1
0
mirror of https://gitlab.com/cznic/sqlite.git synced 2024-11-24 02:26:14 +00:00
go-sqlite/testdata/tcl/e_select2.test
2021-12-08 11:33:15 +01:00

581 lines
19 KiB
Plaintext

# 2010 September 24
#
# 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 tests to verify that the "testable statements" in
# the lang_select.html document are correct.
#
set testdir [file dirname $argv0]
source $testdir/tester.tcl
#-------------------------------------------------------------------------
# te_* commands:
#
#
# te_read_sql DB SELECT-STATEMENT
# te_read_tbl DB TABLENAME
#
# These two commands are used to read a dataset from the database. A dataset
# consists of N rows of M named columns of values each, where each value has a
# type (null, integer, real, text or blob) and a value within the types domain.
# The tcl format for a "dataset" is a list of two elements:
#
# * A list of the column names.
# * A list of data rows. Each row is itself a list, where each element is
# the contents of a column of the row. Each of these is a list of two
# elements, the type name and the actual value.
#
# For example, the contents of table [t1] as a dataset is:
#
# CREATE TABLE t1(a, b);
# INSERT INTO t1 VALUES('abc', NULL);
# INSERT INTO t1 VALUES(43.1, 22);
#
# {a b} {{{TEXT abc} {NULL {}}} {{REAL 43.1} {INTEGER 22}}}
#
# The [te_read_tbl] command returns a dataset read from a table. The
# [te_read_sql] returns the dataset that results from executing a SELECT
# command.
#
#
# te_tbljoin ?SWITCHES? LHS-TABLE RHS-TABLE
# te_join ?SWITCHES? LHS-DATASET RHS-DATASET
#
# This command joins the two datasets and returns the resulting dataset. If
# there are no switches specified, then the results is the cartesian product
# of the two inputs. The [te_tbljoin] command reads the left and right-hand
# datasets from the specified tables. The [te_join] command is passed the
# datasets directly.
#
# Optional switches are as follows:
#
# -on SCRIPT
# -using COLUMN-LIST
# -left
#
# The -on option specifies a tcl script that is executed for each row in the
# cartesian product of the two datasets. The script has 4 arguments appended
# to it, in the following order:
#
# * The list of column-names from the left-hand dataset.
# * A single row from the left-hand dataset (one "data row" list as
# described above.
# * The list of column-names from the right-hand dataset.
# * A single row from the right-hand dataset.
#
# The script must return a boolean value - true if the combination of rows
# should be included in the output dataset, or false otherwise.
#
# The -using option specifies a list of the columns from the right-hand
# dataset that should be omitted from the output dataset.
#
# If the -left option is present, the join is done LEFT JOIN style.
# Specifically, an extra row is inserted if after the -on script is run there
# exist rows in the left-hand dataset that have no corresponding rows in
# the output. See the implementation for more specific comments.
#
#
# te_equals ?SWITCHES? COLNAME1 COLNAME2 <-on script args>
#
# The only supported switch is "-nocase". If it is present, then text values
# are compared in a case-independent fashion. Otherwise, they are compared
# as if using the SQLite BINARY collation sequence.
#
#
# te_and ONSCRIPT1 ONSCRIPT2...
#
#
#
# te_read_tbl DB TABLENAME
# te_read_sql DB SELECT-STATEMENT
#
# These two procs are used to extract datasets from the database, either
# by reading the contents of a named table (te_read_tbl), or by executing
# a SELECT statement (t3_read_sql).
#
# See the comment above, describing "te_* commands", for details of the
# return values.
#
proc te_read_tbl {db tbl} {
te_read_sql $db "SELECT * FROM '$tbl'"
}
proc te_read_sql {db sql} {
set S [sqlite3_prepare_v2 $db $sql -1 DUMMY]
set cols [list]
for {set i 0} {$i < [sqlite3_column_count $S]} {incr i} {
lappend cols [sqlite3_column_name $S $i]
}
set rows [list]
while {[sqlite3_step $S] == "SQLITE_ROW"} {
set r [list]
for {set i 0} {$i < [sqlite3_column_count $S]} {incr i} {
lappend r [list [sqlite3_column_type $S $i] [sqlite3_column_text $S $i]]
}
lappend rows $r
}
sqlite3_finalize $S
return [list $cols $rows]
