commit Oleg and Teodor's RD-tree implementation ... this provides the
regression tests for the GiST changes ... this should be integrated into
the regular regression tests similar to Vadim's SPI contrib stuff ...

contrib/intarray/Makefile

+subdir = contrib/intarray
+top_builddir = ../..
+include $(top_builddir)/src/Makefile.global
+
+# override libdir to install shlib in contrib not main directory
+libdir := $(libdir)/contrib
+
+# shared library parameters
+NAME= _int
+SO_MAJOR_VERSION= 1
+SO_MINOR_VERSION= 0
+
+override CPPFLAGS += -I$(srcdir) -DPGSQL71
+
+OBJS= _int.o
+
+all: all-lib $(NAME).sql
+
+# Shared library stuff
+include $(top_srcdir)/src/Makefile.shlib
+
+
+$(NAME).sql: $(NAME).sql.in
+	sed -e 's:MODULE_PATHNAME:$(libdir)/$(shlib):g' < $< > $@
+
+.PHONY: submake
+submake:
+	$(MAKE) -C $(top_builddir)/src/test/regress pg_regress
+
+# against installed postmaster
+installcheck: submake
+	@echo "'make installcheck' is not supported."	
+
+installcheck: submake
+	$(top_builddir)/src/test/regress/pg_regress _int
+
+# in-tree test doesn't work yet (no way to install my shared library)
+#check: all submake
+#	$(top_builddir)/src/test/regress/pg_regress --temp-install \
+#	  --top-builddir=$(top_builddir) _int
+check:
+	@echo "'make check' is not supported."
+	@echo "Do 'make install', then 'make installcheck' instead."
+
+install: all installdirs install-lib
+	#$(INSTALL_DATA) $(srcdir)/README.$(NAME)  $(docdir)/contrib
+	$(INSTALL_DATA) $(NAME).sql $(datadir)/contrib
+
+installdirs:
+	$(mkinstalldirs) $(docdir)/contrib $(datadir)/contrib $(libdir)
+
+uninstall: uninstall-lib
+	rm -f $(docdir)/contrib/README.$(NAME) $(datadir)/contrib/$(NAME).sql
+
+clean distclean maintainer-clean: clean-lib
+	rm -f *.so y.tab.c y.tab.h $(OBJS) $(NAME).sql
+# things created by various check targets
+	rm -rf results tmp_check log
+	rm -f regression.diffs regression.out regress.out run_check.out
+ifeq ($(PORTNAME), win)
+	rm -f regress.def
+endif
+
+depend dep:
+	$(CC) -MM $(CFLAGS) *.c >depend
+
+ifeq (depend,$(wildcard depend))
+include depend
+endif

contrib/intarray/Makefile.703

+#-------------------------------------------------------------------------
+#
+# Makefile --
+#
+#    Makefile for Enzyme Commission catalogue number type -- ec_code
+#
+#-------------------------------------------------------------------------
+
+PGDIR = ../..
+SRCDIR = $(PGDIR)/src
+
+include $(SRCDIR)/Makefile.global
+
+INCLUDE_OPT =	-I ./ \
+		-I $(SRCDIR)/ \
+		-I $(SRCDIR)/include \
+		-I $(SRCDIR)/port/$(PORTNAME)
+
+CFLAGS += $(INCLUDE_OPT) $(CFLAGS_SL)
+
+MODNAME =	_int
+OBJFILES =	$(MODNAME).o
+
+SQLDEFS =	$(MODNAME).sql
+
+MODULE =	$(MODNAME)$(DLSUFFIX)
+
+MODDIR =	$(LIBDIR)/modules
+
+SQLDIR =	$(LIBDIR)/sql
+
+all:		module sql
+
+module:		$(MODULE)
+
+sql:		$(SQLDEFS)
+
+$(MODULE):	$(OBJFILES)
+		$(CC) $(CFLAGS) -shared -o $@ $(OBJFILES)
+
+install:	$(MODULE) $(SQLDEFS) $(MODDIR) $(SQLDIR)
+		cp -p $(MODULE) $(MODDIR)/
+		strip $(MODDIR)/$(MODULE)
+		cp -p $(SQLDEFS) $(SQLDIR)/
+
+$(MODDIR):
+		mkdir -p $@
+
+$(SQLDIR):
+		mkdir -p $@
+
+%.sql: %.sql.in
+		sed "s|MODULE_PATHNAME|$(MODDIR)/$(MODULE)|" < $< > $@
+
+depend dep:
+		$(CC) -MM $(INCLUDE_OPT) *.c >depend
+
+clean:
+		rm -f $(MODULE) $(SQLDEFS) *$(DLSUFFIX)
+		rm -f *~ *# *.b *.o *.output *.tab.h $(MODNAME)parse.h $(MODNAME)parse.c $(MODNAME)scan.c 
+
+ifeq (depend,$(wildcard depend))
+include depend
+endif

