• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# 2009 December 03
2#
3#    May you do good and not evil.
4#    May you find forgiveness for yourself and forgive others.
5#    May you share freely, never taking more than you give.
6#
7#***********************************************************************
8#
9# Brute force (random data) tests for FTS3.
10#
11
12#-------------------------------------------------------------------------
13#
14# The FTS3 tests implemented in this file focus on testing that FTS3
15# returns the correct set of documents for various types of full-text
16# query. This is done using pseudo-randomly generated data and queries.
17# The expected result of each query is calculated using Tcl code.
18#
19#   1. The database is initialized to contain a single table with three
20#      columns. 100 rows are inserted into the table. Each of the three
21#      values in each row is a document consisting of between 0 and 100
22#      terms. Terms are selected from a vocabulary of $G(nVocab) terms.
23#
24#   2. The following is performed 100 times:
25#
26#      a. A row is inserted into the database. The row contents are
27#         generated as in step 1. The docid is a pseudo-randomly selected
28#         value between 0 and 1000000.
29#
30#      b. A psuedo-randomly selected row is updated. One of its columns is
31#         set to contain a new document generated in the same way as the
32#         documents in step 1.
33#
34#      c. A psuedo-randomly selected row is deleted.
35#
36#      d. For each of several types of fts3 queries, 10 SELECT queries
37#         of the form:
38#
39#           SELECT docid FROM <tbl> WHERE <tbl> MATCH '<query>'
40#
41#         are evaluated. The results are compared to those calculated by
42#         Tcl code in this file. The patterns used for the different query
43#         types are:
44#
45#           1.  query = <term>
46#           2.  query = <prefix>
47#           3.  query = "<term> <term>"
48#           4.  query = "<term> <term> <term>"
49#           5.  query = "<prefix> <prefix> <prefix>"
50#           6.  query = <term> NEAR <term>
51#           7.  query = <term> NEAR/11 <term> NEAR/11 <term>
52#           8.  query = <term> OR <term>
53#           9.  query = <term> NOT <term>
54#           10. query = <term> AND <term>
55#           11. query = <term> NEAR <term> OR <term> NEAR <term>
56#           12. query = <term> NEAR <term> NOT <term> NEAR <term>
57#           13. query = <term> NEAR <term> AND <term> NEAR <term>
58#
59#         where <term> is a term psuedo-randomly selected from the vocabulary
60#         and prefix is the first 2 characters of such a term followed by
61#         a "*" character.
62#
63#      Every second iteration, steps (a) through (d) above are performed
64#      within a single transaction. This forces the queries in (d) to
65#      read data from both the database and the in-memory hash table
66#      that caches the full-text index entries created by steps (a), (b)
67#      and (c) until the transaction is committed.
68#
69# The procedure above is run 5 times, using advisory fts3 node sizes of 50,
70# 500, 1000 and 2000 bytes.
71#
72# After the test using an advisory node-size of 50, an OOM test is run using
73# the database. This test is similar to step (d) above, except that it tests
74# the effects of transient and persistent OOM conditions encountered while
75# executing each query.
76#
77
78set testdir [file dirname $argv0]
79source $testdir/tester.tcl
80
81# If this build does not include FTS3, skip the tests in this file.
82#
83ifcapable !fts3 { finish_test ; return }
84source $testdir/fts3_common.tcl
85source $testdir/malloc_common.tcl
86
87set G(nVocab) 100
88
89set nVocab 100
90set lVocab [list]
91
92expr srand(0)
93
94# Generate a vocabulary of nVocab words. Each word is 3 characters long.
