw an error in the execution routine instead of

  591.      checking in this function.  However, this function is used to determine

  592.      updatability for display in the <code class="literal">information_schema</code> views.)

  593.     </p><p>

  594.      Some inserts, updates, and deletes to foreign tables can be optimized

  595.      by implementing an alternative set of interfaces.  The ordinary

  596.      interfaces for inserts, updates, and deletes fetch rows from the remote

  597.      server and then modify those rows one at a time.  In some cases, this

  598.      row-by-row approach is necessary, but it can be inefficient.  If it is

  599.      possible for the foreign server to determine which rows should be

  600.      modified without actually retrieving them, and if there are no local

  601.      structures which would affect the operation (row-level local triggers,

  602.      stored generated columns, or <code class="literal">WITH CHECK OPTION</code>

  603.      constraints from parent views), then it is possible to arrange things

  604.      so that the entire operation is performed on the remote server.  The

  605.      interfaces described below make this possible.

  606.     </p><p>

  607. </p><pre class="programlisting">

  608. bool

  609. PlanDirectModify(PlannerInfo *root,

  610.                  ModifyTable *plan,

  611.                  Index resultRelation,

  612.                  int subplan_index);

  613. </pre><p>

  614. 

  615.      Decide whether it is safe to execute a direct modification

  616.      on the remote server.  If so, return <code class="literal">true</code> after performing

  617.      planning actions needed for that.  Otherwise, return <code class="literal">false</code>.

  618.      This optional function is called during query planning.

  619.      If this function succeeds, <code class="function">BeginDirectModify</code>,

  620.      <code class="function">IterateDirectModify</code> and <code class="function">EndDirectModify</code> will

  621.      be called at the execution stage, instead.  Otherwise, the table

  622.      modification will be executed using the table-updating functions

  623.      described above.

  624.      The parameters are the same as for <code class="function">PlanForeignModify</code>.

  625.     </p><p>

  626.      To execute the direct modification on the remote server, this function

  627.      must rewrite the target subplan with a <code class="structname">ForeignScan</code> plan

  628.      node that executes the direct modification on the remote server.  The

  629.      <code class="structfield">operation</code> field of the <code class="structname">ForeignScan</code> must

  630.      be set to the <code class="literal">CmdType</code> enumeration appropriately; that is,

  631.      <code class="literal">CMD_UPDATE</code> for <code class="command">UPDATE</code>,

  632.      <code class="literal">CMD_INSERT</code> for <code class="command">INSERT</code>, and

  633.      <code class="literal">CMD_DELETE</code> for <code class="command">DELETE</code>.

  634.     </p><p>

  635.      See <a class="xref" href="fdw-planning.html" title="56.4. Foreign Data Wrapper Query Planning">Section 56.4</a> for additional information.

  636.     </p><p>

  637.      If the <code class="function">PlanDirectModify</code> pointer is set to

  638.      <code class="literal">NULL</code>, no attempts to execute a direct modification on the

  639.      remote server are taken.

  640.     </p><p>

  641. </p><pre class="programlisting">

  642. void

  643. BeginDirectModify(ForeignScanState *node,

  644.                   int eflags);

  645. </pre><p>

  646. 

  647.      Prepare to execute a direct modification on the remote server.

  648.      This is called during executor startup.  It should perform any

  649.      initialization needed prior to the direct modification (that should be

  650.      done upon the first call to <code class="function">IterateDirectModify</code>).

  651.      The <code class="structname">ForeignScanState</code> node has already been created, but

  652.      its <code class="structfield">fdw_state</code> field is still NULL.  Information about

  653.      the table to modify is accessible through the

  654.      <code class="structname">ForeignScanState</code> node (in particular, from the underlying

  655.      <code class="structname">ForeignScan</code> plan node, which contains any FDW-private

  656.      information provided by <code class="function">PlanDirectModify</code>).

  657.      <code class="literal">eflags</code> contains flag bits describing the executor's

  658.      operating mode for this plan node.

  659.     </p><p>

  660.      Note that when <code class="literal">(eflags &amp; EXEC_FLAG_EXPLAIN_ONLY)</code> is

  661.      true, this function should not perform any externally-visible actions;

  662.      it should only do the minimum required to make the node state valid

  663.      for <code class="function">ExplainDirectModify</code> and <code class="function">EndDirectModify</code>.

