SQLite

# features a cross join of some time. Instead of the usual ",", 
# "CROSS JOIN" or "INNER JOIN" join-op, the string %JOIN% must be 
# substituted.
#
# This test runs the SELECT three times - once with:
#
#   * s/%JOIN%/,/
#   * s/%JOIN%/JOIN/
#   * s/%JOIN%/INNER JOIN/
#   * s/%JOIN%/CROSS JOIN/
#
# and checks that each time the results of the SELECT are $res.
#
proc do_join_test {tn select res} {
  foreach {tn2 joinop} [list    1 ,    2 "CROSS JOIN"    3 "INNER JOIN"] {

do_catchsql_test e_select-0.1.5 {
  SELECT count(*) FROM t1, t2 USING (a) ON (t1.a=t2.a)
} {1 {near "ON": syntax error}}

#-------------------------------------------------------------------------
# The following tests focus on FROM clause (join) processing.
#

# EVIDENCE-OF: R-16074-54196 If the FROM clause is omitted from a simple
# SELECT statement, then the input data is implicitly a single row zero
# columns wide
#
do_execsql_test e_select-1.1.1 { SELECT 'abc' }            {abc}
do_execsql_test e_select-1.1.2 { SELECT 'abc' WHERE NULL } {}
do_execsql_test e_select-1.1.3 { SELECT NULL }             {{}}
do_execsql_test e_select-1.1.4 { SELECT count(*) }         {1}
do_execsql_test e_select-1.1.5 { SELECT count(*) WHERE 0 } {0}
do_execsql_test e_select-1.1.6 { SELECT count(*) WHERE 1 } {1}

# EVIDENCE-OF: R-48114-33255 If there is only a single table in the
# join-source following the FROM clause, then the input data used by the
# SELECT statement is the contents of the named table.
#
#   The results of the SELECT queries suggest that they are operating on the
#   contents of the table 'xx'.
#
do_execsql_test e_select-1.2.1 {
  CREATE TABLE xx(x, y);
  INSERT INTO xx VALUES('IiJlsIPepMuAhU', X'10B00B897A15BAA02E3F98DCE8F2');
  INSERT INTO xx VALUES(NULL, -16.87);
  INSERT INTO xx VALUES(-17.89, 'linguistically');
} {}
do_execsql_test e_select-1.2.2 { 
  SELECT quote(x), quote(y) FROM xx
} [list \
  'IiJlsIPepMuAhU' X'10B00B897A15BAA02E3F98DCE8F2' \
  NULL             -16.87                          \
  -17.89           'linguistically'                \
]
do_execsql_test e_select-1.2.3 { 
  SELECT count(*), count(x), count(y) FROM xx
} {3 2 3}
do_execsql_test e_select-1.2.4 { 
  SELECT sum(x), sum(y) FROM xx
} {-17.89 -16.87}

# EVIDENCE-OF: R-23593-12456 If there is more than one table specified
# as part of the join-source following the FROM keyword, then the
# contents of each named table are joined into a single dataset for the
# simple SELECT statement to operate on.
#
#   There are more detailed tests for subsequent requirements that add 
#   more detail to this idea. We just add a single test that shows that
#   data is coming from each of the three tables following the FROM clause
#   here to show that the statement, vague as it is, is not incorrect.
#
do_execsql_test e_select-1.3.1 {
  SELECT * FROM t1, t2, t3
} [list a one a I a 1 a one a I b 2 a one b II a 1 a one b II b 2 a one c III a 1 a one c III b 2 b two a I a 1 b two a I b 2 b two b II a 1 b two b II b 2 b two c III a 1 b two c III b 2 c three a I a 1 c three a I b 2 c three b II a 1 c three b II b 2 c three c III a 1 c three c III b 2]

