| blob | a0469da12546cb35308b47bea72a43bfcb35f6e9 |
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_stat2 histogram data on tables
14 # with many repeated values and only a few distinct values.
15 #
17 set testdir [file dirname $argv0]
18 source $testdir/tester.tcl
20 ifcapable !stat2 {
21 finish_test
22 return
23 }
25 set testprefix analyze5
27 proc eqp {sql {db db}} {
28 uplevel execsql [list "EXPLAIN QUERY PLAN $sql"] $db
29 }
31 unset -nocomplain i t u v w x y z
32 do_test analyze5-1.0 {
33 db eval {CREATE TABLE t1(t,u,v TEXT COLLATE nocase,w,x,y,z)}
34 for {set i 0} {$i < 1000} {incr i} {
35 set y [expr {$i>=25 && $i<=50}]
36 set z [expr {($i>=400) + ($i>=700) + ($i>=875)}]
37 set x $z
38 set w $z
39 set t [expr {$z+0.5}]
40 switch $z {
41 0 {set u "alpha"; unset x}
42 1 {set u "bravo"}
43 2 {set u "charlie"}
44 3 {set u "delta"; unset w}
45 }
46 if {$i%2} {set v $u} {set v [string toupper $u]}
47 db eval {INSERT INTO t1 VALUES($t,$u,$v,$w,$x,$y,$z)}
48 }
49 db eval {
50 CREATE INDEX t1t ON t1(t); -- 0.5, 1.5, 2.5, and 3.5
51 CREATE INDEX t1u ON t1(u); -- text
52 CREATE INDEX t1v ON t1(v); -- mixed case text
53 CREATE INDEX t1w ON t1(w); -- integers 0, 1, 2 and a few NULLs
54 CREATE INDEX t1x ON t1(x); -- integers 1, 2, 3 and many NULLs
55 CREATE INDEX t1y ON t1(y); -- integers 0 and very few 1s
56 CREATE INDEX t1z ON t1(z); -- integers 0, 1, 2, and 3
57 ANALYZE;
58 SELECT sample FROM sqlite_stat2 WHERE idx='t1u' ORDER BY sampleno;
59 }
60 } {alpha alpha alpha alpha bravo bravo bravo charlie charlie delta}
61 do_test analyze5-1.1 {
62 string tolower \
63 [db eval {SELECT sample from sqlite_stat2 WHERE idx='t1v' ORDER BY sampleno}]
64 } {alpha alpha alpha alpha bravo bravo bravo charlie charlie delta}
65 do_test analyze5-1.2 {
66 db eval {SELECT sample from sqlite_stat2 WHERE idx='t1w' ORDER BY sampleno}
67 } {{} 0 0 0 0 1 1 1 2 2}
68 do_test analyze5-1.3 {
69 db eval {SELECT sample from sqlite_stat2 WHERE idx='t1x' ORDER BY sampleno}
70 } {{} {} {} {} 1 1 1 2 2 3}
71 do_test analyze5-1.4 {
72 db eval {SELECT sample from sqlite_stat2 WHERE idx='t1y' ORDER BY sampleno}
73 } {0 0 0 0 0 0 0 0 0 0}
74 do_test analyze5-1.5 {
75 db eval {SELECT sample from sqlite_stat2 WHERE idx='t1z' ORDER BY sampleno}
76 } {0 0 0 0 1 1 1 2 2 3}
77 do_test analyze5-1.6 {
78 db eval {SELECT sample from sqlite_stat2 WHERE idx='t1t' ORDER BY sampleno}
79 } {0.5 0.5 0.5 0.5 1.5 1.5 1.5 2.5 2.5 3.5}
82 # Verify that range queries generate the correct row count estimates
83 #
84 foreach {testid where index rows} {
85 1 {z>=0 AND z<=0} t1z 400
86 2 {z>=1 AND z<=1} t1z 300
87 3 {z>=2 AND z<=2} t1z 200
88 4 {z>=3 AND z<=3} t1z 100
89 5 {z>=4 AND z<=4} t1z 50
90 6 {z>=-1 AND z<=-1} t1z 50
91 7 {z>1 AND z<3} t1z 200
92 8 {z>0 AND z<100} t1z 600
93 9 {z>=1 AND z<100} t1z 600
94 10 {z>1 AND z<100} t1z 300
95 11 {z>=2 AND z<100} t1z 300
96 12 {z>2 AND z<100} t1z 100
97 13 {z>=3 AND z<100} t1z 100
98 14 {z>3 AND z<100} t1z 50
99 15 {z>=4 AND z<100} t1z 50
100 16 {z>=-100 AND z<=-1} t1z 50
101 17 {z>=-100 AND z<=0} t1z 400
102 18 {z>=-100 AND z<0} t1z 50
103 19 {z>=-100 AND z<=1} t1z 700
104 20 {z>=-100 AND z<2} t1z 700
105 21 {z>=-100 AND z<=2} t1z 900
106 22 {z>=-100 AND z<3} t1z 900