  664.     </p><p>

  665.      If the <code class="function">BeginDirectModify</code> pointer is set to

  666.      <code class="literal">NULL</code>, no attempts to execute a direct modification on the

  667.      remote server are taken.

  668.     </p><p>

  669. </p><pre class="programlisting">

  670. TupleTableSlot *

  671. IterateDirectModify(ForeignScanState *node);

  672. </pre><p>

  673. 

  674.      When the <code class="command">INSERT</code>, <code class="command">UPDATE</code> or <code class="command">DELETE</code>

  675.      query doesn't have a <code class="literal">RETURNING</code> clause, just return NULL

  676.      after a direct modification on the remote server.

  677.      When the query has the clause, fetch one result containing the data

  678.      needed for the <code class="literal">RETURNING</code> calculation, returning it in a

  679.      tuple table slot (the node's <code class="structfield">ScanTupleSlot</code> should be

  680.      used for this purpose).  The data that was actually inserted, updated

  681.      or deleted must be stored in the

  682.      <code class="literal">es_result_relation_info-&gt;ri_projectReturning-&gt;pi_exprContext-&gt;ecxt_scantuple</code>

  683.      of the node's <code class="structname">EState</code>.

  684.      Return NULL if no more rows are available.

  685.      Note that this is called in a short-lived memory context that will be

  686.      reset between invocations.  Create a memory context in

  687.      <code class="function">BeginDirectModify</code> if you need longer-lived storage, or use

  688.      the <code class="structfield">es_query_cxt</code> of the node's <code class="structname">EState</code>.

  689.     </p><p>

  690.      The rows returned must match the <code class="structfield">fdw_scan_tlist</code> target

  691.      list if one was supplied, otherwise they must match the row type of the

  692.      foreign table being updated.  If you choose to optimize away fetching

  693.      columns that are not needed for the <code class="literal">RETURNING</code> calculation,

  694.      you should insert nulls in those column positions, or else generate a

  695.      <code class="structfield">fdw_scan_tlist</code> list with those columns omitted.

  696.     </p><p>

  697.      Whether the query has the clause or not, the query's reported row count

  698.      must be incremented by the FDW itself.  When the query doesn't have the

  699.      clause, the FDW must also increment the row count for the

  700.      <code class="structname">ForeignScanState</code> node in the <code class="command">EXPLAIN ANALYZE</code>

  701.      case.

  702.     </p><p>

  703.      If the <code class="function">IterateDirectModify</code> pointer is set to

  704.      <code class="literal">NULL</code>, no attempts to execute a direct modification on the

  705.      remote server are taken.

  706.     </p><p>

  707. </p><pre class="programlisting">

  708. void

  709. EndDirectModify(ForeignScanState *node);

  710. </pre><p>

  711. 

  712.      Clean up following a direct modification on the remote server.  It is

  713.      normally not important to release palloc'd memory, but for example open

  714.      files and connections to the remote server should be cleaned up.

  715.     </p><p>

  716.      If the <code class="function">EndDirectModify</code> pointer is set to

  717.      <code class="literal">NULL</code>, no attempts to execute a direct modification on the

  718.      remote server are taken.

  719.     </p></div><div class="sect2" id="FDW-CALLBACKS-ROW-LOCKING"><div class="titlepage"><div><div><h3 class="title">56.2.5. FDW Routines for Row Locking</h3></div></div></div><p>

  720.      If an FDW wishes to support <em class="firstterm">late row locking</em> (as described

  721.      in <a class="xref" href="fdw-row-locking.html" title="56.5. Row Locking in Foreign Data Wrappers">Section 56.5</a>), it must provide the following

  722.      callback functions:

  723.     </p><p>

  724. </p><pre class="programlisting">

  725. RowMarkType

  726. GetForeignRowMarkType(RangeTblEntry *rte,

  727.                       LockClauseStrength strength);

  728. </pre><p>

  729. 