95#
96set lChar {a b c d e f g h i j k l m n o p q r s t u v w x y z}
97for {set i 0} {$i < $nVocab} {incr i} {
98  set len [expr int(rand()*3)+2]
99  set    word [lindex $lChar [expr int(rand()*26)]]
100  append word [lindex $lChar [expr int(rand()*26)]]
101  if {$len>2} { append word [lindex $lChar [expr int(rand()*26)]] }
102  if {$len>3} { append word [lindex $lChar [expr int(rand()*26)]] }
103  lappend lVocab $word
104}
105
106proc random_term {} {
107  lindex $::lVocab [expr {int(rand()*$::nVocab)}]
108}
109
110# Return a document consisting of $nWord arbitrarily selected terms
111# from the $::lVocab list.
112#
113proc generate_doc {nWord} {
114  set doc [list]
115  for {set i 0} {$i < $nWord} {incr i} {
116    lappend doc [random_term]
117  }
118  return $doc
119}
120
121
122
123# Primitives to update the table.
124#
125unset -nocomplain t1
126proc insert_row {rowid} {
127  set a [generate_doc [expr int((rand()*100))]]
128  set b [generate_doc [expr int((rand()*100))]]
129  set c [generate_doc [expr int((rand()*100))]]
130  execsql { INSERT INTO t1(docid, a, b, c) VALUES($rowid, $a, $b, $c) }
131  set ::t1($rowid) [list $a $b $c]
132}
133proc delete_row {rowid} {
134  execsql { DELETE FROM t1 WHERE rowid = $rowid }
135  catch {unset ::t1($rowid)}
136}
137proc update_row {rowid} {
138  set cols {a b c}
139  set iCol [expr int(rand()*3)]
140  set doc  [generate_doc [expr int((rand()*100))]]
141  lset ::t1($rowid) $iCol $doc
142  execsql "UPDATE t1 SET [lindex $cols $iCol] = \$doc WHERE rowid = \$rowid"
143}
144
145proc simple_phrase {zPrefix} {
146  set ret [list]
147
148  set reg [string map {* {[^ ]*}} $zPrefix]
149  set reg " $reg "
150
151  foreach key [lsort -integer [array names ::t1]] {
152    set value $::t1($key)
153    set cnt [list]
154    foreach col $value {
155      if {[regexp $reg " $col "]} { lappend ret $key ; break }
156    }
157  }
158
159  #lsort -uniq -integer $ret
160  set ret
161}
162
163# This [proc] is used to test the FTS3 matchinfo() function.
164#
165proc simple_token_matchinfo {zToken} {
166
167  set nDoc(0) 0
168  set nDoc(1) 0
169  set nDoc(2) 0
170  set nHit(0) 0
171  set nHit(1) 0
172  set nHit(2) 0
173
174
175  foreach key [array names ::t1] {
176    set value $::t1($key)
177    set a($key) [list]
178    foreach i {0 1 2} col $value {
179      set hit [llength [lsearch -all $col $zToken]]
180      lappend a($key) $hit
181      incr nHit($i) $hit
182      if {$hit>0} { incr nDoc($i) }
183    }
184  }
185
186  set ret [list]
187  foreach docid [lsort -integer [array names a]] {
188    if { [lindex [lsort -integer $a($docid)] end] } {
189      set matchinfo [list 1 3]
190      foreach i {0 1 2} hit $a($docid) {
191        lappend matchinfo $hit $nHit($i) $nDoc($i)
192      }
193      lappend ret $docid $matchinfo
194    }
195  }
196
197  set ret
198}
199
200proc simple_near {termlist nNear} {
201  set ret [list]
202
203  foreach {key value} [array get ::t1] {
204    foreach v $value {
205
206      set l [lsearch -exact -all $v [lindex $termlist 0]]
207      foreach T [lrange $termlist 1 end] {
208        set l2 [list]
209        foreach i $l {
210          set iStart [expr $i - $nNear - 1]
211          set iEnd [expr $i + $nNear + 1]
212          if {$iStart < 0} {set iStart 0}
213          foreach i2 [lsearch -exact -all [lrange $v $iStart $iEnd] $T] {
214            incr i2 $iStart
215            if {$i2 != $i} { lappend l2 $i2 }
216          }
217        }
218        set l [lsort -uniq -integer $l2]
219      }
220
221      if {[llength $l]} {
222#puts "MATCH($key): $v"
223        lappend ret $key
224      }
225    }
226  }
227
228  lsort -unique -integer $ret
229}
230
231# The following three procs:
232#
233#   setup_not A B
234#   setup_or  A B
235#   setup_and A B
236#
237# each take two arguments. Both arguments must be lists of integer values
238# sorted by value. The return value is the list produced by evaluating
239# the equivalent of "A op B", where op is the FTS3 operator NOT, OR or
240# AND.
