0
0
mirror of https://github.com/sqlite/sqlite.git synced 2024-11-28 07:51:10 +01:00
sqlite/test/intpkey.test
drh 8210233c7b Revise tests cases to align with the new EXPLAIN QUERY PLAN output.
FossilOrigin-Name: 50fbd532602d2c316813046ed6be8be2991c281eb5f295c4c28520a0de73862c
2021-03-20 15:11:29 +00:00

646 lines
14 KiB
Plaintext

# 2001 September 15
#
# The author disclaims copyright to this source code. In place of
# a legal notice, here is a blessing:
#
# May you do good and not evil.
# May you find forgiveness for yourself and forgive others.
# May you share freely, never taking more than you give.
#
#***********************************************************************
# This file implements regression tests for SQLite library.
#
# This file implements tests for the special processing associated
# with INTEGER PRIMARY KEY columns.
#
# $Id: intpkey.test,v 1.24 2007/11/29 17:43:28 danielk1977 Exp $
set testdir [file dirname $argv0]
source $testdir/tester.tcl
# Create a table with a primary key and a datatype other than
# integer
#
do_test intpkey-1.0 {
execsql {
CREATE TABLE t1(a TEXT PRIMARY KEY, b, c);
}
} {}
# There should be an index associated with the primary key
#
do_test intpkey-1.1 {
execsql {
SELECT name FROM sqlite_master
WHERE type='index' AND tbl_name='t1';
}
} {sqlite_autoindex_t1_1}
# Now create a table with an integer primary key and verify that
# there is no associated index.
#
do_test intpkey-1.2 {
execsql {
DROP TABLE t1;
CREATE TABLE t1(a INTEGER PRIMARY KEY, b, c);
SELECT name FROM sqlite_master
WHERE type='index' AND tbl_name='t1';
}
} {}
# Insert some records into the new table. Specify the primary key
# and verify that the key is used as the record number.
#
do_test intpkey-1.3 {
execsql {
INSERT INTO t1 VALUES(5,'hello','world');
}
db last_insert_rowid
} {5}
do_test intpkey-1.4 {
execsql {
SELECT * FROM t1;
}
} {5 hello world}
do_test intpkey-1.5 {
execsql {
SELECT rowid, * FROM t1;
}
} {5 5 hello world}
# Attempting to insert a duplicate primary key should give a constraint
# failure.
#
do_test intpkey-1.6 {
set r [catch {execsql {
INSERT INTO t1 VALUES(5,'second','entry');
}} msg]
lappend r $msg
} {1 {UNIQUE constraint failed: t1.a}}
do_test intpkey-1.7 {
execsql {
SELECT rowid, * FROM t1;
}
} {5 5 hello world}
do_test intpkey-1.8 {
set r [catch {execsql {
INSERT INTO t1 VALUES(6,'second','entry');
}} msg]
lappend r $msg
} {0 {}}
do_test intpkey-1.8.1 {
db last_insert_rowid
} {6}
do_test intpkey-1.9 {
execsql {
SELECT rowid, * FROM t1;
}
} {5 5 hello world 6 6 second entry}
# A ROWID is automatically generated for new records that do not specify
# the integer primary key.
#
do_test intpkey-1.10 {
execsql {
INSERT INTO t1(b,c) VALUES('one','two');
SELECT b FROM t1 ORDER BY b;
}
} {hello one second}
# Try to change the ROWID for the new entry.
#
do_test intpkey-1.11 {
execsql {
UPDATE t1 SET a=4 WHERE b='one';
SELECT * FROM t1;
}
} {4 one two 5 hello world 6 second entry}
# Make sure SELECT statements are able to use the primary key column
# as an index.
#
do_test intpkey-1.12.1 {
execsql {
SELECT * FROM t1 WHERE a==4;
}
} {4 one two}
do_test intpkey-1.12.2 {
execsql {
EXPLAIN QUERY PLAN
SELECT * FROM t1 WHERE a==4;