}
#-------
# Usage: te_join <table-data1> <table-data2> <join spec>...
#
# Where a join-spec is an optional list of arguments as follows:
#
# ?-left?
# ?-using colname-list?
# ?-on on-expr-proc?
#
proc te_join {data1 data2 args} {
set testproc ""
set usinglist [list]
set isleft 0
for {set i 0} {$i < [llength $args]} {incr i} {
set a [lindex $args $i]
switch -- $a {
-on { set testproc [lindex $args [incr i]] }
-using { set usinglist [lindex $args [incr i]] }
-left { set isleft 1 }
default {
error "Unknown argument: $a"
}
}
}
set c1 [lindex $data1 0]
set c2 [lindex $data2 0]
set omitlist [list]
set nullrowlist [list]
set cret $c1
set cidx 0
foreach col $c2 {
set idx [lsearch $usinglist $col]
if {$idx>=0} {lappend omitlist $cidx}
if {$idx<0} {
lappend nullrowlist {NULL {}}
lappend cret $col
}
incr cidx
}
set omitlist [lsort -integer -decreasing $omitlist]
set rret [list]
foreach r1 [lindex $data1 1] {
set one 0
foreach r2 [lindex $data2 1] {
set ok 1
if {$testproc != ""} {
set ok [eval $testproc [list $c1 $r1 $c2 $r2]]
}
if {$ok} {
set one 1
foreach idx $omitlist {set r2 [lreplace $r2 $idx $idx]}
lappend rret [concat $r1 $r2]
}
}
if {$isleft && $one==0} {
lappend rret [concat $r1 $nullrowlist]
}
}
list $cret $rret
}
proc te_tbljoin {db t1 t2 args} {
te_join [te_read_tbl $db $t1] [te_read_tbl $db $t2] {*}$args
}
proc te_apply_affinity {affinity typevar valvar} {
upvar $typevar type
upvar $valvar val
switch -- $affinity {
integer {
if {[string is double $val]} { set type REAL }
if {[string is wideinteger $val]} { set type INTEGER }
if {$type == "REAL" && int($val)==$val} {
set type INTEGER
set val [expr {int($val)}]
}
}
text {
set type TEXT
}
none { }
default { error "invalid affinity: $affinity" }
}
}
#----------
# te_equals ?SWITCHES? c1 c2 cols1 row1 cols2 row2
#
proc te_equals {args} {
if {[llength $args]<6} {error "invalid arguments to te_equals"}
foreach {c1 c2 cols1 row1 cols2 row2} [lrange $args end-5 end] break
set nocase 0
set affinity none
for {set i 0} {$i < ([llength $args]-6)} {incr i} {
set a [lindex $args $i]
switch -- $a {
-nocase {
set nocase 1
}
-affinity {
set affinity [string tolower [lindex $args [incr i]]]
}
default {
error "invalid arguments to te_equals"
}
}
}
set idx2 [if {[string is integer $c2]} { set c2 } else { lsearch $cols2 $c2 }]
set idx1 [if {[string is integer $c1]} { set c1 } else { lsearch $cols1 $c1 }]
set t1 [lindex $row1 $idx1 0]
set t2 [lindex $row2 $idx2 0]
set v1 [lindex $row1 $idx1 1]
set v2 [lindex $row2 $idx2 1]
te_apply_affinity $affinity t1 v1
te_apply_affinity $affinity t2 v2
if {$t1 == "NULL" || $t2 == "NULL"} { return 0 }
if {$nocase && $t1 == "TEXT"} { set v1 [string tolower $v1] }
if {$nocase && $t2 == "TEXT"} { set v2 [string tolower $v2] }
set res [expr {$t1 == $t2 && [string equal $v1 $v2]}]
return $res
}
proc te_false {args} { return 0 }
proc te_true {args} { return 1 }
proc te_and {args} {
foreach a [lrange $args 0 end-4] {
set res [eval $a [lrange $args end-3 end]]
if {$res == 0} {return 0}
}
return 1
}
proc te_dataset_eq {testname got expected} {
uplevel #0 [list do_test $testname [list set {} $got] $expected]
}
proc te_dataset_eq_unordered {testname got expected} {
lset got 1 [lsort [lindex $got 1]]
lset expected 1 [lsort [lindex $expected 1]]
te_dataset_eq $testname $got $expected
}
proc te_dataset_ne {testname got unexpected} {
uplevel #0 [list do_test $testname [list string equal $got $unexpected] 0]
}
proc te_dataset_ne_unordered {testname got unexpected} {
lset got 1 [lsort [lindex $got 1]]
lset unexpected 1 [lsort [lindex $unexpected 1]]
te_dataset_ne $testname $got $unexpected
}
#-------------------------------------------------------------------------
#
proc test_join {tn sqljoin tbljoinargs} {
set sql [te_read_sql db "SELECT * FROM $sqljoin"]
set te [te_tbljoin db {*}$tbljoinargs]
te_dataset_eq_unordered $tn $sql $te
}
drop_all_tables
do_execsql_test e_select-2.0 {
CREATE TABLE t1(a, b);
CREATE TABLE t2(a, b);
CREATE TABLE t3(b COLLATE nocase);
INSERT INTO t1 VALUES(2, 'B');
INSERT INTO t1 VALUES(1, 'A');
INSERT INTO t1 VALUES(4, 'D');
INSERT INTO t1 VALUES(NULL, NULL);
INSERT INTO t1 VALUES(3, NULL);
INSERT INTO t2 VALUES(1, 'A');
INSERT INTO t2 VALUES(2, NULL);
INSERT INTO t2 VALUES(5, 'E');
INSERT INTO t2 VALUES(NULL, NULL);
INSERT INTO t2 VALUES(3, 'C');
INSERT INTO t3 VALUES('a');
INSERT INTO t3 VALUES('c');
INSERT INTO t3 VALUES('b');
