Add the ability to create indexes 'concurrently', that is, without

blocking concurrent writes to the table.  Greg Stark, with a little help
from Tom Lane.

doc/src/sgml/catalogs.sgml

@@ -1,4 +1,4 @@
-<!-- $PostgreSQL: pgsql/doc/src/sgml/catalogs.sgml,v 2.129 2006/07/31 20:08:55 tgl Exp $ -->
+<!-- $PostgreSQL: pgsql/doc/src/sgml/catalogs.sgml,v 2.130 2006/08/25 04:06:44 tgl Exp $ -->
 <!--
  Documentation of the system catalogs, directed toward PostgreSQL developers
  -->
@@ -2420,6 +2420,17 @@
       <entry>If true, the table was last clustered on this index</entry>
      </row>
 
+     <row>
+      <entry><structfield>indisvalid</structfield></entry>
+      <entry><type>bool</type></entry>
+      <entry></entry>
+      <entry>If true, the index is currently valid for queries.
+       False means the index is possibly incomplete: it must still be
+       inserted into by INSERT/UPDATE operations, but it cannot safely be
+       used for queries, and if it is unique, the uniqueness shouldn't be
+       relied on either.</entry>
+     </row>
+
      <row>
       <entry><structfield>indkey</structfield></entry>
       <entry><type>int2vector</type></entry>

doc/src/sgml/indexam.sgml

@@ -1,4 +1,4 @@
-<!-- $PostgreSQL: pgsql/doc/src/sgml/indexam.sgml,v 2.16 2006/07/31 20:08:59 tgl Exp $ -->
+<!-- $PostgreSQL: pgsql/doc/src/sgml/indexam.sgml,v 2.17 2006/08/25 04:06:44 tgl Exp $ -->
 
 <chapter id="indexam">
  <title>Index Access Method Interface Definition</title>
@@ -648,6 +648,16 @@ amrestrpos (IndexScanDesc scan);
     </itemizedlist>
   </para>
 
+  <para>
+   Furthermore, immediately before raising a uniqueness violation
+   according to the above rules, the access method must recheck the
+   liveness of the row being inserted.  If it is committed dead then
+   no error should be raised.  (This case cannot occur during the
+   ordinary scenario of inserting a row that's just been created by
+   the current transaction.  It can happen during
+   <command>CREATE UNIQUE INDEX CONCURRENTLY</>, however.) 
+  </para>
+
   <para>
    We require the index access method to apply these tests itself, which
    means that it must reach into the heap to check the commit status of

doc/src/sgml/indices.sgml

@@ -1,4 +1,4 @@
-<!-- $PostgreSQL: pgsql/doc/src/sgml/indices.sgml,v 1.57 2006/05/24 11:01:39 teodor Exp $ -->
+<!-- $PostgreSQL: pgsql/doc/src/sgml/indices.sgml,v 1.58 2006/08/25 04:06:44 tgl Exp $ -->
 
 <chapter id="indexes">
  <title id="indexes-title">Indexes</title>
@@ -90,6 +90,17 @@ CREATE INDEX test1_id_index ON test1 (id);
    significantly speed up queries with joins.
   </para>
 
+  <para>
+   Creating an index on a large table can take a long time.  By default,
+   <productname>PostgreSQL</productname> allows reads (selects) to occur
+   on the table in parallel with index creation, but writes (inserts,
+   updates, deletes) are blocked until the index build is finished.
+   It is possible to allow writes to occur in parallel with index
+   creation, but there are several caveats to be aware of &mdash;
+   for more information see <xref linkend="SQL-CREATEINDEX-CONCURRENTLY"
+   endterm="SQL-CREATEINDEX-CONCURRENTLY-title">.
+  </para>
+
   <para>
    After an index is created, the system has to keep it synchronized with the
    table.  This adds overhead to data manipulation operations.

doc/src/sgml/mvcc.sgml

@@ -1,4 +1,4 @@
-<!-- $PostgreSQL: pgsql/doc/src/sgml/mvcc.sgml,v 2.56 2006/04/23 03:39:52 momjian Exp $ -->
+<!-- $PostgreSQL: pgsql/doc/src/sgml/mvcc.sgml,v 2.57 2006/08/25 04:06:45 tgl Exp $ -->
 
  <chapter id="mvcc">
   <title>Concurrency Control</title>
@@ -622,7 +622,8 @@ SELECT SUM(value) FROM mytab WHERE class = 2;
 	</para>
 
 	<para>
-	 Acquired by <command>VACUUM</command> (without <option>FULL</option>).
+	 Acquired by <command>VACUUM</command> (without <option>FULL</option>)
+	 and by <command>CREATE INDEX CONCURRENTLY</>.
 	</para>
        </listitem>
       </varlistentry>
@@ -641,7 +642,8 @@ SELECT SUM(value) FROM mytab WHERE class = 2;
 	</para>
 
 	<para>
-	 Acquired by <command>CREATE INDEX</command>.
+	 Acquired by <command>CREATE INDEX</command>
+	 (without <option>CONCURRENTLY</option>).
 	</para>
        </listitem>
       </varlistentry>

doc/src/sgml/ref/create_index.sgml

@@ -1,5 +1,5 @@
 <!--
-$PostgreSQL: pgsql/doc/src/sgml/ref/create_index.sgml,v 1.55 2006/07/11 21:05:57 tgl Exp $
+$PostgreSQL: pgsql/doc/src/sgml/ref/create_index.sgml,v 1.56 2006/08/25 04:06:45 tgl Exp $
 PostgreSQL documentation
 -->
 
@@ -20,7 +20,7 @@ PostgreSQL documentation
 
  <refsynopsisdiv>
 <synopsis>
-CREATE [ UNIQUE ] INDEX <replaceable class="parameter">name</replaceable> ON <replaceable class="parameter">table</replaceable> [ USING <replaceable class="parameter">method</replaceable> ]
+CREATE [ UNIQUE ] INDEX [ CONCURRENTLY ] <replaceable class="parameter">name</replaceable> ON <replaceable class="parameter">table</replaceable> [ USING <replaceable class="parameter">method</replaceable> ]
     ( { <replaceable class="parameter">column</replaceable> | ( <replaceable class="parameter">expression</replaceable> ) } [ <replaceable class="parameter">opclass</replaceable> ] [, ...] )
     [ WITH ( <replaceable class="PARAMETER">storage_parameter</replaceable> = <replaceable class="PARAMETER">value</replaceable> [, ... ] ) ]
     [ TABLESPACE <replaceable class="parameter">tablespace</replaceable> ]
@@ -110,6 +110,21 @@ CREATE [ UNIQUE ] INDEX <replaceable class="parameter">name</replaceable> ON <re
       </listitem>
      </varlistentry>
 
+     <varlistentry>
+      <term><literal>CONCURRENTLY</literal></term>
+      <listitem>
+       <para>
+        When this option is used, <productname>PostgreSQL</> will build the
+        index without taking any locks that prevent concurrent inserts,
+        updates, or deletes on the table; whereas a standard index build
+        locks out writes (but not reads) on the table until it's done.
+        There are several caveats to be aware of when using this option
+        &mdash; see <xref linkend="SQL-CREATEINDEX-CONCURRENTLY"
+        endterm="SQL-CREATEINDEX-CONCURRENTLY-title">.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry>
       <term><replaceable class="parameter">name</replaceable></term>
       <listitem>
@@ -239,6 +254,82 @@ CREATE [ UNIQUE ] INDEX <replaceable class="parameter">name</replaceable> ON <re
    </variablelist>
 
   </refsect2>
+
+  <refsect2 id="SQL-CREATEINDEX-CONCURRENTLY">
+   <title id="SQL-CREATEINDEX-CONCURRENTLY-title">Building Indexes Concurrently</title>
+
+   <indexterm zone="SQL-CREATEINDEX-CONCURRENTLY">
+   <primary>index</primary>
+   <secondary>building concurrently</secondary>
+   </indexterm>
+
+   <para>
+    Creating an index for a large table can be a long operation. In large data
+    warehousing applications it can easily take hours or even days to build
+    indexes. It's important to understand the impact creating indexes has on a
+    system.
+   </para>
+
+   <para>
+    Normally <productname>PostgreSQL</> locks the table to be indexed against
+    writes and performs the entire index build with a single scan of the
+    table. Other transactions can still read the table, but if they try to
+    insert, update, or delete rows in the table they will block until the
+    index build is finished.
+   </para>
+
+   <para>
+    <productname>PostgreSQL</> also supports building indexes without locking
+    out writes.  This method is invoked by specifying the
+    <literal>CONCURRENTLY</> option of <command>CREATE INDEX</>.
+    When this option is used,
+    <productname>PostgreSQL</> must perform two scans of the table, and in
+    addition it must wait for all existing transactions to terminate.  Thus
+    this method requires more total work than a standard index build and takes
+    significantly longer to complete.  However, since it allows normal
+    operations to continue while the index is built, this method is useful for
+    adding new indexes in a production environment.  Of course, the extra CPU
+    and I/O load imposed by the index creation may slow other operations.
+   </para>
+
+   <para>
+    If a problem arises during the second scan of the table, such as a
+    uniqueness violation in a unique index, the <command>CREATE INDEX</>
+    command will fail but leave behind an <quote>invalid</> index. This index
+    will be ignored for querying purposes because it may be incomplete;
+    however it will still consume update overhead.  The recommended recovery
+    method in such cases is to drop the index and try again to perform
+    <command>CREATE INDEX CONCURRENTLY</>.  (Another possibility is to rebuild
+    the index with <command>REINDEX</>.  However, since <command>REINDEX</>
+    does not support concurrent builds, this option is unlikely to seem
+    attractive.)
+   </para>
+
+   <para>
+    Another caveat when building a unique index concurrently is that the
+    uniqueness constraint is already being enforced against other transactions
+    when the second table scan begins.  This means that constraint violations
+    could be reported in other queries prior to the index becoming available
+    for use, or even in cases where the index build eventually fails.  Also,
+    if a failure does occur in the second scan, the <quote>invalid</> index
+    continues to enforce its uniqueness constraint afterwards.
+   </para>
+
+   <para>
+    Concurrent builds of expression indexes and partial indexes are supported.
+    Errors occurring in the evaluation of these expressions could cause
+    behavior similar to that described above for unique constraint violations.
+   </para>
+
+   <para>
+    Regular index builds permit other regular index builds on the
+    same table to occur in parallel, but only one concurrent index build
+    can occur on a table at a time.  In both cases, no other types of schema
+    modification on the table are allowed meanwhile.  Another difference
+    is that a regular <command>CREATE INDEX</> command can be performed within
+    a transaction block, but <command>CREATE INDEX CONCURRENTLY</> cannot.
+   </para>
+  </refsect2>
  </refsect1>
 