  730.      Report which row-marking option to use for a foreign table.

  731.      <code class="literal">rte</code> is the <code class="structname">RangeTblEntry</code> node for the table

  732.      and <code class="literal">strength</code> describes the lock strength requested by the

  733.      relevant <code class="literal">FOR UPDATE/SHARE</code> clause, if any.  The result must be

  734.      a member of the <code class="literal">RowMarkType</code> enum type.

  735.     </p><p>

  736.      This function is called during query planning for each foreign table that

  737.      appears in an <code class="command">UPDATE</code>, <code class="command">DELETE</code>, or <code class="command">SELECT

  738.      FOR UPDATE/SHARE</code> query and is not the target of <code class="command">UPDATE</code>

  739.      or <code class="command">DELETE</code>.

  740.     </p><p>

  741.      If the <code class="function">GetForeignRowMarkType</code> pointer is set to

  742.      <code class="literal">NULL</code>, the <code class="literal">ROW_MARK_COPY</code> option is always used.

  743.      (This implies that <code class="function">RefetchForeignRow</code> will never be called,

  744.      so it need not be provided either.)

  745.     </p><p>

  746.      See <a class="xref" href="fdw-row-locking.html" title="56.5. Row Locking in Foreign Data Wrappers">Section 56.5</a> for more information.

  747.     </p><p>

  748. </p><pre class="programlisting">

  749. void

  750. RefetchForeignRow(EState *estate,

  751.                   ExecRowMark *erm,

  752.                   Datum rowid,

  753.                   TupleTableSlot *slot,

  754.                   bool *updated);

  755. </pre><p>

  756. 

  757.      Re-fetch one tuple slot from the foreign table, after locking it if required.

  758.      <code class="literal">estate</code> is global execution state for the query.

  759.      <code class="literal">erm</code> is the <code class="structname">ExecRowMark</code> struct describing

  760.      the target foreign table and the row lock type (if any) to acquire.

  761.      <code class="literal">rowid</code> identifies the tuple to be fetched.

  762.      <code class="literal">slot</code> contains nothing useful upon call, but can be used to

  763.      hold the returned tuple. <code class="literal">updated</code> is an output parameter.

  764.     </p><p>

  765.      This function should store the tuple into the provided slot, or clear it if

  766.      the row lock couldn't be obtained.  The row lock type to acquire is

  767.      defined by <code class="literal">erm-&gt;markType</code>, which is the value

  768.      previously returned by <code class="function">GetForeignRowMarkType</code>.

  769.      (<code class="literal">ROW_MARK_REFERENCE</code> means to just re-fetch the tuple

  770.      without acquiring any lock, and <code class="literal">ROW_MARK_COPY</code> will

  771.      never be seen by this routine.)

  772.     </p><p>

  773.      In addition, <code class="literal">*updated</code> should be set to <code class="literal">true</code>

  774.      if what was fetched was an updated version of the tuple rather than

  775.      the same version previously obtained.  (If the FDW cannot be sure about

  776.      this, always returning <code class="literal">true</code> is recommended.)

  777.     </p><p>

  778.      Note that by default, failure to acquire a row lock should result in

  779.      raising an error; returning with an empty slot is only appropriate if

  780.      the <code class="literal">SKIP LOCKED</code> option is specified

  781.      by <code class="literal">erm-&gt;waitPolicy</code>.

  782.     </p><p>

  783.      The <code class="literal">rowid</code> is the <code class="structfield">ctid</code> value previously read

  784.      for the row to be re-fetched.  Although the <code class="literal">rowid</code> value is

  785.      passed as a <code class="type">Datum</code>, it can currently only be a <code class="type">tid</code>.  The

  786.      function API is chosen in hopes that it may be possible to allow other

  787.      data types for row IDs in future.

  788.     </p><p>

  789.      If the <code class="function">RefetchForeignRow</code> pointer is set to

  790.      <code class="literal">NULL</code>, attempts to re-fetch rows will fail

  791.      with an error message.

  792.     </p><p>

  793.      See <a class="xref" href="fdw-row-locking.html" title="56.5. Row Locking in Foreign Data Wrappers">Section 56.5</a> for more information.