} {}
foreach {tn indexes} {
e_select-2.1.1 { }
e_select-2.1.2 { CREATE INDEX i1 ON t1(a) }
e_select-2.1.3 { CREATE INDEX i1 ON t2(a) }
e_select-2.1.4 { CREATE INDEX i1 ON t3(b) }
} {
catchsql { DROP INDEX i1 }
catchsql { DROP INDEX i2 }
catchsql { DROP INDEX i3 }
execsql $indexes
# EVIDENCE-OF: R-49872-03192 If the join-operator is "CROSS JOIN",
# "INNER JOIN", "JOIN" or a comma (",") and there is no ON or USING
# clause, then the result of the join is simply the cartesian product of
# the left and right-hand datasets.
#
# EVIDENCE-OF: R-46256-57243 There is no difference between the "INNER
# JOIN", "JOIN" and "," join operators.
#
# EVIDENCE-OF: R-25071-21202 The "CROSS JOIN" join operator produces the
# same result as the "INNER JOIN", "JOIN" and "," operators
#
test_join $tn.1.1 "t1, t2" {t1 t2}
test_join $tn.1.2 "t1 INNER JOIN t2" {t1 t2}
test_join $tn.1.3 "t1 CROSS JOIN t2" {t1 t2}
test_join $tn.1.4 "t1 JOIN t2" {t1 t2}
test_join $tn.1.5 "t2, t3" {t2 t3}
test_join $tn.1.6 "t2 INNER JOIN t3" {t2 t3}
test_join $tn.1.7 "t2 CROSS JOIN t3" {t2 t3}
test_join $tn.1.8 "t2 JOIN t3" {t2 t3}
test_join $tn.1.9 "t2, t2 AS x" {t2 t2}
test_join $tn.1.10 "t2 INNER JOIN t2 AS x" {t2 t2}
test_join $tn.1.11 "t2 CROSS JOIN t2 AS x" {t2 t2}
test_join $tn.1.12 "t2 JOIN t2 AS x" {t2 t2}
# EVIDENCE-OF: R-38465-03616 If there is an ON clause then the ON
# expression is evaluated for each row of the cartesian product as a
# boolean expression. Only rows for which the expression evaluates to
# true are included from the dataset.
#
test_join $tn.2.1 "t1, t2 ON (t1.a=t2.a)" {t1 t2 -on {te_equals a a}}
test_join $tn.2.2 "t2, t1 ON (t1.a=t2.a)" {t2 t1 -on {te_equals a a}}
test_join $tn.2.3 "t2, t1 ON (1)" {t2 t1 -on te_true}
test_join $tn.2.4 "t2, t1 ON (NULL)" {t2 t1 -on te_false}
test_join $tn.2.5 "t2, t1 ON (1.1-1.1)" {t2 t1 -on te_false}
test_join $tn.2.6 "t1, t2 ON (1.1-1.0)" {t1 t2 -on te_true}
test_join $tn.3 "t1 LEFT JOIN t2 ON (t1.a=t2.a)" {t1 t2 -left -on {te_equals a a}}
test_join $tn.4 "t1 LEFT JOIN t2 USING (a)" {
t1 t2 -left -using a -on {te_equals a a}
}
test_join $tn.5 "t1 CROSS JOIN t2 USING(b, a)" {
t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.6 "t1 NATURAL JOIN t2" {
t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.7 "t1 NATURAL INNER JOIN t2" {
t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.8 "t1 NATURAL CROSS JOIN t2" {
t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.9 "t1 NATURAL INNER JOIN t2" {
t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.10 "t1 NATURAL LEFT JOIN t2" {
t1 t2 -left -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.11 "t1 NATURAL LEFT OUTER JOIN t2" {
t1 t2 -left -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.12 "t2 NATURAL JOIN t1" {
t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.13 "t2 NATURAL INNER JOIN t1" {
t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.14 "t2 NATURAL CROSS JOIN t1" {
t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.15 "t2 NATURAL INNER JOIN t1" {