#
# The following block of tests - e_select-1.4.* - test that the description
# of cartesian joins in the SELECT documentation is consistent with SQLite.
# In doing so, we test the following three requirements as a side-effect:
#
# EVIDENCE-OF: R-46122-14930 If the join-op 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.
#
#    The tests are built on this assertion. Really, they test that the output
#    of a CROSS JOIN, JOIN, INNER JOIN or "," join matches the expected result
#    of calculating 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-07544-24155 The "CROSS JOIN" join operator produces the
# same data as the "INNER JOIN", "JOIN" and "," operators
#
#    All tests are run 4 times, with the only difference in each run being
#    which of the 4 equivalent cartesian product join operators are used.
#    Since the output data is the same in all cases, we consider that this
#    qualifies as testing the two statements above.
#
do_execsql_test e_select-1.4.0 {
  CREATE TABLE x1(a, b);
  CREATE TABLE x2(c, d, e);
  CREATE TABLE x3(f, g, h, i);

  -- x1: 3 rows, 2 columns
  INSERT INTO x1 VALUES(24, 'converging');
  INSERT INTO x1 VALUES(NULL, X'CB71');
  INSERT INTO x1 VALUES('blonds', 'proprietary');

  -- x2: 2 rows, 3 columns
  INSERT INTO x2 VALUES(-60.06, NULL, NULL);
  INSERT INTO x2 VALUES(-58, NULL, 1.21);

  -- x3: 5 rows, 4 columns
  INSERT INTO x3 VALUES(-39.24, NULL, 'encompass', -1);
  INSERT INTO x3 VALUES('presenting', 51, 'reformation', 'dignified');
  INSERT INTO x3 VALUES('conducting', -87.24, 37.56, NULL);
  INSERT INTO x3 VALUES('coldest', -96, 'dramatists', 82.3);
  INSERT INTO x3 VALUES('alerting', NULL, -93.79, NULL);
} {}

# EVIDENCE-OF: R-59089-25828 The columns of the cartesian product
# dataset are, in order, all the columns of the left-hand dataset
# followed by all the columns of the right-hand dataset.
#
do_join_test e_select-1.4.1.1 {
  SELECT * FROM x1 %JOIN% x2 LIMIT 1
} [concat {24 converging} {-60.06 {} {}}]

do_join_test e_select-1.4.1.2 {
  SELECT * FROM x2 %JOIN% x1 LIMIT 1
} [concat {-60.06 {} {}} {24 converging}]

do_join_test e_select-1.4.1.3 {
  SELECT * FROM x3 %JOIN% x2 LIMIT 1
} [concat {-39.24 {} encompass -1} {-60.06 {} {}}]

do_join_test e_select-1.4.1.4 {
  SELECT * FROM x2 %JOIN% x3 LIMIT 1
} [concat {-60.06 {} {}} {-39.24 {} encompass -1}]

# EVIDENCE-OF: R-44414-54710 There is a row in the cartesian product
# dataset formed by combining each unique combination of a row from the
# left-hand and right-hand datasets.
#
do_join_test e_select-1.4.2.1 {
  SELECT * FROM x2 %JOIN% x3
} [list -60.06 {} {}      -39.24 {} encompass -1                 \
        -60.06 {} {}      presenting 51 reformation dignified    \
        -60.06 {} {}      conducting -87.24 37.56 {}             \
        -60.06 {} {}      coldest -96 dramatists 82.3            \
        -60.06 {} {}      alerting {} -93.79 {}                  \
        -58 {} 1.21       -39.24 {} encompass -1                 \
        -58 {} 1.21       presenting 51 reformation dignified    \
        -58 {} 1.21       conducting -87.24 37.56 {}             \
        -58 {} 1.21       coldest -96 dramatists 82.3            \
        -58 {} 1.21       alerting {} -93.79 {}                  \
]
# TODO: Come back and add a few more like the above.

# EVIDENCE-OF: R-20659-43267 In other words, if the left-hand dataset
# consists of Nlhs rows of Mlhs columns, and the right-hand dataset of
# Nrhs rows of Mrhs columns, then the cartesian product is a dataset of
# Nlhs.Nrhs rows, each containing Mlhs+Mrhs columns.
#
# x1, x2    (Nlhs=3, Nrhs=2)   (Mlhs=2, Mrhs=3)
do_join_test e_select-1.4.3.1 { 
  SELECT count(*) FROM x1 %JOIN% x2 
} [expr 3*2]
do_test e_select-1.4.3.2 { 
  expr {[llength [execsql {SELECT * FROM x1, x2}]] / 6}
} [expr 2+3]

