1# 2011 January 19 2# 3# The author disclaims copyright to this source code. In place of 4# a legal notice, here is a blessing: 5# 6# May you do good and not evil. 7# May you find forgiveness for yourself and forgive others. 8# May you share freely, never taking more than you give. 9# 10#*********************************************************************** 11# 12# This file implements tests for SQLite library. The focus of the tests 13# in this file is the use of the sqlite_stat4 histogram data on tables 14# with many repeated values and only a few distinct values. 15# 16 17set testdir [file dirname $argv0] 18source $testdir/tester.tcl 19 20ifcapable !stat4 { 21 finish_test 22 return 23} 24 25set testprefix analyze5 26 27proc eqp {sql {db db}} { 28 uplevel execsql [list "EXPLAIN QUERY PLAN $sql"] $db 29} 30 31proc alpha {blob} { 32 set ret "" 33 foreach c [split $blob {}] { 34 if {[string is alpha $c]} {append ret $c} 35 } 36 return $ret 37} 38db func alpha alpha 39 40unset -nocomplain i t u v w x y z 41do_test analyze5-1.0 { 42 db eval {CREATE TABLE t1(t,u,v TEXT COLLATE nocase,w,x,y,z)} 43 for {set i 0} {$i < 1000} {incr i} { 44 set y [expr {$i>=25 && $i<=50}] 45 set z [expr {($i>=400) + ($i>=700) + ($i>=875)}] 46 set x $z 47 set w $z 48 set t [expr {$z+0.5}] 49 switch $z { 50 0 {set u "alpha"; unset x} 51 1 {set u "bravo"} 52 2 {set u "charlie"} 53 3 {set u "delta"; unset w} 54 } 55 if {$i%2} {set v $u} {set v [string toupper $u]} 56 db eval {INSERT INTO t1 VALUES($t,$u,$v,$w,$x,$y,$z)} 57 } 58 db eval { 59 CREATE INDEX t1t ON t1(t); -- 0.5, 1.5, 2.5, and 3.5 60 CREATE INDEX t1u ON t1(u); -- text 61 CREATE INDEX t1v ON t1(v); -- mixed case text 62 CREATE INDEX t1w ON t1(w); -- integers 0, 1, 2 and a few NULLs 63 CREATE INDEX t1x ON t1(x); -- integers 1, 2, 3 and many NULLs 64 CREATE INDEX t1y ON t1(y); -- integers 0 and very few 1s 65 CREATE INDEX t1z ON t1(z); -- integers 0, 1, 2, and 3 66 ANALYZE; 67 SELECT alpha(sample) FROM sqlite_stat4 WHERE idx='t1u' ORDER BY nlt; 68 } 69} {alpha bravo charlie delta} 70 71do_test analyze5-1.1 { 72 db eval { 73 SELECT DISTINCT lower(alpha(sample)) FROM sqlite_stat4 WHERE idx='t1v' 74 ORDER BY 1 75 } 76} {alpha bravo charlie delta} 77do_test analyze5-1.2 { 78 db eval {SELECT idx, count(*) FROM sqlite_stat4 GROUP BY 1 ORDER BY 1} 79} {t1t 4 t1u 4 t1v 4 t1w 4 t1x 4 t1y 2 t1z 4} 80 81# Verify that range queries generate the correct row count estimates 82# 83foreach {testid where index rows} { 84 1 {z>=0 AND z<=0} t1z 400 85 2 {z>=1 AND z<=1} t1z 300 86 3 {z>=2 AND z<=2} t1z 175 87 4 {z>=3 AND z<=3} t1z 125 88 5 {z>=4 AND z<=4} t1z 1 89 6 {z>=-1 AND z<=-1} t1z 1 90 7 {z>1 AND z<3} t1z 175 91 8 {z>0 AND z<100} t1z 600 92 9 {z>=1 AND z<100} t1z 600 93 10 {z>1 AND z<100} t1z 300 94 11 {z>=2 AND z<100} t1z 300 95 12 {z>2 AND z<100} t1z 125 96 13 {z>=3 AND z<100} t1z 125 97 14 {z>3 AND z<100} t1z 1 98 15 {z>=4 AND z<100} t1z 1 99 16 {z>=-100 AND z<=-1} t1z 1 100 17 {z>=-100 AND z<=0} t1z 400 101 18 {z>=-100 AND z<0} t1z 1 102 19 {z>=-100 AND z<=1} t1z 700 103 20 {z>=-100 AND z<2} t1z 700 104 21 {z>=-100 AND z<=2} t1z 875 105 22 {z>=-100 AND z<3} t1z 875 106 107 31 {z>=0.0 AND z<=0.0} t1z 400 108 32 {z>=1.0 AND z<=1.0} t1z 300 109 33 {z>=2.0 AND z<=2.0} t1z 175 110 34 {z>=3.0 AND z<=3.0} t1z 125 111 35 {z>=4.0 AND z<=4.0} t1z 1 112 36 {z>=-1.0 AND z<=-1.0} t1z 1 113 37 {z>1.5 AND z<3.0} t1z 174 114 38 {z>0.5 AND z<100} t1z 599 115 39 {z>=1.0 AND z<100} t1z 600 116 40 {z>1.5 AND z<100} t1z 299 117 41 {z>=2.0 AND z<100} t1z 300 118 42 {z>2.1 AND