t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.16 "t2 NATURAL LEFT JOIN t1" {
t2 t1 -left -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.17 "t2 NATURAL LEFT OUTER JOIN t1" {
t2 t1 -left -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
}
test_join $tn.18 "t1 LEFT JOIN t2 USING (b)" {
t1 t2 -left -using b -on {te_equals b b}
}
test_join $tn.19 "t1 JOIN t3 USING(b)" {t1 t3 -using b -on {te_equals b b}}
test_join $tn.20 "t3 JOIN t1 USING(b)" {
t3 t1 -using b -on {te_equals -nocase b b}
}
test_join $tn.21 "t1 NATURAL JOIN t3" {
t1 t3 -using b -on {te_equals b b}
}
test_join $tn.22 "t3 NATURAL JOIN t1" {
t3 t1 -using b -on {te_equals -nocase b b}
}
test_join $tn.23 "t1 NATURAL LEFT JOIN t3" {
t1 t3 -left -using b -on {te_equals b b}
}
test_join $tn.24 "t3 NATURAL LEFT JOIN t1" {
t3 t1 -left -using b -on {te_equals -nocase b b}
}
test_join $tn.25 "t1 LEFT JOIN t3 ON (t3.b=t1.b)" {
t1 t3 -left -on {te_equals -nocase b b}
}
test_join $tn.26 "t1 LEFT JOIN t3 ON (t1.b=t3.b)" {
t1 t3 -left -on {te_equals b b}
}
test_join $tn.27 "t1 JOIN t3 ON (t1.b=t3.b)" { t1 t3 -on {te_equals b b} }
# EVIDENCE-OF: R-28760-53843 When more than two tables are joined
# together as part of a FROM clause, the join operations are processed
# in order from left to right. In other words, the FROM clause (A
# join-op-1 B join-op-2 C) is computed as ((A join-op-1 B) join-op-2 C).
#
# Tests 28a and 28b show that the statement above is true for this case.
# Test 28c shows that if the parenthesis force a different order of
# evaluation the result is different. Test 28d verifies that the result
# of the query with the parenthesis forcing a different order of evaluation
# is as calculated by the [te_*] procs.
#
set t3_natural_left_join_t2 [
te_tbljoin db t3 t2 -left -using {b} -on {te_equals -nocase b b}
]
set t1 [te_read_tbl db t1]
te_dataset_eq_unordered $tn.28a [
te_read_sql db "SELECT * FROM t3 NATURAL LEFT JOIN t2 NATURAL JOIN t1"
] [te_join $t3_natural_left_join_t2 $t1 \
-using {a b} -on {te_and {te_equals a a} {te_equals -nocase b b}} \
]
te_dataset_eq_unordered $tn.28b [
te_read_sql db "SELECT * FROM (t3 NATURAL LEFT JOIN t2) NATURAL JOIN t1"
] [te_join $t3_natural_left_join_t2 $t1 \
-using {a b} -on {te_and {te_equals a a} {te_equals -nocase b b}} \
]
te_dataset_ne_unordered $tn.28c [
te_read_sql db "SELECT * FROM (t3 NATURAL LEFT JOIN t2) NATURAL JOIN t1"
] [
te_read_sql db "SELECT * FROM t3 NATURAL LEFT JOIN (t2 NATURAL JOIN t1)"
]
set t2_natural_join_t1 [te_tbljoin db t2 t1 -using {a b} \
-using {a b} -on {te_and {te_equals a a} {te_equals -nocase b b}} \
]
set t3 [te_read_tbl db t3]
te_dataset_eq_unordered $tn.28d [
te_read_sql db "SELECT * FROM t3 NATURAL LEFT JOIN (t2 NATURAL JOIN t1)"
] [te_join $t3 $t2_natural_join_t1 \
-left -using {b} -on {te_equals -nocase b b} \
]
}
do_execsql_test e_select-2.2.0 {
CREATE TABLE t4(x TEXT COLLATE nocase);
CREATE TABLE t5(y INTEGER, z TEXT COLLATE binary);
INSERT INTO t4 VALUES('2.0');
INSERT INTO t4 VALUES('TWO');
INSERT INTO t5 VALUES(2, 'two');
