[CF 5467] v10 - speedup COPY TO for partitioned table

This branch was automatically generated by a robot using patches from an
email thread registered at:

https://commitfest.postgresql.org/patch/5467

The branch will be overwritten each time a new patch version is posted to
the thread, and also periodically to check for bitrot caused by changes
on the master branch.

Patch(es): https://www.postgresql.org/message-id/CACJufxEp7btcpXiroNXcCvjwDwVT5-A6-A9ryWnDJiYfB7PO=Q@mail.gmail.com
Author(s): Jian He

Author: Commitfest Bot

doc/src/sgml/ref/copy.sgml

@@ -521,13 +521,13 @@ COPY <replaceable class="parameter">count</replaceable>
 
    <para>
     <command>COPY TO</command> can be used with plain
-    tables and populated materialized views.
-    For example,
+    tables, populated materialized views and partitioned tables.
+    For example, if <replaceable class="parameter">table</replaceable> is not partitioned table,
     <literal>COPY <replaceable class="parameter">table</replaceable>
     TO</literal> copies the same rows as
     <literal>SELECT * FROM ONLY <replaceable class="parameter">table</replaceable></literal>.
     However it doesn't directly support other relation types,
-    such as partitioned tables, inheritance child tables, or views.
+    such as inheritance child tables, or views.
     To copy all rows from such relations, use <literal>COPY (SELECT * FROM
     <replaceable class="parameter">table</replaceable>) TO</literal>.
    </para>

src/backend/commands/copyto.c

@@ -19,6 +19,8 @@
 #include <sys/stat.h>
 
 #include "access/tableam.h"
+#include "access/table.h"
+#include "catalog/pg_inherits.h"
 #include "commands/copyapi.h"
 #include "commands/progress.h"
 #include "executor/execdesc.h"
@@ -82,6 +84,7 @@ typedef struct CopyToStateData
 	List	   *attnumlist;		/* integer list of attnums to copy */
 	char	   *filename;		/* filename, or NULL for STDOUT */
 	bool		is_program;		/* is 'filename' a program to popen? */
+	List	   *partitions;		/* oid list of partition oid for copy to */
 	copy_data_dest_cb data_dest_cb; /* function for writing data */
 
 	CopyFormatOptions opts;
@@ -116,6 +119,8 @@ static void CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot);
 static void CopyAttributeOutText(CopyToState cstate, const char *string);
 static void CopyAttributeOutCSV(CopyToState cstate, const char *string,
 								bool use_quote);
+static void CopyThisRelTo(CopyToState cstate, Relation rel,
+						  Relation root_rel, uint64 *processed);
 
 /* built-in format-specific routines */
 static void CopyToTextLikeStart(CopyToState cstate, TupleDesc tupDesc);
@@ -643,6 +648,8 @@ BeginCopyTo(ParseState *pstate,
 		PROGRESS_COPY_COMMAND_TO,
 		0
 	};
+	List	   *children = NIL;
+	List	   *scan_oids = NIL;
 
 	if (rel != NULL && rel->rd_rel->relkind != RELKIND_RELATION)
 	{
@@ -673,11 +680,35 @@ BeginCopyTo(ParseState *pstate,
 					 errmsg("cannot copy from sequence \"%s\"",
 							RelationGetRelationName(rel))));
 		else if (rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE)
-			ereport(ERROR,
-					(errcode(ERRCODE_WRONG_OBJECT_TYPE),
-					 errmsg("cannot copy from partitioned table \"%s\"",
-							RelationGetRelationName(rel)),
-					 errhint("Try the COPY (SELECT ...) TO variant.")));
+		{
+			children = find_all_inheritors(RelationGetRelid(rel),
+										   AccessShareLock,
+										   NULL);
+
+			foreach_oid(childreloid, children)
+			{
+				char		 relkind = get_rel_relkind(childreloid);
+
+				if (relkind == RELKIND_FOREIGN_TABLE)
+				{
+					char	   *relation_name;
+
+					relation_name = get_rel_name(childreloid);
+					ereport(ERROR,
+							errcode(ERRCODE_WRONG_OBJECT_TYPE),
+							errmsg("cannot copy from foreign table \"%s\"", relation_name),
+							errdetail("Partition \"%s\" is a foreign table in the partitioned table \"%s.%s\"",
+									  relation_name, RelationGetRelationName(rel),
+									  get_namespace_name(rel->rd_rel->relnamespace)),
+							errhint("Try the COPY (SELECT ...) TO variant."));
+				}
+
+				if (RELKIND_HAS_PARTITIONS(relkind))
+					continue;
+
+				scan_oids = lappend_oid(scan_oids, childreloid);
+			}
+		}
 		else
 			ereport(ERROR,
 					(errcode(ERRCODE_WRONG_OBJECT_TYPE),
@@ -713,6 +744,7 @@ BeginCopyTo(ParseState *pstate,
 		cstate->rel = rel;
 