  794.     </p><p>

  795. </p><pre class="programlisting">

  796. bool

  797. RecheckForeignScan(ForeignScanState *node,

  798.                    TupleTableSlot *slot);

  799. </pre><p>

  800.      Recheck that a previously-returned tuple still matches the relevant

  801.      scan and join qualifiers, and possibly provide a modified version of

  802.      the tuple.  For foreign data wrappers which do not perform join pushdown,

  803.      it will typically be more convenient to set this to <code class="literal">NULL</code> and

  804.      instead set <code class="structfield">fdw_recheck_quals</code> appropriately.

  805.      When outer joins are pushed down, however, it isn't sufficient to

  806.      reapply the checks relevant to all the base tables to the result tuple,

  807.      even if all needed attributes are present, because failure to match some

  808.      qualifier might result in some attributes going to NULL, rather than in

  809.      no tuple being returned.  <code class="literal">RecheckForeignScan</code> can recheck

  810.      qualifiers and return true if they are still satisfied and false

  811.      otherwise, but it can also store a replacement tuple into the supplied

  812.      slot.

  813.     </p><p>

  814.      To implement join pushdown, a foreign data wrapper will typically

  815.      construct an alternative local join plan which is used only for

  816.      rechecks; this will become the outer subplan of the

  817.      <code class="literal">ForeignScan</code>.  When a recheck is required, this subplan

  818.      can be executed and the resulting tuple can be stored in the slot.

  819.      This plan need not be efficient since no base table will return more

  820.      than one row; for example, it may implement all joins as nested loops.

  821.      The function <code class="literal">GetExistingLocalJoinPath</code> may be used to search

  822.      existing paths for a suitable local join path, which can be used as the

  823.      alternative local join plan.  <code class="literal">GetExistingLocalJoinPath</code>

  824.      searches for an unparameterized path in the path list of the specified

  825.      join relation.  (If it does not find such a path, it returns NULL, in

  826.      which case a foreign data wrapper may build the local path by itself or

  827.      may choose not to create access paths for that join.)

  828.     </p></div><div class="sect2" id="FDW-CALLBACKS-EXPLAIN"><div class="titlepage"><div><div><h3 class="title">56.2.6. FDW Routines for <code class="command">EXPLAIN</code></h3></div></div></div><p>

  829. </p><pre class="programlisting">

  830. void

  831. ExplainForeignScan(ForeignScanState *node,

  832.                    ExplainState *es);

  833. </pre><p>

  834. 

  835.      Print additional <code class="command">EXPLAIN</code> output for a foreign table scan.

  836.      This function can call <code class="function">ExplainPropertyText</code> and

  837.      related functions to add fields to the <code class="command">EXPLAIN</code> output.

  838.      The flag fields in <code class="literal">es</code> can be used to determine what to

  839.      print, and the state of the <code class="structname">ForeignScanState</code> node

  840.      can be inspected to provide run-time statistics in the <code class="command">EXPLAIN

  841.      ANALYZE</code> case.

  842.     </p><p>

  843.      If the <code class="function">ExplainForeignScan</code> pointer is set to

  844.      <code class="literal">NULL</code>, no additional information is printed during

  845.      <code class="command">EXPLAIN</code>.

  846.     </p><p>

  847. </p><pre class="programlisting">

  848. void

  849. ExplainForeignModify(ModifyTableState *mtstate,

  850.                      ResultRelInfo *rinfo,

  851.                      List *fdw_private,

  852.                      int subplan_index,

  853.                      struct ExplainState *es);

  854. </pre><p>

  855. 

  856.      Print additional <code class="command">EXPLAIN</code> output for a foreign table update.

  857.      This function can call <code class="function">ExplainPropertyText</code> and

  858.      related functions to add fields to the <code class="command">EXPLAIN</code> output.

  859.      The flag fields in <code class="literal">es</code> can be used to determine what to

  860.      print, and the state of the <code class="structname">ModifyTableState</code> node

  861.      can be inspected to provide run-time statistics in the <code class="command">EXPLAIN

  862.      ANALYZE</code> case.  The first four arguments are the same as for

  863.      <code class="function">BeginForeignModify</code>.

  864.     </p><p>

  865.      If the <code class="function">ExplainForeignModify</code> pointer is set to

  866.      <code class="literal">NULL</code>, no additional information is printed during

  867.      <code class="command">EXPLAIN</code>.

