sl@0: # 2007 June 21
sl@0: #
sl@0: # The author disclaims copyright to this source code.  In place of
sl@0: # a legal notice, here is a blessing:
sl@0: #
sl@0: #    May you do good and not evil.
sl@0: #    May you find forgiveness for yourself and forgive others.
sl@0: #    May you share freely, never taking more than you give.
sl@0: #
sl@0: #*************************************************************************
sl@0: # This file implements regression tests for SQLite library. The focus 
sl@0: # of this script is testing the pluggable tokeniser feature of the 
sl@0: # FTS2 module.
sl@0: #
sl@0: # $Id: fts2token.test,v 1.3 2007/06/25 12:05:40 danielk1977 Exp $
sl@0: #
sl@0: 
sl@0: set testdir [file dirname $argv0]
sl@0: source $testdir/tester.tcl
sl@0: 
sl@0: # If SQLITE_ENABLE_FTS2 is defined, omit this file.
sl@0: ifcapable !fts2 {
sl@0:   finish_test
sl@0:   return
sl@0: }
sl@0: 
sl@0: proc escape_string {str} {
sl@0:   set out ""
sl@0:   foreach char [split $str ""] {
sl@0:     scan $char %c i
sl@0:     if {$i<=127} {
sl@0:       append out $char
sl@0:     } else {
sl@0:       append out [format {\x%.4x} $i]
sl@0:     }
sl@0:   }
sl@0:   set out
sl@0: }
sl@0: 
sl@0: #--------------------------------------------------------------------------
sl@0: # Test cases fts2token-1.* are the warm-body test for the SQL scalar
sl@0: # function fts2_tokenizer(). The procedure is as follows:
sl@0: #
sl@0: #   1: Verify that there is no such fts2 tokenizer as 'blah'.
sl@0: #
sl@0: #   2: Query for the built-in tokenizer 'simple'. Insert a copy of the
sl@0: #      retrieved value as tokenizer 'blah'.
sl@0: #
sl@0: #   3: Test that the value returned for tokenizer 'blah' is now the
sl@0: #      same as that retrieved for 'simple'.
sl@0: #
sl@0: #   4: Test that it is now possible to create an fts2 table using 
sl@0: #      tokenizer 'blah' (it was not possible in step 1).
sl@0: #
sl@0: #   5: Test that the table created to use tokenizer 'blah' is usable.
sl@0: #
sl@0: do_test fts2token-1.1 {
sl@0:   catchsql {
sl@0:     CREATE VIRTUAL TABLE t1 USING fts2(content, tokenize blah);
sl@0:   }
sl@0: } {1 {unknown tokenizer: blah}}
sl@0: do_test fts2token-1.2 {
sl@0:   execsql {
sl@0:     SELECT fts2_tokenizer('blah', fts2_tokenizer('simple')) IS NULL;
sl@0:   }
sl@0: } {0}
sl@0: do_test fts2token-1.3 {
sl@0:   execsql {
sl@0:     SELECT fts2_tokenizer('blah') == fts2_tokenizer('simple');
sl@0:   }
sl@0: } {1}
sl@0: do_test fts2token-1.4 {
sl@0:   catchsql {
sl@0:     CREATE VIRTUAL TABLE t1 USING fts2(content, tokenize blah);
sl@0:   }
sl@0: } {0 {}}
sl@0: do_test fts2token-1.5 {
sl@0:   execsql {
sl@0:     INSERT INTO t1(content) VALUES('There was movement at the station');
sl@0:     INSERT INTO t1(content) VALUES('For the word has passed around');
sl@0:     INSERT INTO t1(content) VALUES('That the colt from ol regret had got away');
sl@0:     SELECT content FROM t1 WHERE content MATCH 'movement'
sl@0:   }
sl@0: } {{There was movement at the station}}
sl@0: 
sl@0: #--------------------------------------------------------------------------
sl@0: # Test cases fts2token-2.* test error cases in the scalar function based
sl@0: # API for getting and setting tokenizers.
sl@0: #
sl@0: do_test fts2token-2.1 {
sl@0:   catchsql {
sl@0:     SELECT fts2_tokenizer('nosuchtokenizer');
sl@0:   }
sl@0: } {1 {unknown tokenizer: nosuchtokenizer}}
sl@0: 
sl@0: #--------------------------------------------------------------------------
sl@0: # Test cases fts2token-3.* test the three built-in tokenizers with a
sl@0: # simple input string via the built-in test function. This is as much
sl@0: # to test the test function as the tokenizer implementations.
