1
# See the file LICENSE for redistribution information.
3
# Copyright (c) 1996-2002
4
# Sleepycat Software. All rights reserved.
9
# TEST Test duplicate assisted joins. Executes 1, 2, 3 and 4-way joins
10
# TEST with differing index orders and selectivity.
12
# TEST We'll test 2-way, 3-way, and 4-way joins and figure that if those
13
# TEST work, everything else does as well. We'll create test databases
14
# TEST called join1.db, join2.db, join3.db, and join4.db. The number on
15
# TEST the database describes the duplication -- duplicates are of the
16
# TEST form 0, N, 2N, 3N, ... where N is the number of the database.
17
# TEST Primary.db is the primary database, and null.db is the database
18
# TEST that has no matching duplicates.
20
# TEST We should test this on all btrees, all hash, and a combination thereof
21
proc jointest { {psize 8192} {with_dup_dups 0} {flags 0} } {
27
berkdb srand $rand_init
29
# Use one environment for all database opens so we don't
30
# need oodles of regions.
31
set env [berkdb_env -create -home $testdir]
32
error_check_good env_open [is_valid_env $env] TRUE
34
# With the new offpage duplicate code, we don't support
35
# duplicate duplicates in sorted dup sets. Thus, if with_dup_dups
36
# is greater than one, run only with "-dup".
37
if { $with_dup_dups > 1 } {
38
set doptarray {"-dup"}
40
set doptarray {"-dup -dupsort" "-dup" RANDOMMIX RANDOMMIX }
43
# NB: these flags are internal only, ok
44
foreach m "DB_BTREE DB_HASH DB_BOTH" {
45
# run with two different random mixes.
46
foreach dopt $doptarray {
47
set opt [list "-env" $env $dopt]
49
puts "Join test: ($m $dopt) psize $psize,\
53
build_all $m $psize $opt oa $with_dup_dups
55
# null.db is db_built fifth but is referenced by
56
# zero; set up the option array appropriately.
60
puts "\tBuilding the primary database $m"
61
set oflags "-create -truncate -mode 0644 -env $env\
62
[conv $m [berkdb random_int 1 2]]"
63
set db [eval {berkdb_open} $oflags primary.db]
64
error_check_good dbopen [is_valid_db $db] TRUE
65
for { set i 0 } { $i < 1000 } { incr i } {
66
set key [format "%04d" $i]
67
set ret [$db put $key stub]