  <refsect1>
@@ -339,15 +430,22 @@ Is this example correct?
    To create a GiST index on a point attribute so that we
    can efficiently use box operators on the result of the
    conversion function:
-  </para>
   <programlisting>
 CREATE INDEX pointloc
     ON points USING GIST (point2box(location) box_ops);
 SELECT * FROM points
     WHERE point2box(points.pointloc) = boxes.box;
   </programlisting>
+  </para>
 -->
 
+  <para>
+   To create an index without locking out writes to the table:
+<programlisting>
+CREATE INDEX CONCURRENTLY sales_quantity_index ON sales_table (quantity);
+</programlisting>
+  </para>
+
  </refsect1>
 
  <refsect1>

doc/src/sgml/ref/reindex.sgml

@@ -1,5 +1,5 @@
 <!--
-$PostgreSQL: pgsql/doc/src/sgml/ref/reindex.sgml,v 1.31 2006/07/04 18:07:24 tgl Exp $
+$PostgreSQL: pgsql/doc/src/sgml/ref/reindex.sgml,v 1.32 2006/08/25 04:06:45 tgl Exp $
 PostgreSQL documentation
 -->
 
@@ -30,7 +30,7 @@ REINDEX { INDEX | TABLE | DATABASE | SYSTEM } <replaceable class="PARAMETER">nam
   <para>
    <command>REINDEX</command> rebuilds an index using the data
    stored in the index's table, replacing the old copy of the index. There are
-   three main reasons to use <command>REINDEX</command>:
+   several scenarios in which to use <command>REINDEX</command>:
 
    <itemizedlist>
     <listitem>
@@ -61,6 +61,18 @@ REINDEX { INDEX | TABLE | DATABASE | SYSTEM } <replaceable class="PARAMETER">nam
       for an index, and wish to ensure that the change has taken full effect.
      </para>
     </listitem>
+
+    <listitem>
+     <para>
+      An index build with the <literal>CONCURRENTLY</> option failed, leaving
+      an <quote>invalid</> index. Such indexes are useless but it can be
+      convenient to use <command>REINDEX</> to rebuild them. Note that
+      <command>REINDEX</> will not perform a concurrent build. To build the
+      index without interfering with production you should drop the index and
+      reissue the <command>CREATE INDEX CONCURRENTLY</> command.
+     </para>
+    </listitem>
+
    </itemizedlist>
   </para>
  </refsect1>

src/backend/access/nbtree/nbtinsert.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/nbtree/nbtinsert.c,v 1.142 2006/07/25 19:13:00 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/nbtree/nbtinsert.c,v 1.143 2006/08/25 04:06:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -244,8 +244,33 @@ _bt_check_unique(Relation rel, IndexTuple itup, Relation heapRel,
 					}
 
 					/*
-					 * Otherwise we have a definite conflict.
+					 * Otherwise we have a definite conflict.  But before
+					 * complaining, look to see if the tuple we want to insert
+					 * is itself now committed dead --- if so, don't complain.
+					 * This is a waste of time in normal scenarios but we must
+					 * do it to support CREATE INDEX CONCURRENTLY.
 					 */
+					htup.t_self = itup->t_tid;
+					if (heap_fetch(heapRel, SnapshotSelf, &htup, &hbuffer,
+									false, NULL))
+					{
+						/* Normal case --- it's still live */
+						ReleaseBuffer(hbuffer);
+					}
+					else if (htup.t_data != NULL)
+					{
+						/*
+						 * It's been deleted, so no error, and no need to
+						 * continue searching
+						 */
+						break;
+					}
+					else
+					{
+						/* couldn't find the tuple?? */
+						elog(ERROR, "failed to fetch tuple being inserted");
+					}
+
 					ereport(ERROR,
 							(errcode(ERRCODE_UNIQUE_VIOLATION),
 					errmsg("duplicate key violates unique constraint \"%s\"",

src/backend/bootstrap/bootparse.y

@@ -9,7 +9,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/bootstrap/bootparse.y,v 1.83 2006/07/31 01:16:36 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/bootstrap/bootparse.y,v 1.84 2006/08/25 04:06:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -257,7 +257,7 @@ Boot_DeclareIndexStmt:
 								$10,
 								NULL, NIL, NIL,
 								false, false, false,
-								false, false, true, false);
+								false, false, true, false, false);
 					do_end();
 				}
 		;
@@ -275,7 +275,7 @@ Boot_DeclareUniqueIndexStmt:
 								$11,
 								NULL, NIL, NIL,
 								true, false, false,
-								false, false, true, false);
+								false, false, true, false, false);
 					do_end();
 				}
 		;

src/backend/catalog/index.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/catalog/index.c,v 1.272 2006/07/31 20:09:00 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/catalog/index.c,v 1.273 2006/08/25 04:06:46 tgl Exp $
  *
  *
  * INTERFACE ROUTINES
@@ -34,6 +34,7 @@
 #include "catalog/index.h"
 #include "catalog/indexing.h"
 #include "catalog/pg_constraint.h"
+#include "catalog/pg_operator.h"
 #include "catalog/pg_opclass.h"
 #include "catalog/pg_type.h"
 #include "executor/executor.h"
@@ -49,8 +50,17 @@
 #include "utils/memutils.h"
 #include "utils/relcache.h"
 #include "utils/syscache.h"
+#include "utils/tuplesort.h"
 
 
+/* state info for validate_index bulkdelete callback */
+typedef struct
+{
+	Tuplesortstate *tuplesort;	/* for sorting the index TIDs */
+	/* statistics (for debug purposes only): */
+	double htups, itups, tups_inserted;
+} v_i_state;
+
 /* non-export function prototypes */
 static TupleDesc ConstructTupleDescriptor(Relation heapRelation,
 						 IndexInfo *indexInfo,
@@ -61,9 +71,16 @@ static void AppendAttributeTuples(Relation indexRelation, int numatts);
 static void UpdateIndexRelation(Oid indexoid, Oid heapoid,
 					IndexInfo *indexInfo,
 					Oid *classOids,
-					bool primary);
+					bool primary,
+					bool isvalid);
 static void index_update_stats(Relation rel, bool hasindex, bool isprimary,
 				   Oid reltoastidxid, double reltuples);
+static bool validate_index_callback(ItemPointer itemptr, void *opaque);
+static void validate_index_heapscan(Relation heapRelation,
+						Relation indexRelation,
+						IndexInfo *indexInfo,
+						Snapshot snapshot,
+						v_i_state *state);
 static Oid	IndexGetRelation(Oid indexId);
 
 
@@ -308,6 +325,8 @@ AppendAttributeTuples(Relation indexRelation, int numatts)
 
 /* ----------------------------------------------------------------
  *		UpdateIndexRelation
+ *
+ * Construct and insert a new entry in the pg_index catalog
  * ----------------------------------------------------------------
  */
 static void
@@ -315,7 +334,8 @@ UpdateIndexRelation(Oid indexoid,
 					Oid heapoid,
 					IndexInfo *indexInfo,
 					Oid *classOids,
-					bool primary)
+					bool primary,
+					bool isvalid)
 {
 	int2vector *indkey;
 	oidvector  *indclass;
@@ -383,6 +403,7 @@ UpdateIndexRelation(Oid indexoid,
 	values[Anum_pg_index_indisunique - 1] = BoolGetDatum(indexInfo->ii_Unique);
 	values[Anum_pg_index_indisprimary - 1] = BoolGetDatum(primary);
 	values[Anum_pg_index_indisclustered - 1] = BoolGetDatum(false);
+	values[Anum_pg_index_indisvalid - 1] = BoolGetDatum(isvalid);
 	values[Anum_pg_index_indkey - 1] = PointerGetDatum(indkey);
 	values[Anum_pg_index_indclass - 1] = PointerGetDatum(indclass);
 	values[Anum_pg_index_indexprs - 1] = exprsDatum;
@@ -428,6 +449,9 @@ UpdateIndexRelation(Oid indexoid,
  * allow_system_table_mods: allow table to be a system catalog
  * skip_build: true to skip the index_build() step for the moment; caller
  *		must do it later (typically via reindex_index())
+ * concurrent: if true, do not lock the table against writers.  The index
+ *		will be marked "invalid" and the caller must take additional steps
+ *		to fix it up.
  *
  * Returns OID of the created index.
  */
@@ -443,7 +467,8 @@ index_create(Oid heapRelationId,
 			 bool isprimary,
 			 bool isconstraint,
 			 bool allow_system_table_mods,
-			 bool skip_build)
+			 bool skip_build,
+			 bool concurrent)
 {
 	Relation	pg_class;
 	Relation	heapRelation;
@@ -456,9 +481,12 @@ index_create(Oid heapRelationId,
 	pg_class = heap_open(RelationRelationId, RowExclusiveLock);
 
 	/*
-	 * Only SELECT ... FOR UPDATE/SHARE are allowed while doing this
+	 * Only SELECT ... FOR UPDATE/SHARE are allowed while doing a standard
+	 * index build; but for concurrent builds we allow INSERT/UPDATE/DELETE
+	 * (but not VACUUM).
 	 */
-	heapRelation = heap_open(heapRelationId, ShareLock);
+	heapRelation = heap_open(heapRelationId,
+							 (concurrent ? ShareUpdateExclusiveLock : ShareLock));
 
 	/*
 	 * The index will be in the same namespace as its parent table, and is
@@ -480,6 +508,16 @@ index_create(Oid heapRelationId,
 				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
 				 errmsg("user-defined indexes on system catalog tables are not supported")));
 
