# See the file LICENSE for redistribution information. # # Copyright (c) 1996-2009 Oracle. All rights reserved. # # $Id$ # # TEST test010 # TEST Duplicate test # TEST Small key/data pairs. # TEST # TEST Use the first 10,000 entries from the dictionary. # TEST Insert each with self as key and data; add duplicate records for each. # TEST After all are entered, retrieve all; verify output. # TEST Close file, reopen, do retrieve and re-verify. # TEST This does not work for recno proc test010 { method {nentries 10000} {ndups 5} {tnum "010"} args } { source ./include.tcl set omethod $method set args [convert_args $method $args] set omethod [convert_method $method] # Btree with compression does not support unsorted duplicates. if { [is_compressed $args] == 1 } { puts "Test$tnum skipping for btree with compression." return } if { [is_record_based $method] == 1 || \ [is_rbtree $method] == 1 } { puts "Test$tnum skipping for method $method" return } # Create the database and open the dictionary set txnenv 0 set eindex [lsearch -exact $args "-env"] # # If we are using an env, then testfile should just be the db name. # Otherwise it is the test directory and the name. if { $eindex == -1 } { set testfile $testdir/test$tnum.db set env NULL } else { set testfile test$tnum.db incr eindex set env [lindex $args $eindex] set txnenv [is_txnenv $env] if { $txnenv == 1 } { append args " -auto_commit " # # If we are using txns and running with the # default, set the default down a bit. # if { $nentries == 10000 } { set nentries 100 } reduce_dups nentries ndups } set testdir [get_home $env] } puts "Test$tnum: $method ($args) $nentries \ small $ndups dup key/data pairs" set t1 $testdir/t1 set t2 $testdir/t2 set t3 $testdir/t3 cleanup $testdir $env set db [eval {berkdb_open \ -create -mode 0644 -dup} $args {$omethod $testfile}] error_check_good dbopen [is_valid_db $db] TRUE set did [open $dict] set pflags "" set gflags "" set txn "" set count 0 # Here is the loop where we put and get each key/data pair while { [gets $did str] != -1 && $count < $nentries } { for { set i 1 } { $i <= $ndups } { incr i } { set datastr $i:$str if { $txnenv == 1 } { set t [$env txn] error_check_good txn [is_valid_txn $t $env] TRUE set txn "-txn $t" } set ret [eval {$db put} \ $txn $pflags {$str [chop_data $method $datastr]}] error_check_good put $ret 0 if { $txnenv == 1 } { error_check_good txn [$t commit] 0 } } # Now retrieve all the keys matching this key set x 1 if { $txnenv == 1 } { set t [$env txn] error_check_good txn [is_valid_txn $t $env] TRUE set txn "-txn $t" } set dbc [eval {$db cursor} $txn] for {set ret [$dbc get "-set" $str]} \ {[llength $ret] != 0} \ {set ret [$dbc get "-next"] } { if {[llength $ret] == 0} { break } set k [lindex [lindex $ret 0] 0] if { [string compare $k $str] != 0 } { break } set datastr [lindex [lindex $ret 0] 1] set d [data_of $datastr] error_check_good "Test$tnum:get" $d $str set id [ id_of $datastr ] error_check_good "Test$tnum:dup#" $id $x incr x } error_check_good "Test$tnum:ndups:$str" [expr $x - 1] $ndups error_check_good cursor_close [$dbc close] 0 if { $txnenv == 1 } { error_check_good txn [$t commit] 0 } incr count } close $did # Now we will get each key from the DB and compare the results # to the original. puts "\tTest$tnum.a: Checking file for correct duplicates" set dlist "" for { set i 1 } { $i <= $ndups } {incr i} { lappend dlist $i } if { $txnenv == 1 } { set t [$env txn] error_check_good txn [is_valid_txn $t $env] TRUE set txn "-txn $t" } dup_check $db $txn $t1 $dlist if { $txnenv == 1 } { error_check_good txn [$t commit] 0 } # Now compare the keys to see if they match the dictionary entries set q q filehead $nentries $dict $t3 filesort $t3 $t2 filesort $t1 $t3 error_check_good Test$tnum:diff($t3,$t2) \ [filecmp $t3 $t2] 0 error_check_good db_close [$db close] 0 set db [eval {berkdb_open} $args $testfile] error_check_good dbopen [is_valid_db $db] TRUE puts "\tTest$tnum.b: Checking file for correct duplicates after close" if { $txnenv == 1 } { set t [$env txn] error_check_good txn [is_valid_txn $t $env] TRUE set txn "-txn $t" } dup_check $db $txn $t1 $dlist if { $txnenv == 1 } { error_check_good txn [$t commit] 0 } # Now compare the keys to see if they match the dictionary entries filesort $t1 $t3 error_check_good Test$tnum:diff($t3,$t2) \ [filecmp $t3 $t2] 0 error_check_good db_close [$db close] 0 }