68
error_check_good "primary put" $ret 0
70
error_check_good "primary close" [$db close] 0
73
do_join primary.db "1 0" $str oa $flags\
76
do_join primary.db "2 0" $str oa $flags\
79
do_join primary.db "3 0" $str oa $flags\
82
do_join primary.db "4 0" $str oa $flags\
85
do_join primary.db "1" $str oa $flags $with_dup_dups
87
do_join primary.db "2" $str oa $flags $with_dup_dups
89
do_join primary.db "3" $str oa $flags $with_dup_dups
91
do_join primary.db "4" $str oa $flags $with_dup_dups
93
do_join primary.db "1 2" $str oa $flags\
96
do_join primary.db "1 2 3" $str oa $flags\
99
do_join primary.db "1 2 3 4" $str oa $flags\
102
do_join primary.db "2 1" $str oa $flags\
105
do_join primary.db "3 2 1" $str oa $flags\
108
do_join primary.db "4 3 2 1" $str oa $flags\
111
do_join primary.db "1 3" $str oa $flags $with_dup_dups
113
do_join primary.db "3 1" $str oa $flags $with_dup_dups
115
do_join primary.db "1 4" $str oa $flags $with_dup_dups
117
do_join primary.db "4 1" $str oa $flags $with_dup_dups
119
do_join primary.db "2 3" $str oa $flags $with_dup_dups
121
do_join primary.db "3 2" $str oa $flags $with_dup_dups
123
do_join primary.db "2 4" $str oa $flags $with_dup_dups
125
do_join primary.db "4 2" $str oa $flags $with_dup_dups
127
do_join primary.db "3 4" $str oa $flags $with_dup_dups
129
do_join primary.db "4 3" $str oa $flags $with_dup_dups
131
do_join primary.db "2 3 4" $str oa $flags\
134
do_join primary.db "3 4 1" $str oa $flags\
137
do_join primary.db "4 2 1" $str oa $flags\
140
do_join primary.db "0 2 1" $str oa $flags\
143
do_join primary.db "3 2 0" $str oa $flags\
146
do_join primary.db "4 3 2 1" $str oa $flags\
149
do_join primary.db "4 3 0 1" $str oa $flags\
152
do_join primary.db "3 3 3" $str oa $flags\
155
do_join primary.db "2 2 3 3" $str oa $flags\
159
do_join primary.db "1 2" $str oa $flags\
160
$with_dup_dups "3" $str2
162
# You really don't want to run this section
163
# with $with_dup_dups > 2.
164
if { $with_dup_dups <= 2 } {
167
do_join primary.db "1 2 3" $str\
168
oa $flags $with_dup_dups "3 3 1" $str2
171
do_join primary.db "4 0 2" $str\
172
oa $flags $with_dup_dups "4 3 3" $str2
175
do_join primary.db "3 2 1" $str\
176
oa $flags $with_dup_dups "0 2" $str2
179
do_join primary.db "2 2 3 3" $str\
180
oa $flags $with_dup_dups "1 4 4" $str2
183
do_join primary.db "2 2 3 3" $str\
184
oa $flags $with_dup_dups "0 0 4 4" $str2
187
do_join primary.db "2 2 3 3" $str2\
188
oa $flags $with_dup_dups "2 4 4" $str
191
do_join primary.db "2 2 3 3" $str2\
192
oa $flags $with_dup_dups "0 0 4 4" $str
198
error_check_good env_close [$env close] 0
201
proc build_all { method psize opt oaname with_dup_dups {nentries 100} } {
203
db_build join1.db $nentries 50 1 [conv $method 1]\
204
$psize $opt $oaname $with_dup_dups
205
db_build join2.db $nentries 25 2 [conv $method 2]\
206
$psize $opt $oaname $with_dup_dups
207
db_build join3.db $nentries 16 3 [conv $method 3]\
208
$psize $opt $oaname $with_dup_dups
209
db_build join4.db $nentries 12 4 [conv $method 4]\
210
$psize $opt $oaname $with_dup_dups
211
db_build null.db $nentries 0 5 [conv $method 5]\
212
$psize $opt $oaname $with_dup_dups
217
DB_HASH { return "-hash"}
218
"-hash" { return "-hash"}
219
DB_BTREE { return "-btree"}
220
"-btree" { return "-btree"}
222
if { [expr $i % 2] == 0 } {
231
proc random_opts { } {
232
set j [berkdb random_int 0 1]
236
return " -dup -dupsort"
240
proc db_build { name nkeys ndups dup_interval method psize lopt oaname \
244
# Get array of arg names (from two levels up the call stack)
247
# Search for "RANDOMMIX" in $opt, and if present, replace
248
# with " -dup" or " -dup -dupsort" at random.
249
set i [lsearch $lopt RANDOMMIX]
251
set lopt [lreplace $lopt $i $i [random_opts]]
254
# Save off db_open arguments for this database.
255
set opt [eval concat $lopt]
256
set oa($dup_interval) $opt
258
# Create the database and open the dictionary
259
set oflags "-create -truncate -mode 0644 $method\
261
set db [eval {berkdb_open} $oflags $opt $name]
262
error_check_good dbopen [is_valid_db $db] TRUE
265
puts -nonewline "\tBuilding $name: $nkeys keys "
266
puts -nonewline "with $ndups duplicates at interval of $dup_interval"
267
if { $with_dup_dups > 0 } {
269
puts "\t\tand $with_dup_dups duplicate duplicates."