+	/*
+	 * concurrent index build on a system catalog is unsafe because we tend
+	 * to release locks before committing in catalogs
+	 */
+	if (concurrent &&
+		IsSystemRelation(heapRelation))
+		ereport(ERROR,
+				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+				 errmsg("concurrent index creation on system catalog tables is not supported")));
+
 	/*
 	 * We cannot allow indexing a shared relation after initdb (because
 	 * there's no way to make the entry in other databases' pg_class).
@@ -578,7 +616,7 @@ index_create(Oid heapRelationId,
 	 * ----------------
 	 */
 	UpdateIndexRelation(indexRelationId, heapRelationId, indexInfo,
-						classObjectId, isprimary);
+						classObjectId, isprimary, !concurrent);
 
 	/*
 	 * Register constraint and dependencies for the index.
@@ -745,9 +783,8 @@ index_create(Oid heapRelationId,
 	}
 
 	/*
-	 * Close the heap and index; but we keep the ShareLock on the heap and
+	 * Close the heap and index; but we keep the locks that we acquired above
-	 * the exclusive lock on the index that we acquired above, until end of
+	 * until end of transaction.
-	 * transaction.
 	 */
 	index_close(indexRelation, NoLock);
 	heap_close(heapRelation, NoLock);
@@ -895,6 +932,7 @@ BuildIndexInfo(Relation index)
 
 	/* other info */
 	ii->ii_Unique = indexStruct->indisunique;
+	ii->ii_Concurrent = false;			/* assume normal case */
 
 	return ii;
 }
@@ -1327,13 +1365,22 @@ IndexBuildHeapScan(Relation heapRelation,
 						estate);
 
 	/*
-	 * Ok, begin our scan of the base relation.  We use SnapshotAny because we
+	 * Prepare for scan of the base relation.  In a normal index build,
-	 * must retrieve all tuples and do our own time qual checks.
+	 * we use SnapshotAny because we must retrieve all tuples and do our own
+	 * time qual checks (because we have to index RECENTLY_DEAD tuples).
+	 * In a concurrent build, we take a regular MVCC snapshot and index
+	 * whatever's live according to that.  During bootstrap we just use
+	 * SnapshotNow.
 	 */
 	if (IsBootstrapProcessingMode())
 	{
 		snapshot = SnapshotNow;
-		OldestXmin = InvalidTransactionId;
+		OldestXmin = InvalidTransactionId;			/* not used */
+	}
+	else if (indexInfo->ii_Concurrent)
+	{
+		snapshot = CopySnapshot(GetTransactionSnapshot());
+		OldestXmin = InvalidTransactionId;			/* not used */
 	}
 	else
 	{
@@ -1374,10 +1421,12 @@ IndexBuildHeapScan(Relation heapRelation,
 											 scan->rs_cbuf))
 			{
 				case HEAPTUPLE_DEAD:
+					/* Definitely dead, we can ignore it */
 					indexIt = false;
 					tupleIsAlive = false;
 					break;
 				case HEAPTUPLE_LIVE:
+					/* Normal case, index and unique-check it */
 					indexIt = true;
 					tupleIsAlive = true;
 					break;
@@ -1388,6 +1437,7 @@ IndexBuildHeapScan(Relation heapRelation,
 					 * anyway to preserve MVCC semantics.  (Pre-existing
 					 * transactions could try to use the index after we
 					 * finish building it, and may need to see such tuples.)
+					 * Exclude it from unique-checking, however.
 					 */
 					indexIt = true;
 					tupleIsAlive = false;
@@ -1499,6 +1549,309 @@ IndexBuildHeapScan(Relation heapRelation,
 }
 
 
+/*
+ * validate_index - support code for concurrent index builds
+ *
+ * We do a concurrent index build by first building the index normally via
+ * index_create(), while holding a weak lock that allows concurrent
+ * insert/update/delete.  Also, we index only tuples that are valid
+ * as of the start of the scan (see IndexBuildHeapScan), whereas a normal
+ * build takes care to include recently-dead tuples.  This is OK because
+ * we won't mark the index valid until all transactions that might be able
+ * to see those tuples are gone.  The reason for doing that is to avoid
+ * bogus unique-index failures due to concurrent UPDATEs (we might see
+ * different versions of the same row as being valid when we pass over them,
+ * if we used HeapTupleSatisfiesVacuum).  This leaves us with an index that
+ * does not contain any tuples added to the table while we built the index.
+ *
+ * Next, we commit the transaction so that the index becomes visible to other
+ * backends, but it is marked not "indisvalid" to prevent the planner from
+ * relying on it for indexscans.  Then we wait for all transactions that
+ * could have been modifying the table to terminate.  At this point we
+ * know that any subsequently-started transactions will see the index and
+ * insert their new tuples into it.  We then take a new reference snapshot
+ * which is passed to validate_index().  Any tuples that are valid according
+ * to this snap, but are not in the index, must be added to the index.
+ * (Any tuples committed live after the snap will be inserted into the
+ * index by their originating transaction.  Any tuples committed dead before
+ * the snap need not be indexed, because we will wait out all transactions
+ * that might care about them before we mark the index valid.)
+ *
+ * validate_index() works by first gathering all the TIDs currently in the
+ * index, using a bulkdelete callback that just stores the TIDs and doesn't
+ * ever say "delete it".  (This should be faster than a plain indexscan;
+ * also, not all index AMs support full-index indexscan.)  Then we sort the
+ * TIDs, and finally scan the table doing a "merge join" against the TID list
+ * to see which tuples are missing from the index.  Thus we will ensure that
+ * all tuples valid according to the reference snapshot are in the index.
+ *
+ * Building a unique index this way is tricky: we might try to insert a
+ * tuple that is already dead or is in process of being deleted, and we
+ * mustn't have a uniqueness failure against an updated version of the same
+ * row.  We can check the tuple to see if it's already dead and tell
+ * index_insert() not to do the uniqueness check, but that still leaves us
+ * with a race condition against an in-progress update.  To handle that,
+ * we expect the index AM to recheck liveness of the to-be-inserted tuple
+ * before it declares a uniqueness error.
+ *
+ * After completing validate_index(), we wait until all transactions that
+ * were alive at the time of the reference snapshot are gone; this is
+ * necessary to be sure there are none left with a serializable snapshot
+ * older than the reference (and hence possibly able to see tuples we did
+ * not index).  Then we mark the index valid and commit.
+ *
+ * Doing two full table scans is a brute-force strategy.  We could try to be
+ * cleverer, eg storing new tuples in a special area of the table (perhaps
+ * making the table append-only by setting use_fsm).  However that would
+ * add yet more locking issues.
+ */
+void
+validate_index(Oid heapId, Oid indexId, Snapshot snapshot)
+{
+	Relation heapRelation, indexRelation;
+	IndexInfo *indexInfo;
+	IndexVacuumInfo ivinfo;
+	v_i_state state;
+
+	/* Open and lock the parent heap relation */
+	heapRelation = heap_open(heapId, ShareUpdateExclusiveLock);
+	/* And the target index relation */
+	indexRelation = index_open(indexId, RowExclusiveLock);
+
+	/*
+	 * Fetch info needed for index_insert.  (You might think this should
+	 * be passed in from DefineIndex, but its copy is long gone due to
+	 * having been built in a previous transaction.)
+	 */
+	indexInfo = BuildIndexInfo(indexRelation);
+
+	/* mark build is concurrent just for consistency */
+	indexInfo->ii_Concurrent = true;
+
+	/*
+	 * Scan the index and gather up all the TIDs into a tuplesort object.
+	 */
+	ivinfo.index = indexRelation;
+	ivinfo.vacuum_full = false;
+	ivinfo.message_level = DEBUG2;
+	ivinfo.num_heap_tuples = -1;
+
+	state.tuplesort = tuplesort_begin_datum(TIDOID,
+											TIDLessOperator,
+											maintenance_work_mem,
+											false);
+	state.htups = state.itups = state.tups_inserted = 0;
+
+	(void) index_bulk_delete(&ivinfo, NULL,
+							 validate_index_callback, (void *) &state);
+
+	/* Execute the sort */
+	tuplesort_performsort(state.tuplesort);
+
+	/*
+	 * Now scan the heap and "merge" it with the index
+	 */
+	validate_index_heapscan(heapRelation,
+							indexRelation,
+							indexInfo,
+							snapshot,
+							&state);
+
+	/* Done with tuplesort object */
+	tuplesort_end(state.tuplesort);
+
+	elog(DEBUG2,
+		 "validate_index found %.0f heap tuples, %.0f index tuples; inserted %.0f missing tuples",
+		 state.htups, state.itups, state.tups_inserted);
+
+	/* Close rels, but keep locks */
+	index_close(indexRelation, NoLock);
+	heap_close(heapRelation, NoLock);
+}
+
+/*
+ * validate_index_callback - bulkdelete callback to collect the index TIDs
+ */
+static bool
+validate_index_callback(ItemPointer itemptr, void *opaque)
+{
+	v_i_state *state = (v_i_state *) opaque;
+
+	tuplesort_putdatum(state->tuplesort, PointerGetDatum(itemptr), false);
+	state->itups += 1;
+	return false;				/* never actually delete anything */
+}
+
+/*
+ * validate_index_heapscan - second table scan for concurrent index build
+ *
+ * This has much code in common with IndexBuildHeapScan, but it's enough
+ * different that it seems cleaner to have two routines not one.
+ */
+static void
+validate_index_heapscan(Relation heapRelation,
+						Relation indexRelation,
+						IndexInfo *indexInfo,
+						Snapshot snapshot,
+						v_i_state *state)
+{
+	HeapScanDesc scan;
+	HeapTuple	heapTuple;
+	Datum		values[INDEX_MAX_KEYS];
+	bool		isnull[INDEX_MAX_KEYS];
+	List	   *predicate;
+	TupleTableSlot *slot;
+	EState	   *estate;
+	ExprContext *econtext;
+	/* state variables for the merge */
+	ItemPointer indexcursor = NULL;
+	bool tuplesort_empty = false;
+
+	/*
+	 * sanity checks
+	 */
+	Assert(OidIsValid(indexRelation->rd_rel->relam));
+
+	/*
+	 * Need an EState for evaluation of index expressions and partial-index
+	 * predicates.	Also a slot to hold the current tuple.
+	 */
+	estate = CreateExecutorState();
+	econtext = GetPerTupleExprContext(estate);
+	slot = MakeSingleTupleTableSlot(RelationGetDescr(heapRelation));
+
+	/* Arrange for econtext's scan tuple to be the tuple under test */
+	econtext->ecxt_scantuple = slot;
+
+	/* Set up execution state for predicate, if any. */
+	predicate = (List *)
+		ExecPrepareExpr((Expr *) indexInfo->ii_Predicate,
+						estate);
+
+	/*
+	 * Prepare for scan of the base relation.  We need just those tuples
+	 * satisfying the passed-in reference snapshot.
+	 */
+	scan = heap_beginscan(heapRelation, /* relation */
+						  snapshot,		/* seeself */
+						  0,			/* number of keys */
+						  NULL);		/* scan key */
+
+	/*
+	 * Scan all tuples matching the snapshot.
+	 */
+	while ((heapTuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
+	{
+		ItemPointer heapcursor = &heapTuple->t_self;
+
+		CHECK_FOR_INTERRUPTS();
+
+		state->htups += 1;
+
+		/*
+		 * "merge" by skipping through the index tuples until we find or
+		 * pass the current heap tuple.
+		 */
+		while (!tuplesort_empty &&
+			   (!indexcursor ||
+				ItemPointerCompare(indexcursor, heapcursor) < 0))
+		{
+			Datum ts_val;
+			bool ts_isnull;
+
+			if (indexcursor)
+				pfree(indexcursor);
+			tuplesort_empty = !tuplesort_getdatum(state->tuplesort, true,
+												  &ts_val, &ts_isnull);
+			Assert(tuplesort_empty || !ts_isnull);
+			indexcursor = (ItemPointer) DatumGetPointer(ts_val);
+		}
+
+		if (tuplesort_empty ||
+			ItemPointerCompare(indexcursor, heapcursor) > 0)
+		{
+			/*
+			 * We've overshot which means this heap tuple is missing from the
+			 * index, so insert it.
+			 */
+			bool	check_unique;
+
+			MemoryContextReset(econtext->ecxt_per_tuple_memory);
+
+			/* Set up for predicate or expression evaluation */
+			ExecStoreTuple(heapTuple, slot, InvalidBuffer, false);
+
+			/*
+			 * In a partial index, discard tuples that don't satisfy the
+			 * predicate.
+			 */
+			if (predicate != NIL)
+			{
+				if (!ExecQual(predicate, econtext, false))
+					continue;
+			}
+
+			/*
+			 * For the current heap tuple, extract all the attributes we use
+			 * in this index, and note which are null.  This also performs
+			 * evaluation of any expressions needed.
+			 */
+			FormIndexDatum(indexInfo,
+						   slot,
+						   estate,
+						   values,
+						   isnull);
+
+			/*
+			 * If the tuple is already committed dead, we still have to
+			 * put it in the index (because some xacts might be able to
+			 * see it), but we might as well suppress uniqueness checking.
+			 * This is just an optimization because the index AM is not
+			 * supposed to raise a uniqueness failure anyway.
+			 */
+			if (indexInfo->ii_Unique)
+			{
+				/* must hold a buffer lock to call HeapTupleSatisfiesNow */
+				LockBuffer(scan->rs_cbuf, BUFFER_LOCK_SHARE);
+
+				if (HeapTupleSatisfiesNow(heapTuple->t_data, scan->rs_cbuf))
+					check_unique = true;
+				else
+					check_unique = false;
+
+				LockBuffer(scan->rs_cbuf, BUFFER_LOCK_UNLOCK);
+			}
+			else
+				check_unique = false;
+
+			/*
+			 * You'd think we should go ahead and build the index tuple here,
+			 * but some index AMs want to do further processing on the data
+			 * first. So pass the values[] and isnull[] arrays, instead.
+			 */
+			index_insert(indexRelation,
+						 values,
+						 isnull,
+						 heapcursor,
+						 heapRelation,
+						 check_unique);
+
+			state->tups_inserted += 1;
+		}
+	}
+
+	heap_endscan(scan);
+
+	ExecDropSingleTupleTableSlot(slot);
+
+	FreeExecutorState(estate);
+
+	/* These may have been pointing to the now-gone estate */
+	indexInfo->ii_ExpressionsState = NIL;
+	indexInfo->ii_PredicateState = NIL;
+}
+
+
 /*
  * IndexGetRelation: given an index's relation OID, get the OID of the
  * relation it is an index on.	Uses the system cache.
@@ -1530,9 +1883,12 @@ void
 reindex_index(Oid indexId)
 {
 	Relation	iRel,
-				heapRelation;
+				heapRelation,
+				pg_index;
 	Oid			heapId;
 	bool		inplace;
+	HeapTuple	indexTuple;
+	Form_pg_index indexForm;
 