}
} {/SEARCH t1 /}
# Try to insert a non-integer value into the primary key field. This
# should result in a data type mismatch.
#
do_test intpkey-1.13.1 {
set r [catch {execsql {
INSERT INTO t1 VALUES('x','y','z');
}} msg]
lappend r $msg
} {1 {datatype mismatch}}
do_test intpkey-1.13.2 {
set r [catch {execsql {
INSERT INTO t1 VALUES('','y','z');
}} msg]
lappend r $msg
} {1 {datatype mismatch}}
do_test intpkey-1.14 {
set r [catch {execsql {
INSERT INTO t1 VALUES(3.4,'y','z');
}} msg]
lappend r $msg
} {1 {datatype mismatch}}
do_test intpkey-1.15 {
set r [catch {execsql {
INSERT INTO t1 VALUES(-3,'y','z');
}} msg]
lappend r $msg
} {0 {}}
do_test intpkey-1.16 {
execsql {SELECT * FROM t1}
} {-3 y z 4 one two 5 hello world 6 second entry}
#### INDICES
# Check to make sure indices work correctly with integer primary keys
#
do_test intpkey-2.1 {
execsql {
CREATE INDEX i1 ON t1(b);
SELECT * FROM t1 WHERE b=='y'
}
} {-3 y z}
do_test intpkey-2.1.1 {
execsql {
SELECT * FROM t1 WHERE b=='y' AND rowid<0
}
} {-3 y z}
do_test intpkey-2.1.2 {
execsql {
SELECT * FROM t1 WHERE b=='y' AND rowid<0 AND rowid>=-20
}
} {-3 y z}
do_test intpkey-2.1.3 {
execsql {
SELECT * FROM t1 WHERE b>='y'
}
} {-3 y z}
do_test intpkey-2.1.4 {
execsql {
SELECT * FROM t1 WHERE b>='y' AND rowid<10
}
} {-3 y z}
do_test intpkey-2.2 {
execsql {
UPDATE t1 SET a=8 WHERE b=='y';
SELECT * FROM t1 WHERE b=='y';
}
} {8 y z}
do_test intpkey-2.3 {
execsql {
SELECT rowid, * FROM t1;
}
} {4 4 one two 5 5 hello world 6 6 second entry 8 8 y z}
do_test intpkey-2.4 {
execsql {
SELECT rowid, * FROM t1 WHERE b<'second'
}
} {5 5 hello world 4 4 one two}
do_test intpkey-2.4.1 {
execsql {
SELECT rowid, * FROM t1 WHERE 'second'>b
}
} {5 5 hello world 4 4 one two}
do_test intpkey-2.4.2 {
execsql {
SELECT rowid, * FROM t1 WHERE 8>rowid AND 'second'>b
}
} {4 4 one two 5 5 hello world}
do_test intpkey-2.4.3 {
execsql {
SELECT rowid, * FROM t1 WHERE 8>rowid AND 'second'>b AND 0<rowid
}
} {4 4 one two 5 5 hello world}
do_test intpkey-2.5 {
execsql {
SELECT rowid, * FROM t1 WHERE b>'a'
}
} {5 5 hello world 4 4 one two 6 6 second entry 8 8 y z}
do_test intpkey-2.6 {
execsql {
DELETE FROM t1 WHERE rowid=4;
SELECT * FROM t1 WHERE b>'a';
}
} {5 hello world 6 second entry 8 y z}
do_test intpkey-2.7 {
execsql {
UPDATE t1 SET a=-4 WHERE rowid=8;
SELECT * FROM t1 WHERE b>'a';
}
} {5 hello world 6 second entry -4 y z}
do_test intpkey-2.7 {
execsql {
SELECT * FROM t1
}
} {-4 y z 5 hello world 6 second entry}
# Do an SQL statement. Append the search count to the end of the result.
#
proc count sql {
set ::sqlite_search_count 0
return [concat [execsql $sql] $::sqlite_search_count]