contrib/intarray/README.intarray

+This is an implementation of RD-tree data structure using GiST interface
+of PostgreSQL. It has built-in lossy compression - must be declared
+in index creation - with (islossy). Current implementation has index support 
+for one-dimensional array of int4's.
+All work was done by Teodor Sigaev (teodor@stack.net) and Oleg Bartunov
+(oleg@sai.msu.su). See http://www.sai.msu.su/~megera/postgres/gist
+for additional information.
+
+INSTALLATION:
+
+  gmake
+  gmake install
+  -- load functions
+  psql <database> < _int.sql 
+
+REGRESSION TEST:
+
+   gmake installcheck
+
+EXAMPLE USAGE:
+
+  create table message (mid int not null,sections int[]);
+  create table message_section_map (mid int not null,sid int not null);
+
+  -- create indices
+CREATE unique index message_key on message ( mid );
+CREATE unique index message_section_map_key2 on message_section_map (sid, mid );
+CREATE INDEX message_rdtree_idx on message using gist ( sections ) with ( islossy );
+
+  -- select some messages with section in 1 OR 2 - OVERLAP operator
+  select message.mid from message where message.sections && '{1,2}';  
+
+  -- select messages contains in sections 1 AND 2 - CONTAINS operator
+  select message.mid from message where message.sections @ '{1,2}';
+  -- the same, CONTAINED operator
+  select message.mid from message where '{1,2}' ~ message.sections;
+
+BENCHMARK:
+
+  subdirectory bench contains benchmark suite.
+  cd ./bench
+  1. createdb TEST
+  2. psql TEST < ../_int.sql
+  3. ./create_test.pl | psql TEST
+  4. ./bench.pl - perl script to benchmark queries, supports OR, AND queries
+                  with/without RD-Tree. Run script without arguments to 
+                  see availbale options.
+
+     a)test without RD-Tree (OR)
+       ./bench.pl -d TEST -s 1,2 -v
+     b)test with RD-Tree 
+       ./bench.pl -d TEST -s 1,2 -v -r
+
+BENCHMARKS:
+
+Size of table <message>: 200000
+Size of table <message_section_map>: 268538 
+
+Distribution of messages by sections:
+
+section 0: 73899 messages
+section 1: 16298 messages
+section 50: 1241 messages
+section 99: 705 messages
+
+old - without RD-Tree support,
+new - with RD-Tree
+
++----------+---------------+----------------+
+|Search set|OR, time in sec|AND, time in sec|
+|          +-------+-------+--------+-------+
+|          |  old  |  new  |   old  |  new  |
++----------+-------+-------+--------+-------+
+|         1|  1.427|  0.215|       -|      -|
++----------+-------+-------+--------+-------+
+|        99|  1.029|  0.018|       -|      -|
++----------+-------+-------+--------+-------+
+|       1,2|  1.829|  0.334|   5.654|  0.042|
++----------+-------+-------+--------+-------+
+| 1,2,50,60|  2.057|  0.359|   5.044|  0.007|
++----------+-------+-------+--------+-------+