 	/*
 	 * Open and lock the parent heap relation.	ShareLock is sufficient since
@@ -1600,6 +1956,28 @@ reindex_index(Oid indexId)
 	PG_END_TRY();
 	ResetReindexProcessing();
 
+	/*
+	 * If the index is marked invalid (ie, it's from a failed CREATE INDEX
+	 * CONCURRENTLY), we can now mark it valid.  This allows REINDEX to be
+	 * used to clean up in such cases.
+	 */
+	pg_index = heap_open(IndexRelationId, RowExclusiveLock);
+
+	indexTuple = SearchSysCacheCopy(INDEXRELID,
+									ObjectIdGetDatum(indexId),
+									0, 0, 0);
+	if (!HeapTupleIsValid(indexTuple))
+		elog(ERROR, "cache lookup failed for index %u", indexId);
+	indexForm = (Form_pg_index) GETSTRUCT(indexTuple);
+
+	if (!indexForm->indisvalid)
+	{
+		indexForm->indisvalid = true;
+		simple_heap_update(pg_index, &indexTuple->t_self, indexTuple);
+		CatalogUpdateIndexes(pg_index, indexTuple);
+	}
+	heap_close(pg_index, RowExclusiveLock);
+
 	/* Close rels, but keep locks */
 	index_close(iRel, NoLock);
 	heap_close(heapRelation, NoLock);

src/backend/catalog/toasting.c

@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/catalog/toasting.c,v 1.1 2006/07/31 01:16:37 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/catalog/toasting.c,v 1.2 2006/08/25 04:06:47 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -218,6 +218,7 @@ create_toast_table(Relation rel, Oid toastOid, Oid toastIndexOid)
 	indexInfo->ii_Predicate = NIL;
 	indexInfo->ii_PredicateState = NIL;
 	indexInfo->ii_Unique = true;
+	indexInfo->ii_Concurrent = false;
 
 	classObjectId[0] = OID_BTREE_OPS_OID;
 	classObjectId[1] = INT4_BTREE_OPS_OID;
@@ -227,7 +228,7 @@ create_toast_table(Relation rel, Oid toastOid, Oid toastIndexOid)
 							   BTREE_AM_OID,
 							   rel->rd_rel->reltablespace,
 							   classObjectId, (Datum) 0,
-							   true, false, true, false);
+							   true, false, true, false, false);
 
 	/*
 	 * Store the toast table's OID in the parent relation's pg_class row

src/backend/commands/indexcmds.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/commands/indexcmds.c,v 1.146 2006/07/31 01:16:37 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/commands/indexcmds.c,v 1.147 2006/08/25 04:06:48 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -18,6 +18,7 @@
 #include "access/genam.h"
 #include "access/heapam.h"
 #include "access/reloptions.h"
+#include "access/transam.h"
 #include "access/xact.h"
 #include "catalog/catalog.h"
 #include "catalog/dependency.h"
@@ -85,6 +86,7 @@ static bool relationHasPrimaryKey(Relation rel);
  * 'skip_build': make the catalog entries but leave the index file empty;
  *		it will be filled later.
  * 'quiet': suppress the NOTICE chatter ordinarily provided for constraints.
+ * 'concurrent': avoid blocking writers to the table while building.
  */
 void
 DefineIndex(RangeVar *heapRelation,
@@ -102,7 +104,8 @@ DefineIndex(RangeVar *heapRelation,
 			bool is_alter_table,
 			bool check_rights,
 			bool skip_build,
-			bool quiet)
+			bool quiet,
+			bool concurrent)
 {
 	Oid		   *classObjectId;
 	Oid			accessMethodId;
@@ -116,6 +119,12 @@ DefineIndex(RangeVar *heapRelation,
 	Datum		reloptions;
 	IndexInfo  *indexInfo;
 	int			numberOfAttributes;
+	uint32		ixcnt;
+	LockRelId	heaprelid;
+	Snapshot	snapshot;
+	Relation pg_index;
+	HeapTuple indexTuple;
+	Form_pg_index indexForm;
 
 	/*
 	 * count attributes in index
@@ -133,8 +142,16 @@ DefineIndex(RangeVar *heapRelation,
 
 	/*
 	 * Open heap relation, acquire a suitable lock on it, remember its OID
+	 *
+	 * Only SELECT ... FOR UPDATE/SHARE are allowed while doing a standard
+	 * index build; but for concurrent builds we allow INSERT/UPDATE/DELETE
+	 * (but not VACUUM).
 	 */
-	rel = heap_openrv(heapRelation, ShareLock);
+	rel = heap_openrv(heapRelation,
+					  (concurrent ? ShareUpdateExclusiveLock : ShareLock));
+
+	relationId = RelationGetRelid(rel);
+	namespaceId = RelationGetNamespace(rel);
 
 	/* Note: during bootstrap may see uncataloged relation */
 	if (rel->rd_rel->relkind != RELKIND_RELATION &&
@@ -144,8 +161,13 @@ DefineIndex(RangeVar *heapRelation,
 				 errmsg("\"%s\" is not a table",
 						heapRelation->relname)));
 
-	relationId = RelationGetRelid(rel);
+	/*
-	namespaceId = RelationGetNamespace(rel);
+	 * Don't try to CREATE INDEX on temp tables of other backends.
+	 */
+	if (isOtherTempNamespace(namespaceId))
+		ereport(ERROR,
+				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+				 errmsg("cannot create indexes on temporary tables of other sessions")));
 
