1# 2007 November 29
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# This file tests the optimisations made in November 2007 of expressions 
12# of the following form:
13#
14#     <value> IN (SELECT <column> FROM <table>)
15#
16# $Id: in3.test,v 1.5 2008/08/04 03:51:24 danielk1977 Exp $
17
18set testdir [file dirname $argv0]
19source $testdir/tester.tcl
20
21ifcapable !subquery {
22  finish_test
23  return
24}
25
26# Return the number of OpenEphemeral instructions used in the
27# implementation of the sql statement passed as a an argument.
28#
29proc nEphemeral {sql} {
30  set nEph 0
31  foreach op [execsql "EXPLAIN $sql"] {
32    if {$op eq "OpenEphemeral"} {incr nEph}
33  }
34  set nEph
35}
36
37# This proc works the same way as execsql, except that the number
38# of OpenEphemeral instructions used in the implementation of the
39# statement is inserted into the start of the returned list.
40#
41proc exec_neph {sql} {
42  return [concat [nEphemeral $sql] [execsql $sql]]
43}
44
45do_test in3-1.1 {
46  execsql {
47    CREATE TABLE t1(a PRIMARY KEY, b);
48    INSERT INTO t1 VALUES(1, 2);
49    INSERT INTO t1 VALUES(3, 4);
50    INSERT INTO t1 VALUES(5, 6);
51  }
52} {}
53
54# All of these queries should avoid using a temp-table:
55#
56do_test in3-1.2 {
57  exec_neph { SELECT rowid FROM t1 WHERE rowid IN (SELECT rowid FROM t1); }
58} {0 1 2 3}
59do_test in3-1.3 {
60  exec_neph { SELECT a FROM t1 WHERE a IN (SELECT a FROM t1); }
61} {0 1 3 5}
62do_test in3-1.4 {
63  exec_neph { SELECT rowid FROM t1 WHERE rowid+0 IN (SELECT rowid FROM t1); }
64} {0 1 2 3}
65do_test in3-1.5 {
66  exec_neph { SELECT a FROM t1 WHERE a+0 IN (SELECT a FROM t1); }
67} {0 1 3 5}
68
69# Because none of the sub-select queries in the following statements
70# match the pattern ("SELECT <column> FROM <table>"), the following do 
71# require a temp table.
72#
73do_test in3-1.6 {
74  exec_neph { SELECT rowid FROM t1 WHERE rowid IN (SELECT rowid+0 FROM t1); }
75} {1 1 2 3}
76do_test in3-1.7 {
77  exec_neph { SELECT a FROM t1 WHERE a IN (SELECT a+0 FROM t1); }
78} {1 1 3 5}
79do_test in3-1.8 {
80  exec_neph { SELECT a FROM t1 WHERE a IN (SELECT a FROM t1 WHERE 1); }
81} {1 1 3 5}
82do_test in3-1.9 {
83  exec_neph { SELECT a FROM t1 WHERE a IN (SELECT a FROM t1 GROUP BY a); }
84} {1 1 3 5}
85
86# This should not use a temp-table. Even though the sub-select does
87# not exactly match the pattern "SELECT <column> FROM <table>", in
88# this case the ORDER BY is a no-op and can be ignored.
89do_test in3-1.10 {
90  exec_neph { SELECT a FROM t1 WHERE a IN (SELECT a FROM t1 ORDER BY a); }
91} {0 1 3 5}
92
93# These do use the temp-table. Adding the LIMIT clause means the 
94# ORDER BY cannot be ignored.
95do_test in3-1.11 {
96  exec_neph {SELECT a FROM t1 WHERE a IN (SELECT a FROM t1 ORDER BY a LIMIT 1)}
97} {1 1}
98do_test in3-1.12 {
99  exec_neph {
100    SELECT a FROM t1 WHERE a IN (SELECT a FROM t1 ORDER BY a LIMIT 1 OFFSET 1)
101  }
102} {1 3}
103
104# Has to use a temp-table because of the compound sub-select.