contrib/intarray/_int.sql.in

+-- Create the user-defined type for the 1-D frloating point indervals (_int4)
+-- 
+BEGIN TRANSACTION;
+
+--
+-- External C-functions for R-tree methods
+--
+
+-- Comparison methods
+
+CREATE FUNCTION _int_contains(_int4, _int4) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+INSERT INTO pg_description (objoid, description)
+   SELECT oid, 'contains'::text
+   FROM pg_proc
+   WHERE proname = '_int_contains'::name;
+
+CREATE FUNCTION _int_contained(_int4, _int4) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+INSERT INTO pg_description (objoid, description)
+   SELECT oid, 'contained in'::text
+   FROM pg_proc
+   WHERE proname = '_int_contained'::name;
+
+CREATE FUNCTION _int_overlap(_int4, _int4) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+INSERT INTO pg_description (objoid, description)
+   SELECT oid, 'overlaps'::text
+   FROM pg_proc
+   WHERE proname = '_int_overlap'::name;
+
+CREATE FUNCTION _int_same(_int4, _int4) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+INSERT INTO pg_description (objoid, description)
+   SELECT oid, 'same as'::text
+   FROM pg_proc
+   WHERE proname = '_int_same'::name;
+
+CREATE FUNCTION _int_different(_int4, _int4) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+INSERT INTO pg_description (objoid, description)
+   SELECT oid, 'different'::text
+   FROM pg_proc
+   WHERE proname = '_int_different'::name;
+
+-- support routines for indexing
+
+CREATE FUNCTION _int_union(_int4, _int4) RETURNS _int4
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+CREATE FUNCTION _int_inter(_int4, _int4) RETURNS _int4
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+--
+-- OPERATORS
+--
+
+CREATE OPERATOR && (
+   LEFTARG = _int4, RIGHTARG = _int4, PROCEDURE = _int_overlap,
+   COMMUTATOR = '&&',
+   RESTRICT = contsel, JOIN = contjoinsel
+);
+
+--CREATE OPERATOR = (
+--   LEFTARG = _int4, RIGHTARG = _int4, PROCEDURE = _int_same,
+--   COMMUTATOR = '=', NEGATOR = '<>',
+--   RESTRICT = eqsel, JOIN = eqjoinsel,
+--   SORT1 = '<', SORT2 = '<'
+--);
+
+CREATE OPERATOR <> (
+   LEFTARG = _int4, RIGHTARG = _int4, PROCEDURE = _int_different,
+   COMMUTATOR = '<>', NEGATOR = '=',
+   RESTRICT = neqsel, JOIN = neqjoinsel
+);
+
+CREATE OPERATOR @ (
+   LEFTARG = _int4, RIGHTARG = _int4, PROCEDURE = _int_contains,
+   COMMUTATOR = '~', RESTRICT = contsel, JOIN = contjoinsel
+);
+
+CREATE OPERATOR ~ (
+   LEFTARG = _int4, RIGHTARG = _int4, PROCEDURE = _int_contained,
+   COMMUTATOR = '@', RESTRICT = contsel, JOIN = contjoinsel
+);
+
+
+-- define the GiST support methods
+CREATE FUNCTION g_int_consistent(opaque,_int4,int4) RETURNS bool
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+CREATE FUNCTION g_int_compress(opaque) RETURNS opaque 
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+CREATE FUNCTION g_int_decompress(opaque) RETURNS opaque 
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+CREATE FUNCTION g_int_penalty(opaque,opaque,opaque) RETURNS opaque
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+CREATE FUNCTION g_int_picksplit(opaque, opaque) RETURNS opaque
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+CREATE FUNCTION g_int_union(bytea, opaque) RETURNS _int4 
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+CREATE FUNCTION g_int_same(_int4, _int4, opaque) RETURNS opaque 
+	AS 'MODULE_PATHNAME' LANGUAGE 'c';
+
+
+-- register the default opclass for indexing