 	/*
 	 * Verify we (still) have CREATE rights in the rel's namespace.
@@ -391,6 +413,7 @@ DefineIndex(RangeVar *heapRelation,
 	indexInfo->ii_Predicate = make_ands_implicit(predicate);
 	indexInfo->ii_PredicateState = NIL;
 	indexInfo->ii_Unique = unique;
+	indexInfo->ii_Concurrent = concurrent;
 
 	classObjectId = (Oid *) palloc(numberOfAttributes * sizeof(Oid));
 	ComputeIndexAttrs(indexInfo, classObjectId, attributeList,
@@ -410,10 +433,122 @@ DefineIndex(RangeVar *heapRelation,
 				  primary ? "PRIMARY KEY" : "UNIQUE",
 				  indexRelationName, RelationGetRelationName(rel))));
 
+	indexRelationId =
 		index_create(relationId, indexRelationName, indexRelationId,
 					 indexInfo, accessMethodId, tablespaceId, classObjectId,
 					 reloptions, primary, isconstraint,
-				 allowSystemTableMods, skip_build);
+					 allowSystemTableMods, skip_build, concurrent);
+
+	if (!concurrent)
+		return;					/* We're done, in the standard case */
+
+	/*
+	 * Phase 2 of concurrent index build (see comments for validate_index()
+	 * for an overview of how this works)
+	 *
+	 * We must commit our current transaction so that the index becomes
+	 * visible; then start another.  Note that all the data structures
+	 * we just built are lost in the commit.  The only data we keep past
+	 * here are the relation IDs.
+	 *
+	 * Before committing, get a session-level lock on the table, to ensure
+	 * that neither it nor the index can be dropped before we finish.
+	 * This cannot block, even if someone else is waiting for access, because
+	 * we already have the same lock within our transaction.
+	 *
+	 * Note: we don't currently bother with a session lock on the index,
+	 * because there are no operations that could change its state while
+	 * we hold lock on the parent table.  This might need to change later.
+	 */
+	heaprelid = rel->rd_lockInfo.lockRelId;
+	LockRelationIdForSession(&heaprelid, ShareUpdateExclusiveLock);
+
+	CommitTransactionCommand();
+	StartTransactionCommand();
+
+	/* Establish transaction snapshot ... else GetLatestSnapshot complains */
+	(void) GetTransactionSnapshot();
+
+	/*
+	 * Now we must wait until no running transaction could have the table open
+	 * with the old list of indexes.  If we can take an exclusive lock then
+	 * there are none now and anybody who opens it later will get the new
+	 * index in their relcache entry.  Alternatively, if our Xmin reaches our
+	 * own (new) transaction then we know no transactions that started before
+	 * the index was visible are left anyway.
+	 */
+	for (;;)
+	{
+		CHECK_FOR_INTERRUPTS();
+
+		if (ConditionalLockRelationOid(relationId, ExclusiveLock))
+		{
+			/* Release the lock right away to avoid blocking anyone */
+			UnlockRelationOid(relationId, ExclusiveLock);
+			break;
+		}
+
+		if (TransactionIdEquals(GetLatestSnapshot()->xmin,
+								GetTopTransactionId()))
+			break;
+
+		pg_usleep(1000000L);		/* 1 sec */
+	}
+
+	/*
+	 * Now take the "reference snapshot" that will be used by validate_index()
+	 * to filter candidate tuples.  All other transactions running at this
+	 * time will have to be out-waited before we can commit, because we can't
+	 * guarantee that tuples deleted just before this will be in the index.
+	 *
+	 * We also set ActiveSnapshot to this snap, since functions in indexes
+	 * may need a snapshot.
+	 */
+	snapshot = CopySnapshot(GetTransactionSnapshot());
+	ActiveSnapshot = snapshot;
+
+	/*
+	 * Scan the index and the heap, insert any missing index entries.
+	 */
+	validate_index(relationId, indexRelationId, snapshot);
+
+	/*
+	 * The index is now valid in the sense that it contains all currently
+	 * interesting tuples.  But since it might not contain tuples deleted
+	 * just before the reference snap was taken, we have to wait out any
+	 * transactions older than the reference snap.  We can do this by
+	 * waiting for each xact explicitly listed in the snap.
+	 *
+	 * Note: GetSnapshotData() never stores our own xid into a snap,
+	 * hence we need not check for that.
+	 */
+	for (ixcnt = 0; ixcnt < snapshot->xcnt; ixcnt++)
+		XactLockTableWait(snapshot->xip[ixcnt]);
+
+	/* Index can now be marked valid -- update its pg_index entry */
+	pg_index = heap_open(IndexRelationId, RowExclusiveLock);
+
+	indexTuple = SearchSysCacheCopy(INDEXRELID,
+									ObjectIdGetDatum(indexRelationId),
+									0, 0, 0);
+	if (!HeapTupleIsValid(indexTuple))
+		elog(ERROR, "cache lookup failed for index %u", indexRelationId);
+	indexForm = (Form_pg_index) GETSTRUCT(indexTuple);
+
+	Assert(indexForm->indexrelid = indexRelationId);
+	Assert(!indexForm->indisvalid);
+
+	indexForm->indisvalid = true;
+
+	simple_heap_update(pg_index, &indexTuple->t_self, indexTuple);
+	CatalogUpdateIndexes(pg_index, indexTuple);
+
+	heap_close(pg_index, RowExclusiveLock);
+
+	/*
+	 * Last thing to do is release the session-level lock on the parent table.
+	 */
+	UnlockRelationIdForSession(&heaprelid, ShareUpdateExclusiveLock);
 }
 
 

src/backend/commands/tablecmds.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/commands/tablecmds.c,v 1.200 2006/08/21 00:57:24 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/commands/tablecmds.c,v 1.201 2006/08/25 04:06:48 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -3832,7 +3832,8 @@ ATExecAddIndex(AlteredTableInfo *tab, Relation rel,
 				true,			/* is_alter_table */
 				check_rights,
 				skip_build,
-				quiet);
+				quiet,
+				false);
 }
 
 /*

src/backend/nodes/copyfuncs.c

@@ -15,7 +15,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/copyfuncs.c,v 1.348 2006/08/21 00:57:24 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/copyfuncs.c,v 1.349 2006/08/25 04:06:49 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -2049,6 +2049,7 @@ _copyIndexStmt(IndexStmt *from)
 	COPY_SCALAR_FIELD(unique);
 	COPY_SCALAR_FIELD(primary);
 	COPY_SCALAR_FIELD(isconstraint);
+	COPY_SCALAR_FIELD(concurrent);
 
 	return newnode;
 }

src/backend/nodes/equalfuncs.c

@@ -18,7 +18,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/equalfuncs.c,v 1.282 2006/08/21 00:57:24 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/equalfuncs.c,v 1.283 2006/08/25 04:06:49 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -962,6 +962,7 @@ _equalIndexStmt(IndexStmt *a, IndexStmt *b)
 	COMPARE_SCALAR_FIELD(unique);
 	COMPARE_SCALAR_FIELD(primary);
 	COMPARE_SCALAR_FIELD(isconstraint);
+	COMPARE_SCALAR_FIELD(concurrent);
 
 	return true;
 }

src/backend/nodes/outfuncs.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/outfuncs.c,v 1.283 2006/08/21 00:57:24 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/outfuncs.c,v 1.284 2006/08/25 04:06:50 tgl Exp $
  *
  * NOTES
  *	  Every node type that can appear in stored rules' parsetrees *must*
@@ -1353,6 +1353,7 @@ _outIndexStmt(StringInfo str, IndexStmt *node)
 	WRITE_BOOL_FIELD(unique);
 	WRITE_BOOL_FIELD(primary);
 	WRITE_BOOL_FIELD(isconstraint);
+	WRITE_BOOL_FIELD(concurrent);
 }
 
 static void

src/backend/optimizer/util/plancat.c

@@ -9,7 +9,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/util/plancat.c,v 1.124 2006/08/05 00:22:49 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/util/plancat.c,v 1.125 2006/08/25 04:06:50 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -138,6 +138,18 @@ get_relation_info(PlannerInfo *root, Oid relationObjectId, RelOptInfo *rel)
 			indexRelation = index_open(indexoid, lmode);
 			index = indexRelation->rd_index;
 
+			/*
+			 * Ignore invalid indexes, since they can't safely be used for
+			 * queries.  Note that this is OK because the data structure
+			 * we are constructing is only used by the planner --- the
+			 * executor still needs to insert into "invalid" indexes!
+			 */
+			if (!index->indisvalid)
+			{
+				index_close(indexRelation, NoLock);
+				continue;
+			}
+
 			info = makeNode(IndexOptInfo);
 
 			info->indexoid = index->indexrelid;

src/backend/parser/analyze.c

@@ -6,7 +6,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- *	$PostgreSQL: pgsql/src/backend/parser/analyze.c,v 1.347 2006/08/21 00:57:24 tgl Exp $
+ *	$PostgreSQL: pgsql/src/backend/parser/analyze.c,v 1.348 2006/08/25 04:06:51 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1488,6 +1488,7 @@ transformIndexConstraints(ParseState *pstate, CreateStmtContext *cxt)
 		index->tableSpace = constraint->indexspace;
 		index->indexParams = NIL;
 		index->whereClause = NULL;
+		index->concurrent = false;
 
 		/*
 		 * Make sure referenced keys exist.  If we are making a PRIMARY KEY

src/backend/parser/gram.y

@@ -11,7 +11,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/parser/gram.y,v 2.557 2006/08/21 00:57:25 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/parser/gram.y,v 2.558 2006/08/25 04:06:51 tgl Exp $
  *
  * HISTORY
  *	  AUTHOR			DATE			MAJOR EVENT
@@ -364,7 +364,8 @@ static void doNegateFloat(Value *v);
 	CACHE CALLED CASCADE CASCADED CASE CAST CHAIN CHAR_P
 	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLOSE
 	CLUSTER COALESCE COLLATE COLUMN COMMENT COMMIT
-	COMMITTED CONNECTION CONSTRAINT CONSTRAINTS CONVERSION_P CONVERT COPY CREATE CREATEDB
+	COMMITTED CONCURRENTLY CONNECTION CONSTRAINT CONSTRAINTS
+	CONVERSION_P CONVERT COPY CREATE CREATEDB
 	CREATEROLE CREATEUSER CROSS CSV CURRENT_DATE CURRENT_ROLE CURRENT_TIME
 	CURRENT_TIMESTAMP CURRENT_USER CURSOR CYCLE
 