 		tupDesc = RelationGetDescr(cstate->rel);
+		cstate->partitions = list_copy(scan_oids);
 	}
 	else
 	{
@@ -1030,7 +1062,7 @@ DoCopyTo(CopyToState cstate)
 	TupleDesc	tupDesc;
 	int			num_phys_attrs;
 	ListCell   *cur;
-	uint64		processed;
+	uint64		processed = 0;
 
 	if (fe_copy)
 		SendCopyBegin(cstate);
@@ -1068,36 +1100,25 @@ DoCopyTo(CopyToState cstate)
 
 	cstate->routine->CopyToStart(cstate, tupDesc);
 
-	if (cstate->rel)
+	/*
+	 * if COPY TO source table is a partitioned table, then open each
+	 * partition and process each individual partition.
+	 */
+	if (cstate->rel && cstate->rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE)
 	{
-		TupleTableSlot *slot;
-		TableScanDesc scandesc;
-
-		scandesc = table_beginscan(cstate->rel, GetActiveSnapshot(), 0, NULL);
-		slot = table_slot_create(cstate->rel, NULL);
-
-		processed = 0;
-		while (table_scan_getnextslot(scandesc, ForwardScanDirection, slot))
+		foreach_oid(scan_oid, cstate->partitions)
 		{
-			CHECK_FOR_INTERRUPTS();
+			Relation		scan_rel;
 
-			/* Deconstruct the tuple ... */
-			slot_getallattrs(slot);
+			scan_rel = table_open(scan_oid, AccessShareLock);
 
-			/* Format and send the data */
-			CopyOneRowTo(cstate, slot);
+			CopyThisRelTo(cstate, scan_rel, cstate->rel, &processed);
 
-			/*
-			 * Increment the number of processed tuples, and report the
-			 * progress.
-			 */
-			pgstat_progress_update_param(PROGRESS_COPY_TUPLES_PROCESSED,
-										 ++processed);
+			table_close(scan_rel, AccessShareLock);
 		}
-
-		ExecDropSingleTupleTableSlot(slot);
-		table_endscan(scandesc);
 	}
+	else if (cstate->rel)
+		CopyThisRelTo(cstate, cstate->rel, NULL, &processed);
 	else
 	{
 		/* run the plan --- the dest receiver will send tuples */
@@ -1115,6 +1136,71 @@ DoCopyTo(CopyToState cstate)
 	return processed;
 }
 