+INSERT INTO pg_opclass (opcname, opcdeftype)
+   SELECT 'gist__int_ops', oid
+   FROM pg_type
+   WHERE typname = '_int4';
+
+
+-- get the comparators for _intments and store them in a tmp table
+SELECT o.oid AS opoid, o.oprname
+INTO TABLE _int_ops_tmp
+FROM pg_operator o, pg_type t
+WHERE o.oprleft = t.oid and o.oprright = t.oid
+   and t.typname = '_int4';
+
+-- make sure we have the right operators
+-- SELECT * from _int_ops_tmp;
+
+-- using the tmp table, generate the amop entries 
+
+-- _int_overlap
+INSERT INTO pg_amop (amopid, amopclaid, amopopr, amopstrategy)
+   SELECT am.oid, opcl.oid, c.opoid, 3
+   FROM pg_am am, pg_opclass opcl, _int_ops_tmp c
+   WHERE amname = 'gist' and opcname = 'gist__int_ops' 
+      and c.oprname = '&&';
+
+-- _int_same
+INSERT INTO pg_amop (amopid, amopclaid, amopopr, amopstrategy)
+   SELECT am.oid, opcl.oid, c.opoid, 6
+   FROM pg_am am, pg_opclass opcl, _int_ops_tmp c
+   WHERE amname = 'gist' and opcname = 'gist__int_ops' 
+      and c.oprname = '=';
+
+-- _int_contains
+INSERT INTO pg_amop (amopid, amopclaid, amopopr, amopstrategy)
+   SELECT am.oid, opcl.oid, c.opoid, 7
+   FROM pg_am am, pg_opclass opcl, _int_ops_tmp c
+   WHERE amname = 'gist' and opcname = 'gist__int_ops' 
+      and c.oprname = '@';
+
+-- _int_contained
+INSERT INTO pg_amop (amopid, amopclaid, amopopr, amopstrategy)
+   SELECT am.oid, opcl.oid, c.opoid, 8
+   FROM pg_am am, pg_opclass opcl, _int_ops_tmp c
+   WHERE amname = 'gist' and opcname = 'gist__int_ops' 
+      and c.oprname = '~';
+
+DROP TABLE _int_ops_tmp;
+
+
+-- add the entries to amproc for the support methods
+-- note the amprocnum numbers associated with each are specific!
+
+INSERT INTO pg_amproc (amid, amopclaid, amproc, amprocnum)
+   SELECT am.oid, opcl.oid, pro.oid, 1
+   FROM pg_am am, pg_opclass opcl, pg_proc pro
+   WHERE  amname = 'gist' and opcname = 'gist__int_ops'
+      and proname = 'g_int_consistent';
+
+INSERT INTO pg_amproc (amid, amopclaid, amproc, amprocnum)
+   SELECT am.oid, opcl.oid, pro.oid, 2
+   FROM pg_am am, pg_opclass opcl, pg_proc pro
+   WHERE  amname = 'gist' and opcname = 'gist__int_ops'
+      and proname = 'g_int_union';
+
+INSERT INTO pg_amproc (amid, amopclaid, amproc, amprocnum)
+   SELECT am.oid, opcl.oid, pro.oid, 3
+   FROM pg_am am, pg_opclass opcl, pg_proc pro
+   WHERE  amname = 'gist' and opcname = 'gist__int_ops'
+      and proname = 'g_int_compress';
+
+INSERT INTO pg_amproc (amid, amopclaid, amproc, amprocnum)
+   SELECT am.oid, opcl.oid, pro.oid, 4
+   FROM pg_am am, pg_opclass opcl, pg_proc pro
+   WHERE  amname = 'gist' and opcname = 'gist__int_ops'
+      and proname = 'g_int_decompress';
+
+INSERT INTO pg_amproc (amid, amopclaid, amproc, amprocnum)
+   SELECT am.oid, opcl.oid, pro.oid, 5
+   FROM pg_am am, pg_opclass opcl, pg_proc pro
+   WHERE  amname = 'gist' and opcname = 'gist__int_ops'
+      and proname = 'g_int_penalty';
+
+INSERT INTO pg_amproc (amid, amopclaid, amproc, amprocnum)
+   SELECT am.oid, opcl.oid, pro.oid, 6
+   FROM pg_am am, pg_opclass opcl, pg_proc pro
+   WHERE  amname = 'gist' and opcname = 'gist__int_ops'
+      and proname = 'g_int_picksplit';
+
+INSERT INTO pg_amproc (amid, amopclaid, amproc, amprocnum)
+   SELECT am.oid, opcl.oid, pro.oid, 7
+   FROM pg_am am, pg_opclass opcl, pg_proc pro
+   WHERE  amname = 'gist' and opcname = 'gist__int_ops'
+      and proname = 'g_int_same';
+
+END TRANSACTION;