  868.     </p><p>

  869. </p><pre class="programlisting">

  870. void

  871. ExplainDirectModify(ForeignScanState *node,

  872.                     ExplainState *es);

  873. </pre><p>

  874. 

  875.      Print additional <code class="command">EXPLAIN</code> output for a direct modification

  876.      on the remote server.

  877.      This function can call <code class="function">ExplainPropertyText</code> and

  878.      related functions to add fields to the <code class="command">EXPLAIN</code> output.

  879.      The flag fields in <code class="literal">es</code> can be used to determine what to

  880.      print, and the state of the <code class="structname">ForeignScanState</code> node

  881.      can be inspected to provide run-time statistics in the <code class="command">EXPLAIN

  882.      ANALYZE</code> case.

  883.     </p><p>

  884.      If the <code class="function">ExplainDirectModify</code> pointer is set to

  885.      <code class="literal">NULL</code>, no additional information is printed during

  886.      <code class="command">EXPLAIN</code>.

  887.     </p></div><div class="sect2" id="FDW-CALLBACKS-ANALYZE"><div class="titlepage"><div><div><h3 class="title">56.2.7. FDW Routines for <code class="command">ANALYZE</code></h3></div></div></div><p>

  888. </p><pre class="programlisting">

  889. bool

  890. AnalyzeForeignTable(Relation relation,

  891.                     AcquireSampleRowsFunc *func,

  892.                     BlockNumber *totalpages);

  893. </pre><p>

  894. 

  895.      This function is called when <a class="xref" href="sql-analyze.html" title="ANALYZE"><span class="refentrytitle">ANALYZE</span></a> is executed on

  896.      a foreign table.  If the FDW can collect statistics for this

  897.      foreign table, it should return <code class="literal">true</code>, and provide a pointer

  898.      to a function that will collect sample rows from the table in

  899.      <em class="parameter"><code>func</code></em>, plus the estimated size of the table in pages in

  900.      <em class="parameter"><code>totalpages</code></em>.  Otherwise, return <code class="literal">false</code>.

  901.     </p><p>

  902.      If the FDW does not support collecting statistics for any tables, the

  903.      <code class="function">AnalyzeForeignTable</code> pointer can be set to <code class="literal">NULL</code>.

  904.     </p><p>

  905.      If provided, the sample collection function must have the signature

  906. </p><pre class="programlisting">

  907. int

  908. AcquireSampleRowsFunc(Relation relation,

  909.                       int elevel,

  910.                       HeapTuple *rows,

  911.                       int targrows,

  912.                       double *totalrows,

  913.                       double *totaldeadrows);

  914. </pre><p>

  915. 

  916.      A random sample of up to <em class="parameter"><code>targrows</code></em> rows should be collected

  917.      from the table and stored into the caller-provided <em class="parameter"><code>rows</code></em>

  918.      array.  The actual number of rows collected must be returned.  In

  919.      addition, store estimates of the total numbers of live and dead rows in

  920.      the table into the output parameters <em class="parameter"><code>totalrows</code></em> and

  921.      <em class="parameter"><code>totaldeadrows</code></em>.  (Set <em class="parameter"><code>totaldeadrows</code></em> to zero

  922.      if the FDW does not have any concept of dead rows.)

  923.     </p></div><div class="sect2" id="FDW-CALLBACKS-IMPORT"><div class="titlepage"><div><div><h3 class="title">56.2.8. FDW Routines for <code class="command">IMPORT FOREIGN SCHEMA</code></h3></div></div></div><p>

  924. </p><pre class="programlisting">

  925. List *

  926. ImportForeignSchema(ImportForeignSchemaStmt *stmt, Oid serverOid);

  927. </pre><p>

  928. 

  929.      Obtain a list of foreign table creation commands.  This function is

  930.      called when executing <a class="xref" href="sql-importforeignschema.html" title="IMPORT FOREIGN SCHEMA"><span class="refentrytitle">IMPORT FOREIGN SCHEMA</span></a>, and is

  931.      passed the parse tree for that statement, as well as the OID of the

  932.      foreign server to use.  It should return a list of C strings, each of

  933.      which must contain a <a class="xref" href="sql-createforeigntable.html" title="CREATE FOREIGN TABLE"><span class="refentrytitle">CREATE FOREIGN TABLE</span></a> command.