sl@0: #
sl@0: do_test fts2token-3.1 {
sl@0:   execsql {
sl@0:     SELECT fts2_tokenizer_test('simple', 'I don''t see how');
sl@0:   }
sl@0: } {{0 i I 1 don don 2 t t 3 see see 4 how how}}
sl@0: do_test fts2token-3.2 {
sl@0:   execsql {
sl@0:     SELECT fts2_tokenizer_test('porter', 'I don''t see how');
sl@0:   }
sl@0: } {{0 i I 1 don don 2 t t 3 see see 4 how how}}
sl@0: ifcapable icu {
sl@0:   do_test fts2token-3.3 {
sl@0:     execsql {
sl@0:       SELECT fts2_tokenizer_test('icu', 'I don''t see how');
sl@0:     }
sl@0:   } {{0 i I 1 don't don't 2 see see 3 how how}}
sl@0: }
sl@0: 
sl@0: #--------------------------------------------------------------------------
sl@0: # Test cases fts2token-4.* test the ICU tokenizer. In practice, this
sl@0: # tokenizer only has two modes - "thai" and "everybody else". Some other
sl@0: # Asian languages (Lao, Khmer etc.) require the same special treatment as 
sl@0: # Thai, but ICU doesn't support them yet.
sl@0: #
sl@0: ifcapable icu {
sl@0: 
sl@0:   proc do_icu_test {name locale input output} {
sl@0:     set ::out [db eval { SELECT fts2_tokenizer_test('icu', $locale, $input) }]
sl@0:     do_test $name {
sl@0:       lindex $::out 0
sl@0:     } $output
sl@0:   }
sl@0:   
sl@0:   do_icu_test fts2token-4.1 en_US  {}   {}
sl@0:   do_icu_test fts2token-4.2 en_US {Test cases fts2} [list \
sl@0:     0 test Test 1 cases cases 2 fts2 fts2
sl@0:   ]
sl@0: 
sl@0:   # The following test shows that ICU is smart enough to recognise
sl@0:   # Thai chararacters, even when the locale is set to English/United 
sl@0:   # States.
sl@0:   #
sl@0:   set input "\u0e2d\u0e30\u0e44\u0e23\u0e19\u0e30\u0e04\u0e23\u0e31\u0e1a"
sl@0:   set output    "0 \u0e2d\u0e30\u0e44\u0e23 \u0e2d\u0e30\u0e44\u0e23 "
sl@0:   append output "1 \u0e19\u0e30 \u0e19\u0e30 "
sl@0:   append output "2 \u0e04\u0e23\u0e31\u0e1a \u0e04\u0e23\u0e31\u0e1a"
sl@0: 
sl@0:   do_icu_test fts2token-4.3 th_TH  $input $output
sl@0:   do_icu_test fts2token-4.4 en_US  $input $output
sl@0: 
sl@0:   # ICU handles an unknown locale by falling back to the default.
sl@0:   # So this is not an error.
sl@0:   do_icu_test fts2token-4.5 MiddleOfTheOcean  $input $output
sl@0: 
sl@0:   set    longtoken "AReallyReallyLongTokenOneThatWillSurelyRequire"
sl@0:   append longtoken "AReallocInTheIcuTokenizerCode"
sl@0: 
sl@0:   set    input "short tokens then "
sl@0:   append input $longtoken
sl@0:   set    output "0 short short "
sl@0:   append output "1 tokens tokens "
sl@0:   append output "2 then then "
sl@0:   append output "3 [string tolower $longtoken] $longtoken"
sl@0: 
sl@0:   do_icu_test fts2token-4.6 MiddleOfTheOcean  $input $output
sl@0:   do_icu_test fts2token-4.7 th_TH  $input $output
sl@0:   do_icu_test fts2token-4.8 en_US  $input $output
sl@0: }
sl@0: 
sl@0: do_test fts2token-internal {
sl@0:   execsql { SELECT fts2_tokenizer_internal_test() }
sl@0: } {ok}
sl@0: 
sl@0: finish_test