273
for { set count 0 } { [gets $did str] != -1 && $count < $nkeys } {
276
# We need to make sure that the dups are inserted in a
277
# random, or near random, order. Do this by generating
278
# them and putting each in a list, then sorting the list
281
for { set i 0 } { $i < $ndups } { incr i } {
282
set data [format "%04d" [expr $i * $dup_interval]]
283
lappend duplist $data
286
for { set i 0 } { $i < $ndups } {incr i } {
287
# set j [berkdb random_int $i [expr $ndups - 1]]
288
set j [expr ($i % 2) + $i]
289
if { $j >= $ndups } { set j $i }
290
set dupi [lindex $duplist $i]
291
set dupj [lindex $duplist $j]
292
set duplist [lreplace $duplist $i $i $dupj]
293
set duplist [lreplace $duplist $j $j $dupi]
295
foreach data $duplist {
296
if { $with_dup_dups != 0 } {
298
{ $j < $with_dup_dups }\
300
set ret [$db put $str $data]
301
error_check_good put$j $ret 0
304
set ret [$db put $str $data]
305
error_check_good put $ret 0
310
set ret [$db put $str NODUP]
311
error_check_good put $ret 0
315
error_check_good close:$name [$db close] 0
318
proc do_join { primary dbs key oanm flags with_dup_dups {dbs2 ""} {key2 ""} } {
324
puts -nonewline "\tJoining: $dbs on $key"
328
puts " with $dbs2 on $key2"
331
# Open all the databases
332
set p [berkdb_open -unknown $testdir/$primary]
333
error_check_good "primary open" [is_valid_db $p] TRUE
338
set ndx [llength $dbs]
340
foreach i [concat $dbs $dbs2] {
342
set db [eval {berkdb_open -unknown} $opt [n_to_name $i]]
343
error_check_good "[n_to_name $i] open" [is_valid_db $db] TRUE
344
set curs [$db cursor]
345
error_check_good "$db cursor" \
346
[is_substr $curs "$db.c"] 1
348
lappend curslist $curs
351
set realkey [concat $key[n_to_name $i]]
353
set realkey [concat $key2[n_to_name $i]]
356
set pair [$curs get -set $realkey]
357
error_check_good cursor_set:$realkey:$pair \
358
[llength [lindex $pair 0]] 2
363
set join_curs [eval {$p join} $curslist]
364
error_check_good join_cursor \
365
[is_substr $join_curs "$p.c"] 1
367
# Calculate how many dups we expect.
368
# We go through the list of indices. If we find a 0, then we
369
# expect 0 dups. For everything else, we look at pairs of numbers,
370
# if the are relatively prime, multiply them and figure out how
371
# many times that goes into 50. If they aren't relatively prime,
372
# take the number of times the larger goes into 50.
375
foreach n [concat $dbs $dbs2] {
384
if { [expr $last % $n] == 0 || [expr $n % $last] == 0 } {
387
set expected [expr 50 / $last]
390
set last [expr $n * $last / [gcd $n $last]]
391
set expected [expr 50 / $last]
395
# If $with_dup_dups is greater than zero, each datum has
396
# been inserted $with_dup_dups times. So we expect the number
397
# of dups to go up by a factor of ($with_dup_dups)^(number of databases)
399
if { $with_dup_dups > 0 } {
400
foreach n [concat $dbs $dbs2] {
401
set expected [expr $expected * $with_dup_dups]
406
if { $flags == " -join_item"} {
412
for { set pair [eval {$join_curs get} $flags] } { \
413
[llength [lindex $pair 0]] == $l } {
414
set pair [eval {$join_curs get} $flags] } {
415
set k [lindex [lindex $pair 0] 0]
417
error_check_bad valid_dup:$i:$dbs $i 0
418
set kval [string trimleft $k 0]
419
if { [string length $kval] == 0 } {
422
error_check_good valid_dup:$i:$dbs [expr $kval % $i] 0
426
error_check_good number_of_dups:$dbs $ndups $expected
428
error_check_good close_primary [$p close] 0
429
foreach i $curslist {
430
error_check_good close_cursor:$i [$i close] 0
433
error_check_good close_index:$i [$i close] 0
437
proc n_to_name { n } {
449
for { set i 2 } { $i <= $a } { incr i } {
450
if { [expr $a % $i] == 0 && [expr $b % $i] == 0 } {