241#
242proc setop_not {A B} {
243  foreach b $B { set n($b) {} }
244  set ret [list]
245  foreach a $A { if {![info exists n($a)]} {lappend ret $a} }
246  return $ret
247}
248proc setop_or {A B} {
249  lsort -integer -uniq [concat $A $B]
250}
251proc setop_and {A B} {
252  foreach b $B { set n($b) {} }
253  set ret [list]
254  foreach a $A { if {[info exists n($a)]} {lappend ret $a} }
255  return $ret
256}
257
258proc mit {blob} {
259  set scan(littleEndian) i*
260  set scan(bigEndian) I*
261  binary scan $blob $scan($::tcl_platform(byteOrder)) r
262  return $r
263}
264db func mit mit
265
266set sqlite_fts3_enable_parentheses 1
267
268foreach nodesize {50 500 1000 2000} {
269  catch { array unset ::t1 }
270
271  # Create the FTS3 table. Populate it (and the Tcl array) with 100 rows.
272  #
273  db transaction {
274    catchsql { DROP TABLE t1 }
275    execsql "CREATE VIRTUAL TABLE t1 USING fts3(a, b, c)"
276    execsql "INSERT INTO t1(t1) VALUES('nodesize=$nodesize')"
277    for {set i 0} {$i < 100} {incr i} { insert_row $i }
278  }
279
280  for {set iTest 1} {$iTest <= 100} {incr iTest} {
281    catchsql COMMIT
282
283    set DO_MALLOC_TEST 0
284    set nRep 10
285    if {$iTest==100 && $nodesize==50} {
286      set DO_MALLOC_TEST 1
287      set nRep 2
288    }
289
290    # Delete one row, update one row and insert one row.
291    #
292    set rows [array names ::t1]
293    set nRow [llength $rows]
294    set iUpdate [lindex $rows [expr {int(rand()*$nRow)}]]
295    set iDelete $iUpdate
296    while {$iDelete == $iUpdate} {
297      set iDelete [lindex $rows [expr {int(rand()*$nRow)}]]
298    }
299    set iInsert $iUpdate
300    while {[info exists ::t1($iInsert)]} {
301      set iInsert [expr {int(rand()*1000000)}]
302    }
303    execsql BEGIN
304      insert_row $iInsert
305      update_row $iUpdate
306      delete_row $iDelete
307    if {0==($iTest%2)} { execsql COMMIT }
308
309    if {0==($iTest%2)} {
310      do_test fts3rnd-1.$nodesize.$iTest.0 { fts3_integrity_check t1 } ok
311    }
312
313    # Pick 10 terms from the vocabulary. Check that the results of querying
314    # the database for the set of documents containing each of these terms
315    # is the same as the result obtained by scanning the contents of the Tcl
316    # array for each term.
317    #
318    for {set i 0} {$i < 10} {incr i} {
319      set term [random_term]
320      do_select_test fts3rnd-1.$nodesize.$iTest.1.$i {
321        SELECT docid, mit(matchinfo(t1)) FROM t1 WHERE t1 MATCH $term
322      } [simple_token_matchinfo $term]
323    }
324
325    # This time, use the first two characters of each term as a term prefix
326    # to query for. Test that querying the Tcl array produces the same results
327    # as querying the FTS3 table for the prefix.