@@ -3638,20 +3639,22 @@ opt_granted_by: GRANTED BY RoleId						{ $$ = $3; }
 
 /*****************************************************************************
  *
- *		QUERY:
+ *		QUERY: CREATE INDEX
- *				create index <indexname> on <relname>
+ *
- *				  [ using <access> ] "(" ( <col> [ using <opclass> ] )+ ")"
+ * Note: we can't factor CONCURRENTLY into a separate production without
- *				  [ tablespace <tablespacename> ] [ where <predicate> ]
+ * making it a reserved word.
  *
  * Note: we cannot put TABLESPACE clause after WHERE clause unless we are
  * willing to make TABLESPACE a fully reserved word.
  *****************************************************************************/
 
-IndexStmt:	CREATE index_opt_unique INDEX index_name ON qualified_name
+IndexStmt:	CREATE index_opt_unique INDEX index_name
-			access_method_clause '(' index_params ')' opt_definition OptTableSpace where_clause
+			ON qualified_name access_method_clause '(' index_params ')'
+			opt_definition OptTableSpace where_clause
 				{
 					IndexStmt *n = makeNode(IndexStmt);
 					n->unique = $2;
+					n->concurrent = false;
 					n->idxname = $4;
 					n->relation = $6;
 					n->accessMethod = $7;
@@ -3661,6 +3664,22 @@ IndexStmt:	CREATE index_opt_unique INDEX index_name ON qualified_name
 					n->whereClause = $13;
 					$$ = (Node *)n;
 				}
+			| CREATE index_opt_unique INDEX CONCURRENTLY index_name
+			ON qualified_name access_method_clause '(' index_params ')'
+			opt_definition OptTableSpace where_clause
+				{
+					IndexStmt *n = makeNode(IndexStmt);
+					n->unique = $2;
+					n->concurrent = true;
+					n->idxname = $5;
+					n->relation = $7;
+					n->accessMethod = $8;
+					n->indexParams = $10;
+					n->options = $12;
+					n->tableSpace = $13;
+					n->whereClause = $14;
+					$$ = (Node *)n;
+				}
 		;
 
 index_opt_unique:
@@ -8491,6 +8510,7 @@ unreserved_keyword:
 			| COMMENT
 			| COMMIT
 			| COMMITTED
+			| CONCURRENTLY
 			| CONNECTION
 			| CONSTRAINTS
 			| CONVERSION_P

src/backend/parser/keywords.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/parser/keywords.c,v 1.175 2006/08/12 02:52:05 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/parser/keywords.c,v 1.176 2006/08/25 04:06:52 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -84,6 +84,7 @@ static const ScanKeyword ScanKeywords[] = {
 	{"comment", COMMENT},
 	{"commit", COMMIT},
 	{"committed", COMMITTED},
+	{"concurrently", CONCURRENTLY},
 	{"connection", CONNECTION},
 	{"constraint", CONSTRAINT},
 	{"constraints", CONSTRAINTS},

src/backend/storage/page/itemptr.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/storage/page/itemptr.c,v 1.17 2006/07/14 14:52:23 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/storage/page/itemptr.c,v 1.18 2006/08/25 04:06:53 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -16,13 +16,14 @@
 
 #include "storage/itemptr.h"
 
+
 /*
  * ItemPointerEquals
  *	Returns true if both item pointers point to the same item,
  *	 otherwise returns false.
  *
  * Note:
- *	Assumes that the disk item pointers are not NULL.
+ *	Asserts that the disk item pointers are both valid!
  */
 bool
 ItemPointerEquals(ItemPointer pointer1, ItemPointer pointer2)
@@ -35,3 +36,30 @@ ItemPointerEquals(ItemPointer pointer1, ItemPointer pointer2)
 	else
 		return false;
 }
+
+/*
+ * ItemPointerCompare
+ *		Generic btree-style comparison for item pointers.
+ */
+int32
+ItemPointerCompare(ItemPointer arg1, ItemPointer arg2)
+{
+	/*
+	 * Don't use ItemPointerGetBlockNumber or ItemPointerGetOffsetNumber here,
+	 * because they assert ip_posid != 0 which might not be true for a
+	 * user-supplied TID.
+	 */
+	BlockNumber b1 = BlockIdGetBlockNumber(&(arg1->ip_blkid));
+	BlockNumber b2 = BlockIdGetBlockNumber(&(arg2->ip_blkid));
+	
+	if (b1 < b2)
+		return -1;
+	else if (b1 > b2)
+		return 1;
+	else if (arg1->ip_posid < arg2->ip_posid)
+		return -1;
+	else if (arg1->ip_posid > arg2->ip_posid)
+		return 1;
+	else
+		return 0;
+}

src/backend/tcop/utility.c

@@ -10,7 +10,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/tcop/utility.c,v 1.266 2006/08/15 18:26:58 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/tcop/utility.c,v 1.267 2006/08/25 04:06:53 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -781,6 +781,9 @@ ProcessUtility(Node *parsetree,
 			{
 				IndexStmt  *stmt = (IndexStmt *) parsetree;
 
+				if (stmt->concurrent)
+					PreventTransactionChain(stmt, "CREATE INDEX CONCURRENTLY");
+
 				CheckRelationOwnership(stmt->relation, true);
 
 				DefineIndex(stmt->relation,		/* relation */
@@ -798,7 +801,8 @@ ProcessUtility(Node *parsetree,
 							false,				/* is_alter_table */
 							true,				/* check_rights */
 							false,				/* skip_build */
-							false);		/* quiet */
+							false,				/* quiet */
+							stmt->concurrent);	/* concurrent */
 			}
 			break;
 

src/backend/utils/adt/tid.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tid.c,v 1.54 2006/07/21 20:51:32 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tid.c,v 1.55 2006/08/25 04:06:53 tgl Exp $
  *
  * NOTES
  *	  input routine largely stolen from boxin().
@@ -158,36 +158,13 @@ tidsend(PG_FUNCTION_ARGS)
  *	 PUBLIC ROUTINES														 *
  *****************************************************************************/
 
-static int32
-tid_cmp_internal(ItemPointer arg1, ItemPointer arg2)
-{
-	/*
-	 * Don't use ItemPointerGetBlockNumber or ItemPointerGetOffsetNumber here,
-	 * because they assert ip_posid != 0 which might not be true for a
-	 * user-supplied TID.
-	 */
-	BlockNumber b1 = BlockIdGetBlockNumber(&(arg1->ip_blkid));
-	BlockNumber b2 = BlockIdGetBlockNumber(&(arg2->ip_blkid));
-	
-	if (b1 < b2)
-		return -1;
-	else if (b1 > b2)
-		return 1;
-	else if (arg1->ip_posid < arg2->ip_posid)
-		return -1;
-	else if (arg1->ip_posid > arg2->ip_posid)
-		return 1;
-	else
-		return 0;
-}
-
 Datum
 tideq(PG_FUNCTION_ARGS)
 {
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_BOOL(tid_cmp_internal(arg1,arg2) == 0);
+	PG_RETURN_BOOL(ItemPointerCompare(arg1,arg2) == 0);
 }
 
 Datum
@@ -196,7 +173,7 @@ tidne(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_BOOL(tid_cmp_internal(arg1,arg2) != 0);
+	PG_RETURN_BOOL(ItemPointerCompare(arg1,arg2) != 0);
 }
 
 Datum
@@ -205,7 +182,7 @@ tidlt(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_BOOL(tid_cmp_internal(arg1,arg2) < 0);
+	PG_RETURN_BOOL(ItemPointerCompare(arg1,arg2) < 0);
 }
 
 Datum
@@ -214,7 +191,7 @@ tidle(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_BOOL(tid_cmp_internal(arg1,arg2) <= 0);
+	PG_RETURN_BOOL(ItemPointerCompare(arg1,arg2) <= 0);
 }
 
 Datum
@@ -223,7 +200,7 @@ tidgt(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_BOOL(tid_cmp_internal(arg1,arg2) > 0);
+	PG_RETURN_BOOL(ItemPointerCompare(arg1,arg2) > 0);
 }
 
 Datum
@@ -232,7 +209,7 @@ tidge(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_BOOL(tid_cmp_internal(arg1,arg2) >= 0);
+	PG_RETURN_BOOL(ItemPointerCompare(arg1,arg2) >= 0);
 }
 
 Datum
@@ -241,7 +218,7 @@ bttidcmp(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_INT32(tid_cmp_internal(arg1, arg2));
+	PG_RETURN_INT32(ItemPointerCompare(arg1, arg2));
 }
 
 Datum
@@ -250,7 +227,7 @@ tidlarger(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_ITEMPOINTER(tid_cmp_internal(arg1,arg2) >= 0 ? arg1 : arg2);
+	PG_RETURN_ITEMPOINTER(ItemPointerCompare(arg1,arg2) >= 0 ? arg1 : arg2);
 }
 
 Datum
@@ -259,7 +236,7 @@ tidsmaller(PG_FUNCTION_ARGS)
 	ItemPointer arg1 = PG_GETARG_ITEMPOINTER(0);
 	ItemPointer arg2 = PG_GETARG_ITEMPOINTER(1);
 
-	PG_RETURN_ITEMPOINTER(tid_cmp_internal(arg1,arg2) <= 0 ? arg1 : arg2);
+	PG_RETURN_ITEMPOINTER(ItemPointerCompare(arg1,arg2) <= 0 ? arg1 : arg2);
 }
 
 

src/bin/psql/common.c

@@ -3,7 +3,7 @@
  *
  * Copyright (c) 2000-2006, PostgreSQL Global Development Group
  *
- * $PostgreSQL: pgsql/src/bin/psql/common.c,v 1.124 2006/08/13 21:10:04 tgl Exp $
+ * $PostgreSQL: pgsql/src/bin/psql/common.c,v 1.125 2006/08/25 04:06:54 tgl Exp $
  */
 #include "postgres_fe.h"
 #include "common.h"
@@ -1075,22 +1075,71 @@ command_no_begin(const char *query)
 	 * Commands not allowed within transactions.  The statements checked for
 	 * here should be exactly those that call PreventTransactionChain() in the
 	 * backend.
-	 *
-	 * Note: we are a bit sloppy about CLUSTER, which is transactional in some
-	 * variants but not others.
 	 */
 	if (wordlen == 6 && pg_strncasecmp(query, "vacuum", 6) == 0)
 		return true;
 	if (wordlen == 7 && pg_strncasecmp(query, "cluster", 7) == 0)
+	{
+		/* CLUSTER with any arguments is allowed in transactions */
+		query += wordlen;
+
+		query = skip_white_space(query);
+
+		if (isalpha((unsigned char) query[0]))
+			return false;		/* has additional words */
+		return true;			/* it's CLUSTER without arguments */
+	}
+
+	if (wordlen == 6 && pg_strncasecmp(query, "create", 6) == 0)
+	{
+		query += wordlen;
+
+		query = skip_white_space(query);
+
+		wordlen = 0;
+		while (isalpha((unsigned char) query[wordlen]))
+			wordlen += PQmblen(&query[wordlen], pset.encoding);
+
+		if (wordlen == 8 && pg_strncasecmp(query, "database", 8) == 0)
+			return true;
+		if (wordlen == 10 && pg_strncasecmp(query, "tablespace", 10) == 0)
 			return true;
 