+/*
+ * rel: the relation from which the actual data will be copied.
+ * root_rel: if not NULL, it indicates that we are copying partitioned relation
+ * data to the destination, and "rel" is the partition of "root_rel".
+ * processed: number of tuples processed.
+*/
+static void
+CopyThisRelTo(CopyToState cstate, Relation rel, Relation root_rel,
+			  uint64 *processed)
+{
+	TupleTableSlot *slot;
+	TableScanDesc scandesc;
+	AttrMap    		*map	= NULL;
+	TupleTableSlot  *root_slot = NULL;
+	TupleTableSlot  *original_slot = NULL;
+	TupleDesc		scan_tupdesc;
+	TupleDesc 		rootdesc	= NULL;
+
+	scan_tupdesc = RelationGetDescr(rel);
+	scandesc = table_beginscan(rel, GetActiveSnapshot(), 0, NULL);
+	slot = table_slot_create(rel, NULL);
+
+	/*
+	 * A partition's rowtype might differ from the root table's.
+	 * Since we are export partitioned table data here,
+	 * we must convert it back to the root table's rowtype.
+	*/
+	if (root_rel != NULL)
+	{
+		rootdesc = RelationGetDescr(root_rel);
+		map = build_attrmap_by_name_if_req(rootdesc, scan_tupdesc, false);
+	}
+
+	while (table_scan_getnextslot(scandesc, ForwardScanDirection, slot))
+	{
+		CHECK_FOR_INTERRUPTS();
+
+		/* Deconstruct the tuple ... */
+		if (map != NULL)
+		{
+			original_slot = slot;
+			root_slot = MakeSingleTupleTableSlot(rootdesc, &TTSOpsBufferHeapTuple);
+			slot = execute_attr_map_slot(map, slot, root_slot);
+		}
+		else
+			slot_getallattrs(slot);
+
+		/* Format and send the data */
+		CopyOneRowTo(cstate, slot);
+
+		/*
+		 * Increment the number of processed tuples, and report the
+		 * progress.
+		 */
+		pgstat_progress_update_param(PROGRESS_COPY_TUPLES_PROCESSED,
+									 ++(*processed));
+
+		if (original_slot != NULL)
+			ExecDropSingleTupleTableSlot(original_slot);
+	}
+
+	ExecDropSingleTupleTableSlot(slot);
+	table_endscan(scandesc);
+}
+
 /*
  * Emit one row during DoCopyTo().
  */

src/test/regress/expected/copy.out

@@ -350,3 +350,21 @@ COPY copytest_mv(id) TO stdout WITH (header);
 id
 1
 DROP MATERIALIZED VIEW copytest_mv;
+-- Tests for COPY TO with partitioned tables.
+CREATE TABLE pp (id int,val int) PARTITION BY RANGE (id);
+CREATE TABLE pp_1 (val int, id int) PARTITION BY RANGE (id);
+CREATE TABLE pp_2 (val int, id int) PARTITION BY RANGE (id);
+ALTER TABLE pp ATTACH PARTITION pp_1 FOR VALUES FROM (1) TO (5);
+ALTER TABLE pp ATTACH PARTITION pp_2 FOR VALUES FROM (5) TO (10);
+CREATE TABLE pp_15 PARTITION OF pp_1 FOR VALUES FROM (1) TO (5);
+CREATE TABLE pp_510 PARTITION OF pp_2 FOR VALUES FROM (5) TO (10);
+INSERT INTO pp SELECT g, 10 + g FROM generate_series(1,6) g;
+COPY pp TO stdout(header);
+id	val
+1	11
+2	12
+3	13
+4	14
+5	15
+6	16
+DROP TABLE PP;

src/test/regress/sql/copy.sql

@@ -375,3 +375,18 @@ COPY copytest_mv(id) TO stdout WITH (header);
 REFRESH MATERIALIZED VIEW copytest_mv;
 COPY copytest_mv(id) TO stdout WITH (header);
 DROP MATERIALIZED VIEW copytest_mv;
+
+-- Tests for COPY TO with partitioned tables.
+CREATE TABLE pp (id int,val int) PARTITION BY RANGE (id);
+CREATE TABLE pp_1 (val int, id int) PARTITION BY RANGE (id);
+CREATE TABLE pp_2 (val int, id int) PARTITION BY RANGE (id);
+ALTER TABLE pp ATTACH PARTITION pp_1 FOR VALUES FROM (1) TO (5);
+ALTER TABLE pp ATTACH PARTITION pp_2 FOR VALUES FROM (5) TO (10);
+
+CREATE TABLE pp_15 PARTITION OF pp_1 FOR VALUES FROM (1) TO (5);
+CREATE TABLE pp_510 PARTITION OF pp_2 FOR VALUES FROM (5) TO (10);
+
+INSERT INTO pp SELECT g, 10 + g FROM generate_series(1,6) g;
+
+COPY pp TO stdout(header);
+DROP TABLE PP;