z<100} t1z 124 119 43 {z>=3.0 AND z<100} t1z 125 120 44 {z>3.2 AND z<100} t1z 1 121 45 {z>=4.0 AND z<100} t1z 1 122 46 {z>=-100 AND z<=-1.0} t1z 1 123 47 {z>=-100 AND z<=0.0} t1z 400 124 48 {z>=-100 AND z<0.0} t1z 1 125 49 {z>=-100 AND z<=1.0} t1z 700 126 50 {z>=-100 AND z<2.0} t1z 700 127 51 {z>=-100 AND z<=2.0} t1z 875 128 52 {z>=-100 AND z<3.0} t1z 875 129 130 101 {z=-1} t1z 1 131 102 {z=0} t1z 400 132 103 {z=1} t1z 300 133 104 {z=2} t1z 175 134 105 {z=3} t1z 125 135 106 {z=4} t1z 1 136 107 {z=-10.0} t1z 1 137 108 {z=0.0} t1z 400 138 109 {z=1.0} t1z 300 139 110 {z=2.0} t1z 175 140 111 {z=3.0} t1z 125 141 112 {z=4.0} t1z 1 142 113 {z=1.5} t1z 1 143 114 {z=2.5} t1z 1 144 145 201 {z IN (-1)} t1z 1 146 202 {z IN (0)} t1z 400 147 203 {z IN (1)} t1z 300 148 204 {z IN (2)} t1z 175 149 205 {z IN (3)} t1z 125 150 206 {z IN (4)} t1z 1 151 207 {z IN (0.5)} t1z 1 152 208 {z IN (0,1)} t1z 700 153 209 {z IN (0,1,2)} t1z 875 154 210 {z IN (0,1,2,3)} {} 100 155 211 {z IN (0,1,2,3,4,5)} {} 100 156 212 {z IN (1,2)} t1z 475 157 213 {z IN (2,3)} t1z 300 158 214 {z=3 OR z=2} t1z 300 159 215 {z IN (-1,3)} t1z 126 160 216 {z=-1 OR z=3} t1z 126 161 162 300 {y=0} t1y 974 163 301 {y=1} t1y 26 164 302 {y=0.1} t1y 1 165 166 400 {x IS NULL} t1x 400 167 168} { 169 # Verify that the expected index is used with the expected row count 170 # No longer valid due to an EXPLAIN QUERY PLAN output format change 171 # do_test analyze5-1.${testid}a { 172 # set x [lindex [eqp "SELECT * FROM t1 WHERE $where"] 3] 173 # set idx {} 174 # regexp {INDEX (t1.) } $x all idx 175 # regexp {~([0-9]+) rows} $x all nrow 176 # list $idx $nrow 177 # } [list $index $rows] 178 179 # Verify that the same result is achieved regardless of whether or not 180 # the index is used 181 do_test analyze5-1.${testid}b { 182 set w2 [string map {y +y z +z} $where] 183 set a1 [db eval "SELECT rowid FROM t1 NOT INDEXED WHERE $w2\ 184 ORDER BY +rowid"] 185 set a2 [db eval "SELECT rowid FROM t1 WHERE $where ORDER BY +rowid"] 186 if {$a1==$a2} { 187 set res ok 188 } else { 189 set res "a1=\[$a1\] a2=\[$a2\]" 190 } 191 set res 192 } {ok} 193} 194 195# Increase the number of NULLs in column x 196# 197db eval { 198 UPDATE t1 SET x=NULL; 199 UPDATE t1 SET x=rowid 200 WHERE rowid IN (SELECT rowid FROM t1 ORDER BY random() LIMIT 5); 201 ANALYZE; 202} 203 204# Verify that range queries generate the correct row count estimates 205# 206foreach {testid where index rows} { 207 500 {x IS NULL AND u='charlie'} t1u 17 208 501 {x=1 AND u='charlie'} t1x 1 209 502 {x IS NULL} t1x 995 210 503 {x=1} t1x 1 211 504 {x IS NOT NULL} t1x 2 212 505 {+x IS NOT NULL} {} 500 213 506 {upper(x) IS NOT NULL} {} 500 214 215} { 216 # Verify that the expected index is used with the expected row count 217 # No longer valid due to an EXPLAIN QUERY PLAN format change 218 # do_test analyze5-1.${testid}a { 219 # set x [lindex [eqp "SELECT * FROM t1 WHERE $where"] 3] 220 # set idx {} 221 # regexp {INDEX (t1.) } $x all idx 222 # regexp {~([0-9]+) rows} $x all nrow 223 # list $idx $nrow 224 # } [list $index $rows] 225 226 # Verify that the same result is achieved regardless of whether or not 227 # the index is used 228 do_test analyze5-1.${testid}b { 229 set w2 [string map {y +y z +z} $where] 230 set a1 [db eval "SELECT rowid FROM t1 NOT INDEXED WHERE $w2\ 231 ORDER BY +rowid"] 232 set a2 [db eval "SELECT rowid FROM t1 WHERE $where ORDER BY +rowid"] 233 if {$a1==$a2} { 234 set res ok 235 } else { 236 set res "a1=\[$a1\] a2=\[$a2\]" 237 } 238 set res 239 } {ok} 240} 241 242finish_test 243