} {}
# EVIDENCE-OF: R-59237-46742 A subquery specified in the
# table-or-subquery following the FROM clause in a simple SELECT
# statement is handled as if it was a table containing the data returned
# by executing the subquery statement.
#
# EVIDENCE-OF: R-27438-53558 Each column of the subquery has the
# collation sequence and affinity of the corresponding expression in the
# subquery statement.
#
foreach {tn subselect select spec} {
1 "SELECT * FROM t2" "SELECT * FROM t1 JOIN %ss%"
{t1 %ss%}
2 "SELECT * FROM t2" "SELECT * FROM t1 JOIN %ss% AS x ON (t1.a=x.a)"
{t1 %ss% -on {te_equals 0 0}}
3 "SELECT * FROM t2" "SELECT * FROM %ss% AS x JOIN t1 ON (t1.a=x.a)"
{%ss% t1 -on {te_equals 0 0}}
4 "SELECT * FROM t1, t2" "SELECT * FROM %ss% AS x JOIN t3"
{%ss% t3}
5 "SELECT * FROM t1, t2" "SELECT * FROM %ss% NATURAL JOIN t3"
{%ss% t3 -using b -on {te_equals 1 0}}
6 "SELECT * FROM t1, t2" "SELECT * FROM t3 NATURAL JOIN %ss%"
{t3 %ss% -using b -on {te_equals -nocase 0 1}}
7 "SELECT * FROM t1, t2" "SELECT * FROM t3 NATURAL LEFT JOIN %ss%"
{t3 %ss% -left -using b -on {te_equals -nocase 0 1}}
8 "SELECT count(*) AS y FROM t4" "SELECT * FROM t5, %ss% USING (y)"
{t5 %ss% -using y -on {te_equals -affinity text 0 0}}
9 "SELECT count(*) AS y FROM t4" "SELECT * FROM %ss%, t5 USING (y)"
{%ss% t5 -using y -on {te_equals -affinity text 0 0}}
10 "SELECT x AS y FROM t4" "SELECT * FROM %ss% JOIN t5 USING (y)"
{%ss% t5 -using y -on {te_equals -nocase -affinity integer 0 0}}
11 "SELECT x AS y FROM t4" "SELECT * FROM t5 JOIN %ss% USING (y)"
{t5 %ss% -using y -on {te_equals -nocase -affinity integer 0 0}}
12 "SELECT y AS x FROM t5" "SELECT * FROM %ss% JOIN t4 USING (x)"
{%ss% t4 -using x -on {te_equals -nocase -affinity integer 0 0}}
13 "SELECT y AS x FROM t5" "SELECT * FROM t4 JOIN %ss% USING (x)"
{t4 %ss% -using x -on {te_equals -nocase -affinity integer 0 0}}
14 "SELECT +y AS x FROM t5" "SELECT * FROM %ss% JOIN t4 USING (x)"
{%ss% t4 -using x -on {te_equals -nocase -affinity text 0 0}}
15 "SELECT +y AS x FROM t5" "SELECT * FROM t4 JOIN %ss% USING (x)"
{t4 %ss% -using x -on {te_equals -nocase -affinity text 0 0}}
} {
# Create a temporary table named %ss% containing the data returned by
# the sub-select. Then have the [te_tbljoin] proc use this table to
# compute the expected results of the $select query. Drop the temporary
# table before continuing.
#
execsql "CREATE TEMP TABLE '%ss%' AS $subselect"
set te [eval te_tbljoin db $spec]
execsql "DROP TABLE '%ss%'"
# Check that the actual data returned by the $select query is the same
# as the expected data calculated using [te_tbljoin] above.
#
te_dataset_eq_unordered e_select-2.2.1.$tn [
te_read_sql db [string map [list %ss% "($subselect)"] $select]
] $te
}
finish_test