  934.      These strings will be parsed and executed by the core server.

  935.     </p><p>

  936.      Within the <code class="structname">ImportForeignSchemaStmt</code> struct,

  937.      <code class="structfield">remote_schema</code> is the name of the remote schema from

  938.      which tables are to be imported.

  939.      <code class="structfield">list_type</code> identifies how to filter table names:

  940.      <code class="literal">FDW_IMPORT_SCHEMA_ALL</code> means that all tables in the remote

  941.      schema should be imported (in this case <code class="structfield">table_list</code> is

  942.      empty), <code class="literal">FDW_IMPORT_SCHEMA_LIMIT_TO</code> means to include only

  943.      tables listed in <code class="structfield">table_list</code>,

  944.      and <code class="literal">FDW_IMPORT_SCHEMA_EXCEPT</code> means to exclude the tables

  945.      listed in <code class="structfield">table_list</code>.

  946.      <code class="structfield">options</code> is a list of options used for the import process.

  947.      The meanings of the options are up to the FDW.

  948.      For example, an FDW could use an option to define whether the

  949.      <code class="literal">NOT NULL</code> attributes of columns should be imported.

  950.      These options need not have anything to do with those supported by the

  951.      FDW as database object options.

  952.     </p><p>

  953.      The FDW may ignore the <code class="structfield">local_schema</code> field of

  954.      the <code class="structname">ImportForeignSchemaStmt</code>, because the core server

  955.      will automatically insert that name into the parsed <code class="command">CREATE

  956.      FOREIGN TABLE</code> commands.

  957.     </p><p>

  958.      The FDW does not have to concern itself with implementing the filtering

  959.      specified by <code class="structfield">list_type</code> and <code class="structfield">table_list</code>,

  960.      either, as the core server will automatically skip any returned commands

  961.      for tables excluded according to those options.  However, it's often

  962.      useful to avoid the work of creating commands for excluded tables in the

  963.      first place.  The function <code class="function">IsImportableForeignTable()</code> may be

  964.      useful to test whether a given foreign-table name will pass the filter.

  965.     </p><p>

  966.      If the FDW does not support importing table definitions, the

  967.      <code class="function">ImportForeignSchema</code> pointer can be set to <code class="literal">NULL</code>.

  968.     </p></div><div class="sect2" id="FDW-CALLBACKS-PARALLEL"><div class="titlepage"><div><div><h3 class="title">56.2.9. FDW Routines for Parallel Execution</h3></div></div></div><p>

  969.      A <code class="structname">ForeignScan</code> node can, optionally, support parallel

  970.      execution.  A parallel <code class="structname">ForeignScan</code> will be executed

  971.      in multiple processes and must return each row exactly once across

  972.      all cooperating processes.  To do this, processes can coordinate through

  973.      fixed-size chunks of dynamic shared memory.  This shared memory is not

  974.      guaranteed to be mapped at the same address in every process, so it

  975.      must not contain pointers.  The following functions are all optional,

  976.      but most are required if parallel execution is to be supported.

  977.     </p><p>

  978. </p><pre class="programlisting">

  979. bool

  980. IsForeignScanParallelSafe(PlannerInfo *root, RelOptInfo *rel,

  981.                           RangeTblEntry *rte);

  982. </pre><p>

  983.     Test whether a scan can be performed within a parallel worker.  This

  984.     function will only be called when the planner believes that a parallel

  985.     plan might be possible, and should return true if it is safe for that scan

  986.     to run within a parallel worker.  This will generally not be the case if

  987.     the remote data source has transaction semantics, unless the worker's

  988.     connection to the data can somehow be made to share the same transaction

  989.     context as the leader.

  990.     </p><p>

  991.     If this function is not defined, it is assumed that the scan must take

  992.     place within the parallel leader.  Note that returning true does not mean

  993.     that the scan itself can be done in parallel, only that the scan can be

  994.     performed within a parallel worker.  Therefore, it can be useful to define

  995.     this method even when parallel execution is not supported.