328    #
329    for {set i 0} {$i < $nRep} {incr i} {
330      set prefix [string range [random_term] 0 end-1]
331      set match "${prefix}*"
332      do_select_test fts3rnd-1.$nodesize.$iTest.2.$i {
333        SELECT docid FROM t1 WHERE t1 MATCH $match
334      } [simple_phrase $match]
335    }
336
337    # Similar to the above, except for phrase queries.
338    #
339    for {set i 0} {$i < $nRep} {incr i} {
340      set term [list [random_term] [random_term]]
341      set match "\"$term\""
342      do_select_test fts3rnd-1.$nodesize.$iTest.3.$i {
343        SELECT docid FROM t1 WHERE t1 MATCH $match
344      } [simple_phrase $term]
345    }
346
347    # Three word phrases.
348    #
349    for {set i 0} {$i < $nRep} {incr i} {
350      set term [list [random_term] [random_term] [random_term]]
351      set match "\"$term\""
352      do_select_test fts3rnd-1.$nodesize.$iTest.4.$i {
353        SELECT docid FROM t1 WHERE t1 MATCH $match
354      } [simple_phrase $term]
355    }
356
357    # Three word phrases made up of term-prefixes.
358    #
359    for {set i 0} {$i < $nRep} {incr i} {
360      set    query "[string range [random_term] 0 end-1]* "
361      append query "[string range [random_term] 0 end-1]* "
362      append query "[string range [random_term] 0 end-1]*"
363
364      set match "\"$query\""
365      do_select_test fts3rnd-1.$nodesize.$iTest.5.$i {
366        SELECT docid FROM t1 WHERE t1 MATCH $match
367      } [simple_phrase $query]
368    }
369
370    # A NEAR query with terms as the arguments.
371    #
372    for {set i 0} {$i < $nRep} {incr i} {
373      set terms [list [random_term] [random_term]]
374      set match [join $terms " NEAR "]
375      do_select_test fts3rnd-1.$nodesize.$iTest.6.$i {
376        SELECT docid FROM t1 WHERE t1 MATCH $match
377      } [simple_near $terms 10]
378    }
379
380    # A 3-way NEAR query with terms as the arguments.
381    #
382    for {set i 0} {$i < $nRep} {incr i} {
383      set terms [list [random_term] [random_term] [random_term]]
384      set nNear 11
385      set match [join $terms " NEAR/$nNear "]
386      do_select_test fts3rnd-1.$nodesize.$iTest.7.$i {
387        SELECT docid FROM t1 WHERE t1 MATCH $match
388      } [simple_near $terms $nNear]
389    }
390
391    # Set operations on simple term queries.
392    #
393    foreach {tn op proc} {
394      8  OR  setop_or
395      9  NOT setop_not
396      10 AND setop_and
397    } {
398      for {set i 0} {$i < $nRep} {incr i} {
399        set term1 [random_term]
400        set term2 [random_term]
401        set match "$term1 $op $term2"
402        do_select_test fts3rnd-1.$nodesize.$iTest.$tn.$i {
403          SELECT docid FROM t1 WHERE t1 MATCH $match
404        } [$proc [simple_phrase $term1] [simple_phrase $term2]]
405      }
406    }
407
408    # Set operations on NEAR queries.
409    #
410    foreach {tn op proc} {
411      8  OR  setop_or
412      9  NOT setop_not
413      10 AND setop_and
414    } {
415      for {set i 0} {$i < $nRep} {incr i} {
416        set term1 [random_term]
417        set term2 [random_term]
418        set term3 [random_term]
419        set term4 [random_term]
420        set match "$term1 NEAR $term2 $op $term3 NEAR $term4"
421        do_select_test fts3rnd-1.$nodesize.$iTest.$tn.$i {
422          SELECT docid FROM t1 WHERE t1 MATCH $match
423        } [$proc                                  \
424            [simple_near [list $term1 $term2] 10] \
425            [simple_near [list $term3 $term4] 10]
426          ]
427      }
428    }
429
430    catchsql COMMIT
431  }
432}
433
434finish_test
435