105#
106ifcapable compound {
107  do_test in3-1.13 {
108    exec_neph {
109      SELECT a FROM t1 WHERE a IN (
110        SELECT a FROM t1 UNION ALL SELECT a FROM t1
111      )
112    }
113  } {1 1 3 5}
114}
115
116# The first of these queries has to use the temp-table, because the 
117# collation sequence used for the index on "t1.a" does not match the
118# collation sequence used by the "IN" comparison. The second does not
119# require a temp-table, because the collation sequences match.
120#
121do_test in3-1.14 {
122  exec_neph { SELECT a FROM t1 WHERE a COLLATE nocase IN (SELECT a FROM t1) }
123} {1 1 3 5}
124do_test in3-1.15 {
125  exec_neph { SELECT a FROM t1 WHERE a COLLATE binary IN (SELECT a FROM t1) }
126} {0 1 3 5}
127
128# Neither of these queries require a temp-table. The collation sequence
129# makes no difference when using a rowid.
130#
131do_test in3-1.16 {
132  exec_neph {SELECT a FROM t1 WHERE a COLLATE nocase IN (SELECT rowid FROM t1)}
133} {0 1 3}
134do_test in3-1.17 {
135  exec_neph {SELECT a FROM t1 WHERE a COLLATE binary IN (SELECT rowid FROM t1)}
136} {0 1 3}
137
138# The following tests - in3.2.* - test a bug that was difficult to track
139# down during development. They are not particularly well focused.
140#
141do_test in3-2.1 {
142  execsql {
143    DROP TABLE IF EXISTS t1;
144    CREATE TABLE t1(w int, x int, y int);
145    CREATE TABLE t2(p int, q int, r int, s int);
146  }
147  for {set i 1} {$i<=100} {incr i} {
148    set w $i
149    set x [expr {int(log($i)/log(2))}]
150    set y [expr {$i*$i + 2*$i + 1}]
151    execsql "INSERT INTO t1 VALUES($w,$x,$y)"
152  }
153  set maxy [execsql {select max(y) from t1}]
154  db eval { INSERT INTO t2 SELECT 101-w, x, $maxy+1-y, y FROM t1 }
155} {}
156do_test in3-2.2 {
157  execsql {
158    SELECT rowid 
159    FROM t1 
160    WHERE rowid IN (SELECT rowid FROM t1 WHERE rowid IN (1, 2));
161  }
162} {1 2}
163do_test in3-2.3 {
164  execsql {
165    select rowid from t1 where rowid IN (-1,2,4)
166  }
167} {2 4}
168do_test in3-2.4 {
169  execsql {
170    SELECT rowid FROM t1 WHERE rowid IN 
171       (select rowid from t1 where rowid IN (-1,2,4))
172  }
173} {2 4}
174
175#-------------------------------------------------------------------------
176# This next block of tests - in3-3.* - verify that column affinity is
177# correctly handled in cases where an index might be used to optimise
178# an IN (SELECT) expression.
179#
180do_test in3-3.1 {
181  catch {execsql {
182    DROP TABLE t1;
183    DROP TABLE t2;
184  }}
185
186  execsql {
187
188    CREATE TABLE t1(a BLOB, b NUMBER ,c TEXT);
189    CREATE UNIQUE INDEX t1_i1 ON t1(a);        /* no affinity */
190    CREATE UNIQUE INDEX t1_i2 ON t1(b);        /* numeric affinity */
191    CREATE UNIQUE INDEX t1_i3 ON t1(c);        /* text affinity */
192
193    CREATE TABLE t2(x BLOB, y NUMBER, z TEXT);
194    CREATE UNIQUE INDEX t2_i1 ON t2(x);        /* no affinity */
195    CREATE UNIQUE INDEX t2_i2 ON t2(y);        /* numeric affinity */
196    CREATE UNIQUE INDEX t2_i3 ON t2(z);        /* text affinity */
197
198    INSERT INTO t1 VALUES(1, 1, 1);
199    INSERT INTO t2 VALUES('1', '1', '1');
200  }
201} {}
202
203do_test in3-3.2 {
204  # No affinity is applied before comparing "x" and "a". Therefore
205  # the index can be used (the comparison is false, text!=number).