+		/* CREATE [UNIQUE] INDEX CONCURRENTLY isn't allowed in xacts */
+		if (wordlen == 6 && pg_strncasecmp(query, "unique", 6) == 0)
+		{
+			query += wordlen;
+
+			query = skip_white_space(query);
+
+			wordlen = 0;
+			while (isalpha((unsigned char) query[wordlen]))
+				wordlen += PQmblen(&query[wordlen], pset.encoding);
+		}
+
+		if (wordlen == 5 && pg_strncasecmp(query, "index", 5) == 0)
+		{
+			query += wordlen;
+
+			query = skip_white_space(query);
+
+			wordlen = 0;
+			while (isalpha((unsigned char) query[wordlen]))
+				wordlen += PQmblen(&query[wordlen], pset.encoding);
+
+			if (wordlen == 12 && pg_strncasecmp(query, "concurrently", 12) == 0)
+				return true;
+		}
+
+		return false;
+	}
+
 	/*
-	 * Note: these tests will match CREATE SYSTEM, DROP SYSTEM, and REINDEX
+	 * Note: these tests will match DROP SYSTEM and REINDEX TABLESPACE,
-	 * TABLESPACE, which aren't really valid commands so we don't care much.
+	 * which aren't really valid commands so we don't care much.
-	 * The other six possible matches are correct.
+	 * The other four possible matches are correct.
 	 */
-	if ((wordlen == 6 && pg_strncasecmp(query, "create", 6) == 0) ||
+	if ((wordlen == 4 && pg_strncasecmp(query, "drop", 4) == 0) ||
-		(wordlen == 4 && pg_strncasecmp(query, "drop", 4) == 0) ||
 		(wordlen == 7 && pg_strncasecmp(query, "reindex", 7) == 0))
 	{
 		query += wordlen;

src/bin/psql/describe.c

@@ -3,7 +3,7 @@
  *
  * Copyright (c) 2000-2006, PostgreSQL Global Development Group
  *
- * $PostgreSQL: pgsql/src/bin/psql/describe.c,v 1.142 2006/07/27 19:52:06 tgl Exp $
+ * $PostgreSQL: pgsql/src/bin/psql/describe.c,v 1.143 2006/08/25 04:06:54 tgl Exp $
  */
 #include "postgres_fe.h"
 #include "describe.h"
@@ -942,7 +942,7 @@ describeOneTableDetails(const char *schemaname,
 		PGresult   *result;
 
 		printfPQExpBuffer(&buf,
-						  "SELECT i.indisunique, i.indisprimary, i.indisclustered, a.amname, c2.relname,\n"
+						  "SELECT i.indisunique, i.indisprimary, i.indisclustered, i.indisvalid, a.amname, c2.relname,\n"
 					"  pg_catalog.pg_get_expr(i.indpred, i.indrelid, true)\n"
 						  "FROM pg_catalog.pg_index i, pg_catalog.pg_class c, pg_catalog.pg_class c2, pg_catalog.pg_am a\n"
 		  "WHERE i.indexrelid = c.oid AND c.oid = '%s' AND c.relam = a.oid\n"
@@ -962,9 +962,10 @@ describeOneTableDetails(const char *schemaname,
 			char	   *indisunique = PQgetvalue(result, 0, 0);
 			char	   *indisprimary = PQgetvalue(result, 0, 1);
 			char	   *indisclustered = PQgetvalue(result, 0, 2);
-			char	   *indamname = PQgetvalue(result, 0, 3);
+			char	   *indisvalid = PQgetvalue(result, 0, 3);
-			char	   *indtable = PQgetvalue(result, 0, 4);
+			char	   *indamname = PQgetvalue(result, 0, 4);
-			char	   *indpred = PQgetvalue(result, 0, 5);
+			char	   *indtable = PQgetvalue(result, 0, 5);
+			char	   *indpred = PQgetvalue(result, 0, 6);
 			int			count_footers = 0;
 
 			if (strcmp(indisprimary, "t") == 0)
@@ -985,6 +986,9 @@ describeOneTableDetails(const char *schemaname,
 			if (strcmp(indisclustered, "t") == 0)
 				appendPQExpBuffer(&tmpbuf, _(", clustered"));
 
+			if (strcmp(indisvalid, "t") != 0)
+				appendPQExpBuffer(&tmpbuf, _(", invalid"));
+
 			footers = pg_malloc_zero(4 * sizeof(*footers));
 			footers[count_footers++] = pg_strdup(tmpbuf.data);
 			add_tablespace_footer(tableinfo.relkind, tableinfo.tablespace,
@@ -1067,7 +1071,7 @@ describeOneTableDetails(const char *schemaname,
 		if (tableinfo.hasindex)
 		{
 			printfPQExpBuffer(&buf,
-							  "SELECT c2.relname, i.indisprimary, i.indisunique, i.indisclustered, "
+							  "SELECT c2.relname, i.indisprimary, i.indisunique, i.indisclustered, i.indisvalid, "
 							  "pg_catalog.pg_get_indexdef(i.indexrelid, 0, true), c2.reltablespace\n"
 							  "FROM pg_catalog.pg_class c, pg_catalog.pg_class c2, pg_catalog.pg_index i\n"
 							  "WHERE c.oid = '%s' AND c.oid = i.indrelid AND i.indexrelid = c2.oid\n"
@@ -1201,7 +1205,7 @@ describeOneTableDetails(const char *schemaname,
 								   ? " UNIQUE,"
 								   : ""));
 				/* Everything after "USING" is echoed verbatim */
-				indexdef = PQgetvalue(result1, i, 4);
+				indexdef = PQgetvalue(result1, i, 5);
 				usingpos = strstr(indexdef, " USING ");
 				if (usingpos)
 					indexdef = usingpos + 7;
@@ -1211,11 +1215,14 @@ describeOneTableDetails(const char *schemaname,
 				if (strcmp(PQgetvalue(result1, i, 3), "t") == 0)
 					appendPQExpBuffer(&buf, " CLUSTER");
 
+				if (strcmp(PQgetvalue(result1, i, 4), "t") != 0)
+					appendPQExpBuffer(&buf, " INVALID");
+
 				/* Print tablespace of the index on the same line */
 				count_footers += 1;
 				initPQExpBuffer(&tmpbuf);
 				if (add_tablespace_footer('i',
-										  atooid(PQgetvalue(result1, i, 5)),
+										  atooid(PQgetvalue(result1, i, 6)),
 									 footers, &count_footers, tmpbuf, false))
 				{
 					appendPQExpBuffer(&buf, ", ");

src/include/catalog/catversion.h

@@ -37,7 +37,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/catalog/catversion.h,v 1.352 2006/08/21 00:57:26 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/catalog/catversion.h,v 1.353 2006/08/25 04:06:54 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -53,6 +53,6 @@
  */
 
 /*							yyyymmddN */
-#define CATALOG_VERSION_NO	200608191
+#define CATALOG_VERSION_NO	200608211
 
 #endif

src/include/catalog/index.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/catalog/index.h,v 1.70 2006/07/31 01:16:37 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/catalog/index.h,v 1.71 2006/08/25 04:06:55 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -39,7 +39,8 @@ extern Oid index_create(Oid heapRelationId,
 			 bool isprimary,
 			 bool isconstraint,
 			 bool allow_system_table_mods,
-			 bool skip_build);
+			 bool skip_build,
+			 bool concurrent);
 
 extern void index_drop(Oid indexId);
 
@@ -64,6 +65,8 @@ extern double IndexBuildHeapScan(Relation heapRelation,
 				   IndexBuildCallback callback,
 				   void *callback_state);
 
+extern void validate_index(Oid heapId, Oid indexId, Snapshot snapshot);
+
 extern void reindex_index(Oid indexId);
 extern bool reindex_relation(Oid relid, bool toast_too);
 

src/include/catalog/pg_attribute.h

@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/catalog/pg_attribute.h,v 1.123 2006/07/10 16:20:51 alvherre Exp $
+ * $PostgreSQL: pgsql/src/include/catalog/pg_attribute.h,v 1.124 2006/08/25 04:06:55 tgl Exp $
  *
  * NOTES
  *	  the genbki.sh script reads this file and generates .bki
@@ -460,9 +460,10 @@ DATA(insert ( 1259 tableoid			26 0  4  -7 0 -1 -1 t p i t f f t 0));
 { 0, {"indisunique"},		16, -1, 1, 4, 0, -1, -1, true, 'p', 'c', true, false, false, true, 0 }, \
 { 0, {"indisprimary"},		16, -1, 1, 5, 0, -1, -1, true, 'p', 'c', true, false, false, true, 0 }, \
 { 0, {"indisclustered"},	16, -1, 1, 6, 0, -1, -1, true, 'p', 'c', true, false, false, true, 0 }, \
-{ 0, {"indkey"},			22, -1, -1, 7, 1, -1, -1, false, 'p', 'i', true, false, false, true, 0 }, \
+{ 0, {"indisvalid"},		16, -1, 1, 7, 0, -1, -1, true, 'p', 'c', true, false, false, true, 0 }, \
-{ 0, {"indclass"},			30, -1, -1, 8, 1, -1, -1, false, 'p', 'i', true, false, false, true, 0 }, \
+{ 0, {"indkey"},			22, -1, -1, 8, 1, -1, -1, false, 'p', 'i', true, false, false, true, 0 }, \
-{ 0, {"indexprs"},			25, -1, -1, 9, 0, -1, -1, false, 'x', 'i', false, false, false, true, 0 }, \
+{ 0, {"indclass"},			30, -1, -1, 9, 1, -1, -1, false, 'p', 'i', true, false, false, true, 0 }, \
-{ 0, {"indpred"},			25, -1, -1, 10, 0, -1, -1, false, 'x', 'i', false, false, false, true, 0 }
+{ 0, {"indexprs"},			25, -1, -1, 10, 0, -1, -1, false, 'x', 'i', false, false, false, true, 0 }, \
+{ 0, {"indpred"},			25, -1, -1, 11, 0, -1, -1, false, 'x', 'i', false, false, false, true, 0 }
 