contrib/intarray/bench/bench.pl

+#!/usr/bin/perl 
+
+use strict;
+# make sure we are in a sane environment.
+use DBI();
+use DBD::Pg();
+use Time::HiRes qw( usleep ualarm gettimeofday tv_interval );
+use Getopt::Std;
+
+my %opt;
+getopts('d:b:s:veorauc', \%opt);
+
+if ( !( scalar %opt && defined $opt{s} ) ) {
+	print <<EOT;
+Usage:
+$0 -d DATABASE -s SECTIONS [-b NUMBER] [-v] [-e] [-o] [-r] [-a] [-u]
+-d DATABASE   	-DATABASE
+-b NUMBER   	-number of repeats
+-s SECTIONS 	-sections, format	sid1[,sid2[,sid3[...]]]]
+-v 		-verbose (show SQL)
+-e		-show explain
+-r		-use RD-tree index
+-a		-AND section
+-o		-show output
+-u		-unique
+-c 		-count
+
+EOT
+	exit;
+}
+
+$opt{d} ||= '_int4';
+my $dbi=DBI->connect('DBI:Pg:dbname='.$opt{d});
+
+my %table;
+my @where;
+
+$table{message}=1;
+
+if ( $opt{a} ) {
+	if ( $opt{r} ) {
+		push @where, "message.sections @ '{$opt{s}}'";
+	} else {
+		foreach my $sid ( split(/[,\s]+/, $opt{s} )) {
+			push @where, "EXISTS ( select  message_section_map.mid from message_section_map where message.mid=message_section_map.mid and message_section_map.sid = $sid )";
+		}
+	}
+} else {
+	if ( $opt{r} ) {
+		push @where, "message.sections && '{$opt{s}}'";
+	} else {
+		$table{message_section_map} = 1;
+		push @where, "message.mid = message_section_map.mid";
+		push @where, "message_section_map.sid in ($opt{s})";
+	}
+}
+
+my $outf;
+if ( $opt{c} ) {
+	$outf = ( $opt{u} ) ? 'count( distinct message.mid )' : 'count( message.mid )';
+} else {
+	$outf = ( $opt{u} ) ? 'distinct( message.mid )' : 'message.mid';
+}
+my $sql = "select $outf from ".join(', ', keys %table)." where ".join(' AND ', @where).';';
+
+if ( $opt{v} ) {
+	print "$sql\n";
+}
+
+if ( $opt{e} ) {
+	$dbi->do("explain $sql");
+}
+
+my $t0 = [gettimeofday];
+my $count=0;
+my $b=$opt{b};
+$b||=1;
+my @a;
+foreach ( 1..$b ) {
+	@a=exec_sql($dbi,$sql);
+	$count=$#a;
+}
+my $elapsed = tv_interval ( $t0, [gettimeofday]);
+if ( $opt{o} ) {
+	foreach ( @a ) {
+		print "$_->{mid}\t$_->{sections}\n";
+	}
+} 
+print sprintf("total: %.02f sec; number: %d; for one: %.03f sec; found %d docs\n", $elapsed, $b, $elapsed/$b, $count+1 );
+$dbi -> disconnect;
+
+sub exec_sql {
+        my ($dbi, $sql, @keys) = @_;
+        my $sth=$dbi->prepare($sql) || die;
+        $sth->execute( @keys ) || die; 
+        my $r;  
+        my @row;
+        while ( defined ( $r=$sth->fetchrow_hashref ) ) {
+                push @row, $r;
+        }               
+        $sth->finish;   
+        return @row;
+}
+