206  exec_neph { SELECT x IN (SELECT a FROM t1) FROM t2 }
207} {0 0}
208do_test in3-3.3 {
209  # Logically, numeric affinity is applied to both sides before 
210  # the comparison.  Therefore it is possible to use index t1_i2.
211  exec_neph { SELECT x IN (SELECT b FROM t1) FROM t2 }
212} {0 1}
213do_test in3-3.4 {
214  # No affinity is applied before the comparison takes place. Making
215  # it possible to use index t1_i3.
216  exec_neph { SELECT x IN (SELECT c FROM t1) FROM t2 }
217} {0 1}
218
219do_test in3-3.5 {
220  # Numeric affinity should be applied to each side before the comparison
221  # takes place. Therefore we cannot use index t1_i1, which has no affinity.
222  exec_neph { SELECT y IN (SELECT a FROM t1) FROM t2 }
223} {1 1}
224do_test in3-3.6 {
225  # Numeric affinity is applied to both sides before 
226  # the comparison.  Therefore it is possible to use index t1_i2.
227  exec_neph { SELECT y IN (SELECT b FROM t1) FROM t2 }
228} {0 1}
229do_test in3-3.7 {
230  # Numeric affinity is applied before the comparison takes place. 
231  # Making it impossible to use index t1_i3.
232  exec_neph { SELECT y IN (SELECT c FROM t1) FROM t2 }
233} {1 1}
234
235#---------------------------------------------------------------------
236#
237# Test using a multi-column index.
238#
239do_test in3-4.1 {
240  execsql {
241    CREATE TABLE t3(a, b, c);
242    CREATE UNIQUE INDEX t3_i ON t3(b, a);
243  }
244
245  execsql {
246    INSERT INTO t3 VALUES(1, 'numeric', 2);
247    INSERT INTO t3 VALUES(2, 'text', 2);
248    INSERT INTO t3 VALUES(3, 'real', 2);
249    INSERT INTO t3 VALUES(4, 'none', 2);
250  }
251} {}
252do_test in3-4.2 {
253  exec_neph { SELECT 'text' IN (SELECT b FROM t3) }
254} {0 1}
255do_test in3-4.3 {
256  exec_neph { SELECT 'TEXT' COLLATE nocase IN (SELECT b FROM t3) }
257} {1 1}
258do_test in3-4.4 {
259  # A temp table must be used because t3_i.b is not guaranteed to be unique.
260  exec_neph { SELECT b FROM t3 WHERE b IN (SELECT b FROM t3) }
261} {1 none numeric real text}
262do_test in3-4.5 {
263  execsql { CREATE UNIQUE INDEX t3_i2 ON t3(b) }
264  exec_neph { SELECT b FROM t3 WHERE b IN (SELECT b FROM t3) }
265} {0 none numeric real text}
266do_test in3-4.6 {
267  execsql { DROP INDEX t3_i2 }
268} {}
269
270# The following two test cases verify that ticket #2991 has been fixed.
271#
272do_test in3-5.1 {
273  execsql {
274    CREATE TABLE Folders(
275      folderid INTEGER PRIMARY KEY, 
276      parentid INTEGER, 
277      rootid INTEGER, 
278      path VARCHAR(255)
279    );
280  }
281} {}
282do_test in3-5.2 {
283  catchsql {
284    DELETE FROM Folders WHERE folderid IN
285    (SELECT folderid FROM Folder WHERE path LIKE 'C:\MP3\Albums\' || '%');
286  }
287} {1 {no such table: Folder}}
288
289finish_test
290