  996.     </p><p>

  997. </p><pre class="programlisting">

  998. Size

  999. EstimateDSMForeignScan(ForeignScanState *node, ParallelContext *pcxt);

  1000. </pre><p>

  1001.     Estimate the amount of dynamic shared memory that will be required

  1002.     for parallel operation.  This may be higher than the amount that will

  1003.     actually be used, but it must not be lower.  The return value is in bytes.

  1004.     This function is optional, and can be omitted if not needed; but if it

  1005.     is omitted, the next three functions must be omitted as well, because

  1006.     no shared memory will be allocated for the FDW's use.

  1007.     </p><p>

  1008. </p><pre class="programlisting">

  1009. void

  1010. InitializeDSMForeignScan(ForeignScanState *node, ParallelContext *pcxt,

  1011.                          void *coordinate);

  1012. </pre><p>

  1013.     Initialize the dynamic shared memory that will be required for parallel

  1014.     operation.  <code class="literal">coordinate</code> points to a shared memory area of

  1015.     size equal to the return value of <code class="function">EstimateDSMForeignScan</code>.

  1016.     This function is optional, and can be omitted if not needed.

  1017.    </p><p>

  1018. </p><pre class="programlisting">

  1019. void

  1020. ReInitializeDSMForeignScan(ForeignScanState *node, ParallelContext *pcxt,

  1021.                            void *coordinate);

  1022. </pre><p>

  1023.     Re-initialize the dynamic shared memory required for parallel operation

  1024.     when the foreign-scan plan node is about to be re-scanned.

  1025.     This function is optional, and can be omitted if not needed.

  1026.     Recommended practice is that this function reset only shared state,

  1027.     while the <code class="function">ReScanForeignScan</code> function resets only local

  1028.     state.  Currently, this function will be called

  1029.     before <code class="function">ReScanForeignScan</code>, but it's best not to rely on

  1030.     that ordering.

  1031.    </p><p>

  1032. </p><pre class="programlisting">

  1033. void

  1034. InitializeWorkerForeignScan(ForeignScanState *node, shm_toc *toc,

  1035.                             void *coordinate);

  1036. </pre><p>

  1037.     Initialize a parallel worker's local state based on the shared state

  1038.     set up by the leader during <code class="function">InitializeDSMForeignScan</code>.

  1039.     This function is optional, and can be omitted if not needed.

  1040.    </p><p>

  1041. </p><pre class="programlisting">

  1042. void

  1043. ShutdownForeignScan(ForeignScanState *node);

  1044. </pre><p>

  1045.     Release resources when it is anticipated the node will not be executed

  1046.     to completion.  This is not called in all cases; sometimes,

  1047.     <code class="literal">EndForeignScan</code> may be called without this function having

  1048.     been called first.  Since the DSM segment used by parallel query is

  1049.     destroyed just after this callback is invoked, foreign data wrappers that

  1050.     wish to take some action before the DSM segment goes away should implement

  1051.     this method.

  1052.    </p></div><div class="sect2" id="FDW-CALLBACKS-REPARAMETERIZE-PATHS"><div class="titlepage"><div><div><h3 class="title">56.2.10. FDW Routines for Reparameterization of Paths</h3></div></div></div><p>

  1053. </p><pre class="programlisting">

  1054. List *

  1055. ReparameterizeForeignPathByChild(PlannerInfo *root, List *fdw_private,

  1056.                                  RelOptInfo *child_rel);

  1057. </pre><p>

  1058.     This function is called while converting a path parameterized by the

  1059.     top-most parent of the given child relation <code class="literal">child_rel</code> to be

  1060.     parameterized by the child relation. The function is used to reparameterize

  1061.     any paths or translate any expression nodes saved in the given

  1062.     <code class="literal">fdw_private</code> member of a <code class="structname">ForeignPath</code>. The

  1063.     callback may use <code class="literal">reparameterize_path_by_child</code>,

  1064.     <code class="literal">adjust_appendrel_attrs</code> or

  1065.     <code class="literal">adjust_appendrel_attrs_multilevel</code> as required.

  1066.     </p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="fdw-functions.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="fdwhandler.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="fdw-helpers.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">56.1. Foreign Data Wrapper Functions </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> 56.3. Foreign Data Wrapper Helper Functions</td></tr></table></div></body></html>
上海开阖软件有限公司 沪ICP备12045867号-1