contrib/intarray/bench/create_test.pl

+#!/usr/bin/perl
+
+use strict;
+print <<EOT;
+create table message (
+	mid	int not null,
+	sections	int[]
+);
+create table message_section_map (
+	mid 	int not null,
+	sid	int not null
+);
+
+EOT
+
+open(MSG,">message.tmp") || die;
+open(MAP,">message_section_map.tmp") || die;
+
+srand( 1 );
+#foreach my $i ( 1..1778 ) {
+#foreach my $i ( 1..3443 ) {
+#foreach my $i ( 1..5000 ) {
+#foreach my $i ( 1..29362 ) {
+#foreach my $i ( 1..33331 ) {
+#foreach my $i ( 1..83268 ) {
+foreach my $i ( 1..200000 ) {
+	my @sect;
+	if ( rand() < 0.7 ) {
+		$sect[0] = int( (rand()**4)*100 );
+	} else {
+		my %hash;
+		@sect = grep { $hash{$_}++; $hash{$_} <= 1 } map { int( (rand()**4)*100) } 0..( int(rand()*5) );
+	}
+	if ( $#sect < 0 || rand() < 0.1 ) {
+		print MSG "$i\t\\N\n";
+	} else {
+		print MSG "$i\t{".join(',',@sect)."}\n";
+		map { print MAP "$i\t$_\n" } @sect;
+	}
+}
+close MAP;
+close MSG;
+
+copytable('message');
+copytable('message_section_map');
+
+print <<EOT;
+
+CREATE unique index message_key on message ( mid );
+--CREATE unique index message_section_map_key1 on message_section_map ( mid, sid );
+CREATE unique index message_section_map_key2 on message_section_map ( sid, mid );
+CREATE INDEX message_rdtree_idx on message using gist ( sections ) with ( islossy );
+VACUUM ANALYZE;
+
+select count(*) from message;
+select count(*) from message_section_map;
+
+
+
+EOT
+
+
+unlink 'message.tmp', 'message_section_map.tmp';
+
+sub copytable {
+	my $t = shift;
+	
+	print "COPY $t from stdin;\n";
+	open( FFF, "$t.tmp") || die;
+	while(<FFF>) { print; }
+	close FFF;
+	print "\\.\n";
+}

contrib/intarray/expected/_int.out

+--
+-- first, define the datatype.  Turn off echoing so that expected file
+-- does not depend on contents of seg.sql.
+--
+\set ECHO none
+CREATE TABLE test__int( a int[] );
+\copy test__int from 'data/test__int.data'
+SELECT count(*) from test__int WHERE a && '{23,50}';
+ count 
+-------
+   345
+(1 row)
+
+SELECT count(*) from test__int WHERE a @ '{23,50}';
+ count 
+-------
+    12
+(1 row)
+

contrib/intarray/sql/_int.sql

+--
+-- first, define the datatype.  Turn off echoing so that expected file
+-- does not depend on contents of seg.sql.
+--
+\set ECHO none
+\i _int.sql
+\set ECHO all
+
+CREATE TABLE test__int( a int[] );
+
+\copy test__int from 'data/test__int.data'
+
+SELECT count(*) from test__int WHERE a && '{23,50}';
+SELECT count(*) from test__int WHERE a @ '{23,50}';
+