# x2, x3    (Nlhs=2, Nrhs=5)   (Mlhs=3, Mrhs=4)
do_join_test e_select-1.4.3.3 { 
  SELECT count(*) FROM x2 %JOIN% x3 
} [expr 2*5]
do_test e_select-1.4.3.4 { 
  expr {[llength [execsql {SELECT * FROM x2 JOIN x3}]] / 10}
} [expr 3+4]

# x3, x1    (Nlhs=5, Nrhs=3)   (Mlhs=4, Mrhs=2)
do_join_test e_select-1.4.3.5 { 
  SELECT count(*) FROM x3 %JOIN% x1 
} [expr 5*3]
do_test e_select-1.4.3.6 { 
  expr {[llength [execsql {SELECT * FROM x3 CROSS JOIN x1}]] / 15}
} [expr 4+2]

# x3, x3    (Nlhs=5, Nrhs=5)   (Mlhs=4, Mrhs=4)
do_join_test e_select-1.4.3.7 { 
  SELECT count(*) FROM x3 %JOIN% x3 
} [expr 5*5]
do_test e_select-1.4.3.8 { 
  expr {[llength [execsql {SELECT * FROM x3 INNER JOIN x3 AS x4}]] / 25}
} [expr 4+4]

# Some extra cartesian product tests using tables t1 and t2.
#
do_execsql_test e_select-1.4.4.1 { SELECT * FROM t1, t2 } $t1_cross_t2
do_execsql_test e_select-1.4.4.2 { SELECT * FROM t1 AS x, t1 AS y} $t1_cross_t1
foreach {tn select res} [list \
    1 { SELECT * FROM t1 CROSS JOIN t2 }           $t1_cross_t2        \
    2 { SELECT * FROM t1 AS y CROSS JOIN t1 AS x } $t1_cross_t1        \
    3 { SELECT * FROM t1 INNER JOIN t2 }           $t1_cross_t2        \
    4 { SELECT * FROM t1 AS y INNER JOIN t1 AS x } $t1_cross_t1        \
] {
  do_execsql_test e_select-1.4.5.$tn $select $res
}


# EVIDENCE-OF: R-45641-53865 If there is an ON clause specified, then
# the ON expression is evaluated for each row of the cartesian product
# and the result cast to a numeric value as if by a CAST expression. All
# rows for which the expression evaluates to NULL or zero (integer value
# 0 or real value 0.0) are excluded from the dataset.





#
foreach {tn select res} [list                                              \
    1 { SELECT * FROM t1 %JOIN% t2 ON (1) }       $t1_cross_t2             \
    2 { SELECT * FROM t1 %JOIN% t2 ON (0) }       [list]                   \
    3 { SELECT * FROM t1 %JOIN% t2 ON (NULL) }    [list]                   \
    4 { SELECT * FROM t1 %JOIN% t2 ON ('abc') }   [list]                   \
    5 { SELECT * FROM t1 %JOIN% t2 ON ('1ab') }   $t1_cross_t2             \

  4b { SELECT * FROM (SELECT a COLLATE nocase, b FROM t6) AS x
       %JOIN% t5 ON (x.a=t5.a) } 
     {aa cc AA cc bb DD BB dd}
} {
  do_join_test e_select-1.7.$tn $select $res
}

# EVIDENCE-OF: R-41434-12448 If the join-op is a "LEFT JOIN" or "LEFT

# OUTER JOIN", then after the ON or USING filtering clauses have been
# applied, an extra row is added to the output for each row in the
# original left-hand input dataset that corresponds to no rows at all in
# the composite dataset (if any).
#
do_execsql_test e_select-1.8.0 {
  CREATE TABLE t7(a, b, c);
  CREATE TABLE t8(a, d, e);

  INSERT INTO t7 VALUES('x', 'ex',  24);
  INSERT INTO t7 VALUES('y', 'why', 25);

  3 {SELECT * FROM t1 NATURAL LEFT JOIN t2 ON (45)}
} {
  do_catchsql_test e_select-1.12.$tn "
    $sql
  " {1 {a NATURAL join may not have an ON or USING clause}}
}

#-------------------------------------------------------------------------
# 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]