}
# Create indices that include the integer primary key as one of their
# columns.
#
do_test intpkey-3.1 {
execsql {
CREATE INDEX i2 ON t1(a);
}
} {}
do_test intpkey-3.2 {
count {
SELECT * FROM t1 WHERE a=5;
}
} {5 hello world 0}
do_test intpkey-3.3 {
count {
SELECT * FROM t1 WHERE a>4 AND a<6;
}
} {5 hello world 2}
do_test intpkey-3.4 {
count {
SELECT * FROM t1 WHERE b>='hello' AND b<'hello2';
}
} {5 hello world 3}
do_test intpkey-3.5 {
execsql {
CREATE INDEX i3 ON t1(c,a);
}
} {}
do_test intpkey-3.6 {
count {
SELECT * FROM t1 WHERE c=='world';
}
} {5 hello world 3}
do_test intpkey-3.7 {
execsql {INSERT INTO t1 VALUES(11,'hello','world')}
count {
SELECT * FROM t1 WHERE c=='world';
}
} {5 hello world 11 hello world 5}
do_test intpkey-3.8 {
count {
SELECT * FROM t1 WHERE c=='world' AND a>7;
}
} {11 hello world 3}
do_test intpkey-3.9 {
count {
SELECT * FROM t1 WHERE 7<a;
}
} {11 hello world 1}
# Test inequality constraints on integer primary keys and rowids
#
do_test intpkey-4.1 {
count {
SELECT * FROM t1 WHERE 11=rowid
}
} {11 hello world 0}
do_test intpkey-4.2 {
count {
SELECT * FROM t1 WHERE 11=rowid AND b=='hello'
}
} {11 hello world 0}
do_test intpkey-4.3 {
count {
SELECT * FROM t1 WHERE 11=rowid AND b=='hello' AND c IS NOT NULL;
}
} {11 hello world 0}
do_test intpkey-4.4 {
count {
SELECT * FROM t1 WHERE rowid==11
}
} {11 hello world 0}
do_test intpkey-4.5 {
count {
SELECT * FROM t1 WHERE oid==11 AND b=='hello'
}
} {11 hello world 0}
do_test intpkey-4.6 {
count {
SELECT * FROM t1 WHERE a==11 AND b=='hello' AND c IS NOT NULL;
}
} {11 hello world 0}
do_test intpkey-4.7 {
count {
SELECT * FROM t1 WHERE 8<rowid;
}
} {11 hello world 1}
do_test intpkey-4.8 {
count {
SELECT * FROM t1 WHERE 8<rowid AND 11>=oid;
}
} {11 hello world 1}
do_test intpkey-4.9 {
count {
SELECT * FROM t1 WHERE 11<=_rowid_ AND 12>=a;
}
} {11 hello world 1}
do_test intpkey-4.10 {
count {
SELECT * FROM t1 WHERE 0>=_rowid_;
}
} {-4 y z 1}
do_test intpkey-4.11 {
count {
SELECT * FROM t1 WHERE a<0;
}
} {-4 y z 1}
do_test intpkey-4.12 {
count {
SELECT * FROM t1 WHERE a<0 AND a>10;
}
} {1}
# Make sure it is OK to insert a rowid of 0
#
do_test intpkey-5.1 {
execsql {
INSERT INTO t1 VALUES(0,'zero','entry');
}
count {
SELECT * FROM t1 WHERE a=0;