 #endif   /* PG_ATTRIBUTE_H */

src/include/catalog/pg_index.h

@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/catalog/pg_index.h,v 1.39 2006/03/05 15:58:54 momjian Exp $
+ * $PostgreSQL: pgsql/src/include/catalog/pg_index.h,v 1.40 2006/08/25 04:06:55 tgl Exp $
  *
  * NOTES
  *	  the genbki.sh script reads this file and generates .bki
@@ -41,6 +41,7 @@ CATALOG(pg_index,2610) BKI_WITHOUT_OIDS
 	bool		indisunique;	/* is this a unique index? */
 	bool		indisprimary;	/* is this index for primary key? */
 	bool		indisclustered; /* is this the index last clustered by? */
+	bool		indisvalid; 	/* is this index valid for use by queries? */
 
 	/* VARIABLE LENGTH FIELDS: */
 	int2vector	indkey;			/* column numbers of indexed cols, or 0 */
@@ -63,16 +64,17 @@ typedef FormData_pg_index *Form_pg_index;
  *		compiler constants for pg_index
  * ----------------
  */
-#define Natts_pg_index					10
+#define Natts_pg_index					11
 #define Anum_pg_index_indexrelid		1
 #define Anum_pg_index_indrelid			2
 #define Anum_pg_index_indnatts			3
 #define Anum_pg_index_indisunique		4
 #define Anum_pg_index_indisprimary		5
 #define Anum_pg_index_indisclustered	6
-#define Anum_pg_index_indkey			7
+#define Anum_pg_index_indisvalid		7
-#define Anum_pg_index_indclass			8
+#define Anum_pg_index_indkey			8
-#define Anum_pg_index_indexprs			9
+#define Anum_pg_index_indclass			9
-#define Anum_pg_index_indpred			10
+#define Anum_pg_index_indexprs			10
+#define Anum_pg_index_indpred			11
 
 #endif   /* PG_INDEX_H */

src/include/commands/defrem.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/commands/defrem.h,v 1.75 2006/07/18 17:42:01 momjian Exp $
+ * $PostgreSQL: pgsql/src/include/commands/defrem.h,v 1.76 2006/08/25 04:06:55 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -33,7 +33,8 @@ extern void DefineIndex(RangeVar *heapRelation,
 			bool is_alter_table,
 			bool check_rights,
 			bool skip_build,
-			bool quiet);
+			bool quiet,
+			bool concurrent);
 extern void RemoveIndex(RangeVar *relation, DropBehavior behavior);
 extern void ReindexIndex(RangeVar *indexRelation);
 extern void ReindexTable(RangeVar *relation);

src/include/nodes/execnodes.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/nodes/execnodes.h,v 1.159 2006/08/12 02:52:06 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/nodes/execnodes.h,v 1.160 2006/08/25 04:06:56 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -37,6 +37,7 @@
  *		Predicate			partial-index predicate, or NIL if none
  *		PredicateState		exec state for predicate, or NIL if none
  *		Unique				is it a unique index?
+ *		Concurrent			are we doing a concurrent index build?
  * ----------------
  */
 typedef struct IndexInfo
@@ -49,6 +50,7 @@ typedef struct IndexInfo
 	List	   *ii_Predicate;	/* list of Expr */
 	List	   *ii_PredicateState;		/* list of ExprState */
 	bool		ii_Unique;
+	bool		ii_Concurrent;
 } IndexInfo;
 
 /* ----------------

src/include/nodes/parsenodes.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/nodes/parsenodes.h,v 1.324 2006/08/21 00:57:26 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/nodes/parsenodes.h,v 1.325 2006/08/25 04:06:56 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1463,6 +1463,7 @@ typedef struct IndexStmt
 	bool		unique;			/* is index unique? */
 	bool		primary;		/* is index on primary key? */
 	bool		isconstraint;	/* is it from a CONSTRAINT clause? */
+	bool		concurrent;		/* should this be a concurrent index build? */
 } IndexStmt;
 
 /* ----------------------

src/include/storage/itemptr.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/storage/itemptr.h,v 1.28 2006/03/05 15:58:59 momjian Exp $
+ * $PostgreSQL: pgsql/src/include/storage/itemptr.h,v 1.29 2006/08/25 04:06:57 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -141,5 +141,6 @@ typedef ItemPointerData *ItemPointer;
  */
 
 extern bool ItemPointerEquals(ItemPointer pointer1, ItemPointer pointer2);
+extern int32 ItemPointerCompare(ItemPointer arg1, ItemPointer arg2);
 
 #endif   /* ITEMPTR_H */

src/test/regress/expected/create_index.out

@@ -360,3 +360,53 @@ INSERT INTO func_index_heap VALUES('QWERTY');
 create unique index hash_f8_index_1 on hash_f8_heap(abs(random));
 create unique index hash_f8_index_2 on hash_f8_heap((seqno + 1), random);
 create unique index hash_f8_index_3 on hash_f8_heap(random) where seqno > 1000;
+--
+-- Try some concurrent index builds
+--
+-- Unfortunately this only tests about half the code paths because there are
+-- no concurrent updates happening to the table at the same time.
+CREATE TABLE concur_heap (f1 text, f2 text);
+-- empty table
+CREATE INDEX CONCURRENTLY concur_index1 ON concur_heap(f2,f1);
+INSERT INTO concur_heap VALUES  ('a','b');
+INSERT INTO concur_heap VALUES  ('b','b');
+-- unique index
+CREATE UNIQUE INDEX CONCURRENTLY concur_index2 ON concur_heap(f1);
+-- check if constraint is set up properly to be enforced
+INSERT INTO concur_heap VALUES ('b','x');
+ERROR:  duplicate key violates unique constraint "concur_index2"
+-- check if constraint is enforced properly at build time
+CREATE UNIQUE INDEX CONCURRENTLY concur_index3 ON concur_heap(f2);
+ERROR:  could not create unique index
+DETAIL:  Table contains duplicated values.
+-- test that expression indexes and partial indexes work concurrently
+CREATE INDEX CONCURRENTLY concur_index4 on concur_heap(f2) WHERE f1='a';
+CREATE INDEX CONCURRENTLY concur_index5 on concur_heap(f2) WHERE f1='x';
+CREATE INDEX CONCURRENTLY concur_index6 on concur_heap((f2||f1));
+-- You can't do a concurrent index build in a transaction
+BEGIN;
+CREATE INDEX CONCURRENTLY concur_index7 ON concur_heap(f1);
+ERROR:  CREATE INDEX CONCURRENTLY cannot run inside a transaction block
+COMMIT;
+-- But you can do a regular index build in a transaction
+BEGIN;
+CREATE INDEX std_index on concur_heap(f2);
+COMMIT;
+-- check to make sure that the failed indexes were cleaned up properly and the
+-- successful indexes are created properly. Notably that they do NOT have the
+-- "invalid" flag set.
+\d concur_heap
+Table "public.concur_heap"
+ Column | Type | Modifiers 
+--------+------+-----------
+ f1     | text | 
+ f2     | text | 
+Indexes:
+    "concur_index2" UNIQUE, btree (f1)
+    "concur_index1" btree (f2, f1)
+    "concur_index4" btree (f2) WHERE f1 = 'a'::text
+    "concur_index5" btree (f2) WHERE f1 = 'x'::text
+    "concur_index6" btree ((f2 || f1))
+    "std_index" btree (f2)
+
+DROP TABLE concur_heap;

src/test/regress/sql/create_index.sql

@@ -219,3 +219,43 @@ INSERT INTO func_index_heap VALUES('QWERTY');
 create unique index hash_f8_index_1 on hash_f8_heap(abs(random));
 create unique index hash_f8_index_2 on hash_f8_heap((seqno + 1), random);
 create unique index hash_f8_index_3 on hash_f8_heap(random) where seqno > 1000;
+
+--
+-- Try some concurrent index builds
+--
+-- Unfortunately this only tests about half the code paths because there are
+-- no concurrent updates happening to the table at the same time.
+
+CREATE TABLE concur_heap (f1 text, f2 text);
+-- empty table
+CREATE INDEX CONCURRENTLY concur_index1 ON concur_heap(f2,f1);
+INSERT INTO concur_heap VALUES  ('a','b');
+INSERT INTO concur_heap VALUES  ('b','b');
+-- unique index
+CREATE UNIQUE INDEX CONCURRENTLY concur_index2 ON concur_heap(f1);
+-- check if constraint is set up properly to be enforced
+INSERT INTO concur_heap VALUES ('b','x');
+-- check if constraint is enforced properly at build time
+CREATE UNIQUE INDEX CONCURRENTLY concur_index3 ON concur_heap(f2);
+-- test that expression indexes and partial indexes work concurrently
+CREATE INDEX CONCURRENTLY concur_index4 on concur_heap(f2) WHERE f1='a';
+CREATE INDEX CONCURRENTLY concur_index5 on concur_heap(f2) WHERE f1='x';
+CREATE INDEX CONCURRENTLY concur_index6 on concur_heap((f2||f1));
+
+-- You can't do a concurrent index build in a transaction
+BEGIN;
+CREATE INDEX CONCURRENTLY concur_index7 ON concur_heap(f1);
+COMMIT;
+
+-- But you can do a regular index build in a transaction
+BEGIN;
+CREATE INDEX std_index on concur_heap(f2);
+COMMIT;
+
+-- check to make sure that the failed indexes were cleaned up properly and the
+-- successful indexes are created properly. Notably that they do NOT have the
+-- "invalid" flag set.
+
+\d concur_heap
+
+DROP TABLE concur_heap;