      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
}

#----------
# te_equals ?SWITCHES? c1 c2 cols1 row1 cols2 row2
#
proc te_equals {args} {

      }
      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]

  if {$t1 == "NULL" || $t2 == "NULL"} { return 0 }
  if {$nocase && $t1 == "TEXT"} { set v1 [string tolower $v1] }
  if {$nocase && $t2 == "TEXT"} { set v2 [string tolower $v2] }
  return [expr {$t1 == $t2 && $v1 == $v2}]
}

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 { }
  e_select-2.2 { CREATE INDEX i1 ON t1(a) }
  e_select-2.3 { CREATE INDEX i1 ON t2(a) }
  e_select-2.4 { CREATE INDEX i1 ON t3(b) }
} {

  catchsql { DROP INDEX i1 }
  catchsql { DROP INDEX i2 }
  catchsql { DROP INDEX i3 }
  execsql $indexes

  # EVIDENCE-OF: R-46122-14930 If the join-op 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-07544-24155 The "CROSS JOIN" join operator produces the
  # same data 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-45641-53865 If there is an ON clause specified, then
  # the ON expression is evaluated for each row of the cartesian product
  # and the result cast to a numeric value as if by a CAST expression. All
  # rows for which the expression evaluates to NULL or zero (integer value
  # 0 or real value 0.0) are excluded 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 3 "t1 LEFT JOIN t2 ON (t1.a=t2.a)" {t1 t2 -left -on {te_equals a a}}
  test_join 4 "t1 LEFT JOIN t2 USING (a)" {
    t1 t2 -left -using a -on {te_equals a a}
  }
  test_join 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 6 "t1 NATURAL JOIN t2" {
    t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 7 "t1 NATURAL INNER JOIN t2" {
    t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 8 "t1 NATURAL CROSS JOIN t2" {
    t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 9 "t1 NATURAL INNER JOIN t2" {
    t1 t2 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 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 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 12 "t2 NATURAL JOIN t1" {
    t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 13 "t2 NATURAL INNER JOIN t1" {
    t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 14 "t2 NATURAL CROSS JOIN t1" {
    t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 15 "t2 NATURAL INNER JOIN t1" {
    t2 t1 -using {a b} -on {te_and {te_equals a a} {te_equals b b}}
  }
  test_join 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 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 18 "t1 LEFT JOIN t2 USING (b)" {
    t1 t2 -left -using b -on {te_equals b b}
  }
  test_join 19 "t1 JOIN t3 USING(b)" {t1 t3 -using b -on {te_equals b b}}
  test_join 20 "t3 JOIN t1 USING(b)" {
    t3 t1 -using b -on {te_equals -nocase b b}
  }
  test_join 21 "t1 NATURAL JOIN t3"  {
    t1 t3 -using b -on {te_equals b b}
  }
  test_join 22 "t3 NATURAL JOIN t1"  {
    t3 t1 -using b -on {te_equals -nocase b b}
  }
  test_join 23 "t1 NATURAL LEFT JOIN t3" {
    t1 t3 -left -using b -on {te_equals b b}
  }
  test_join 24 "t3 NATURAL LEFT JOIN t1" {
    t3 t1 -left -using b -on {te_equals -nocase b b}
  }
  test_join 25 "t1 LEFT JOIN t3 ON (t3.b=t1.b)" {
    t1 t3 -left -on {te_equals -nocase b b}
  }
  test_join 26 "t1 LEFT JOIN t3 ON (t1.b=t3.b)" {
    t1 t3 -left -on {te_equals b b}
  }
  test_join 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}                         \
  ]
}


# XXXEVIDENCE-OF: R-55824-40976 A sub-select specified in the join-source
# following the FROM clause in a simple SELECT statement is handled as
# if it was a table containing the data returned by executing the
# sub-select statement.
#
proc test_subselect_join {tn subselect select script} {
  1   "SELECT * FROM t2"   "SELECT * FROM t1 JOIN (%ss%)" 
    {t1 %ss%}
} {
  execsql "CREATE TEMP TABLE sstemp AS $subselect"
  set ssdata [te_read_tbl db sstemp]
  execsql "DROP TABLE sstemp"



  
}

finish_test