}
} {0 zero entry 0}
do_test intpkey-5.2 {
execsql {
SELECT rowid, a FROM t1 ORDER BY rowid
}
} {-4 -4 0 0 5 5 6 6 11 11}
# Test the ability of the COPY command to put data into a
# table that contains an integer primary key.
#
# COPY command has been removed. But we retain these tests so
# that the tables will contain the right data for tests that follow.
#
do_test intpkey-6.1 {
execsql {
BEGIN;
INSERT INTO t1 VALUES(20,'b-20','c-20');
INSERT INTO t1 VALUES(21,'b-21','c-21');
INSERT INTO t1 VALUES(22,'b-22','c-22');
COMMIT;
SELECT * FROM t1 WHERE a>=20;
}
} {20 b-20 c-20 21 b-21 c-21 22 b-22 c-22}
do_test intpkey-6.2 {
execsql {
SELECT * FROM t1 WHERE b=='hello'
}
} {5 hello world 11 hello world}
do_test intpkey-6.3 {
execsql {
DELETE FROM t1 WHERE b='b-21';
SELECT * FROM t1 WHERE b=='b-21';
}
} {}
do_test intpkey-6.4 {
execsql {
SELECT * FROM t1 WHERE a>=20
}
} {20 b-20 c-20 22 b-22 c-22}
# Do an insert of values with the columns specified out of order.
#
do_test intpkey-7.1 {
execsql {
INSERT INTO t1(c,b,a) VALUES('row','new',30);
SELECT * FROM t1 WHERE rowid>=30;
}
} {30 new row}
do_test intpkey-7.2 {
execsql {
SELECT * FROM t1 WHERE rowid>20;
}
} {22 b-22 c-22 30 new row}
# Do an insert from a select statement.
#
do_test intpkey-8.1 {
execsql {
CREATE TABLE t2(x INTEGER PRIMARY KEY, y, z);
INSERT INTO t2 SELECT * FROM t1;
SELECT rowid FROM t2;
}
} {-4 0 5 6 11 20 22 30}
do_test intpkey-8.2 {
execsql {
SELECT x FROM t2;
}
} {-4 0 5 6 11 20 22 30}
do_test intpkey-9.1 {
execsql {
UPDATE t1 SET c='www' WHERE c='world';
SELECT rowid, a, c FROM t1 WHERE c=='www';
}
} {5 5 www 11 11 www}
# Check insert of NULL for primary key
#
do_test intpkey-10.1 {
execsql {
DROP TABLE t2;
CREATE TABLE t2(x INTEGER PRIMARY KEY, y, z);
INSERT INTO t2 VALUES(NULL, 1, 2);
SELECT * from t2;
}
} {1 1 2}
do_test intpkey-10.2 {
execsql {
INSERT INTO t2 VALUES(NULL, 2, 3);
SELECT * from t2 WHERE x=2;
}
} {2 2 3}
do_test intpkey-10.3 {
execsql {
INSERT INTO t2 SELECT NULL, z, y FROM t2;
SELECT * FROM t2;
}
} {1 1 2 2 2 3 3 2 1 4 3 2}
# This tests checks to see if a floating point number can be used
# to reference an integer primary key.
#
do_test intpkey-11.1 {
execsql {
SELECT b FROM t1 WHERE a=2.0+3.0;
}
} {hello}
do_test intpkey-11.1 {
execsql {
SELECT b FROM t1 WHERE a=2.0+3.5;
}
} {}
integrity_check intpkey-12.1
# Try to use a string that looks like a floating point number as
# an integer primary key. This should actually work when the floating
# point value can be rounded to an integer without loss of data.
#
do_test intpkey-13.1 {
execsql {
SELECT * FROM t1 WHERE a=1;
}
} {}
do_test intpkey-13.2 {
execsql {
INSERT INTO t1 VALUES('1.0',2,3);
SELECT * FROM t1 WHERE a=1;
}
} {1 2 3}
do_test intpkey-13.3 {
catchsql {
INSERT INTO t1 VALUES('1.5',3,4);
}
} {1 {datatype mismatch}}
ifcapable {bloblit} {
do_test intpkey-13.4 {
catchsql {
INSERT INTO t1 VALUES(x'123456',3,4);
}
} {1 {datatype mismatch}}
}
do_test intpkey-13.5 {
catchsql {
INSERT INTO t1 VALUES('+1234567890',3,4);