108 31 {z>=0.0 AND z<=0.0} t1z 400
109 32 {z>=1.0 AND z<=1.0} t1z 300
110 33 {z>=2.0 AND z<=2.0} t1z 200
111 34 {z>=3.0 AND z<=3.0} t1z 100
112 35 {z>=4.0 AND z<=4.0} t1z 50
113 36 {z>=-1.0 AND z<=-1.0} t1z 50
114 37 {z>1.5 AND z<3.0} t1z 200
115 38 {z>0.5 AND z<100} t1z 600
116 39 {z>=1.0 AND z<100} t1z 600
117 40 {z>1.5 AND z<100} t1z 300
118 41 {z>=2.0 AND z<100} t1z 300
119 42 {z>2.1 AND z<100} t1z 100
120 43 {z>=3.0 AND z<100} t1z 100
121 44 {z>3.2 AND z<100} t1z 50
122 45 {z>=4.0 AND z<100} t1z 50
123 46 {z>=-100 AND z<=-1.0} t1z 50
124 47 {z>=-100 AND z<=0.0} t1z 400
125 48 {z>=-100 AND z<0.0} t1z 50
126 49 {z>=-100 AND z<=1.0} t1z 700
127 50 {z>=-100 AND z<2.0} t1z 700
128 51 {z>=-100 AND z<=2.0} t1z 900
129 52 {z>=-100 AND z<3.0} t1z 900
131 101 {z=-1} t1z 50
132 102 {z=0} t1z 400
133 103 {z=1} t1z 300
134 104 {z=2} t1z 200
135 105 {z=3} t1z 100
136 106 {z=4} t1z 50
137 107 {z=-10.0} t1z 50
138 108 {z=0.0} t1z 400
139 109 {z=1.0} t1z 300
140 110 {z=2.0} t1z 200
141 111 {z=3.0} t1z 100
142 112 {z=4.0} t1z 50
143 113 {z=1.5} t1z 50
144 114 {z=2.5} t1z 50
146 201 {z IN (-1)} t1z 50
147 202 {z IN (0)} t1z 400
148 203 {z IN (1)} t1z 300
149 204 {z IN (2)} t1z 200
150 205 {z IN (3)} t1z 100
151 206 {z IN (4)} t1z 50
152 207 {z IN (0.5)} t1z 50
153 208 {z IN (0,1)} t1z 700
154 209 {z IN (0,1,2)} t1z 900
155 210 {z IN (0,1,2,3)} {} 100
156 211 {z IN (0,1,2,3,4,5)} {} 100
157 212 {z IN (1,2)} t1z 500
158 213 {z IN (2,3)} t1z 300
159 214 {z=3 OR z=2} t1z 300
160 215 {z IN (-1,3)} t1z 150
161 216 {z=-1 OR z=3} t1z 150
163 300 {y=0} {} 100
164 301 {y=1} t1y 50
165 302 {y=0.1} t1y 50
167 400 {x IS NULL} t1x 400
169 } {
170 # Verify that the expected index is used with the expected row count
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 $rows]
179 # Verify that the same result is achieved regardless of whether or not
180 # the index is used
181 do_test analyze5-1.${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 }
195 # Increase the number of NULLs in column x
196 #
197 db 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 }
204 # Verify that range queries generate the correct row count estimates
205 #
206 foreach {testid where index rows} {
207 500 {x IS NULL AND u='charlie'} t1u 20
208 501 {x=1 AND u='charlie'} t1x 5
209 502 {x IS NULL} {} 100
210 503 {x=1} t1x 50
211 504 {x IS NOT NULL} t1x 25
212 505 {+x IS NOT NULL} {} 500
213 506 {upper(x) IS NOT NULL} {} 500
215 } {
216 # Verify that the expected index is used with the expected row count
217 do_test analyze5-1.${testid}a {
218 set x [lindex [eqp "SELECT * FROM t1 WHERE $where"] 3]
219 set idx {}
220 regexp {INDEX (t1.) } $x all idx
221 regexp {~([0-9]+) rows} $x all nrow
222 list $idx $nrow
223 } [list $index $rows]
225 # Verify that the same result is achieved regardless of whether or not
226 # the index is used
227 do_test analyze5-1.${testid}b {
228 set w2 [string map {y +y z +z} $where]
229 set a1 [db eval "SELECT rowid FROM t1 NOT INDEXED WHERE $w2\
230 ORDER BY +rowid"]
231 set a2 [db eval "SELECT rowid FROM t1 WHERE $where ORDER BY +rowid"]
232 if {$a1==$a2} {
233 set res ok
234 } else {
235 set res "a1=\[$a1\] a2=\[$a2\]"
236 }
237 set res
238 } {ok}
239 }
241 finish_test