}
} {0 {}}
# Compare an INTEGER PRIMARY KEY against a TEXT expression. The INTEGER
# affinity should be applied to the text value before the comparison
# takes place.
#
do_test intpkey-14.1 {
execsql {
CREATE TABLE t3(a INTEGER PRIMARY KEY, b INTEGER, c TEXT);
INSERT INTO t3 VALUES(1, 1, 'one');
INSERT INTO t3 VALUES(2, 2, '2');
INSERT INTO t3 VALUES(3, 3, 3);
}
} {}
do_test intpkey-14.2 {
execsql {
SELECT * FROM t3 WHERE a>2;
}
} {3 3 3}
do_test intpkey-14.3 {
execsql {
SELECT * FROM t3 WHERE a>'2';
}
} {3 3 3}
do_test intpkey-14.4 {
execsql {
SELECT * FROM t3 WHERE a<'2';
}
} {1 1 one}
do_test intpkey-14.5 {
execsql {
SELECT * FROM t3 WHERE a<c;
}
} {1 1 one}
do_test intpkey-14.6 {
execsql {
SELECT * FROM t3 WHERE a=c;
}
} {2 2 2 3 3 3}
# Check for proper handling of primary keys greater than 2^31.
# Ticket #1188
#
do_test intpkey-15.1 {
execsql {
INSERT INTO t1 VALUES(2147483647, 'big-1', 123);
SELECT * FROM t1 WHERE a>2147483648;
}
} {}
do_test intpkey-15.2 {
execsql {
INSERT INTO t1 VALUES(NULL, 'big-2', 234);
SELECT b FROM t1 WHERE a>=2147483648;
}
} {big-2}
do_test intpkey-15.3 {
execsql {
SELECT b FROM t1 WHERE a>2147483648;
}
} {}
do_test intpkey-15.4 {
execsql {
SELECT b FROM t1 WHERE a>=2147483647;
}
} {big-1 big-2}
do_test intpkey-15.5 {
execsql {
SELECT b FROM t1 WHERE a<2147483648;
}
} {y zero 2 hello second hello b-20 b-22 new 3 big-1}
do_test intpkey-15.6 {
execsql {
SELECT b FROM t1 WHERE a<12345678901;
}
} {y zero 2 hello second hello b-20 b-22 new 3 big-1 big-2}
do_test intpkey-15.7 {
execsql {
SELECT b FROM t1 WHERE a>12345678901;
}
} {}
# 2016-04-18 ticket https://www.sqlite.org/src/tktview/7d7525cb01b68712495d3a
# Be sure to escape quoted typenames.
#
do_execsql_test intpkey-16.0 {
CREATE TABLE t16a(id "INTEGER" PRIMARY KEY AUTOINCREMENT, b [TEXT], c `INT`);
} {}
do_execsql_test intpkey-16.1 {
PRAGMA table_info=t16a;
} {0 id INTEGER 0 {} 1 1 b TEXT 0 {} 0 2 c INT 0 {} 0}
# 2016-05-06 ticket https://www.sqlite.org/src/tktview/16c9801ceba4923939085
# When the schema contains an index on the IPK and no other index
# and a WHERE clause on a delete uses an OR where both sides referencing
# the IPK, then it is possible that the OP_Delete will fail because there
# deferred seek of the OP_Seek is not resolved prior to reaching the OP_Delete.
#
do_execsql_test intpkey-17.0 {
CREATE TABLE t17(x INTEGER PRIMARY KEY, y TEXT);
INSERT INTO t17(x,y) VALUES(123,'elephant'),(248,'giraffe');
CREATE INDEX t17x ON t17(x);
DELETE FROM t17 WHERE x=99 OR x<130;
SELECT * FROM t17;
} {248 giraffe}
do_execsql_test intpkey-17.1 {
DROP INDEX t17x;
DELETE FROM t17;
INSERT INTO t17(x,y) VALUES(123,'elephant'),(248,'giraffe');
CREATE UNIQUE INDEX t17x ON t17(abs(x));
DELETE FROM t17 WHERE abs(x) IS NULL OR abs(x)<130;
SELECT * FROM t17;
} {248 giraffe}
do_execsql_test intpkey-17.2 {
DELETE FROM t17;
INSERT INTO t17(x,y) VALUES(123,'elephant'),(248,'giraffe');
UPDATE t17 SET y='ostrich' WHERE abs(x)=248;
SELECT * FROM t17 ORDER BY +x;
} {123 elephant 248 ostrich}
finish_test