author | Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com> |
Tue, 06 Jul 2010 15:30:04 +0300 | |
changeset 10 | afe194b6b1cd |
parent 2 | 6c1a2771f4b7 |
child 15 | cf5c74390b98 |
permissions | -rw-r--r-- |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
1 |
/*------------------------------------------------------------------------------ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
2 |
* Copyright (C) 2003-2006 Ben van Klinken and the CLucene Team |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
3 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
4 |
* Distributable under the terms of either the Apache License (Version 2.0) or |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
5 |
* the GNU Lesser General Public License, as specified in the COPYING file. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
6 |
------------------------------------------------------------------------------*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
7 |
#include "clucene/stdheader.h" |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
8 |
#include "StandardTokenizer.h" |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
9 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
10 |
CL_NS_USE(analysis) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
11 |
CL_NS_USE(util) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
12 |
CL_NS_DEF2(analysis,standard) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
13 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
14 |
const TCHAR* tokenImageArray[] = { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
15 |
_T("<EOF>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
16 |
_T("<UNKNOWN>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
17 |
_T("<ALPHANUM>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
18 |
_T("<APOSTROPHE>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
19 |
_T("<ACRONYM>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
20 |
_T("<COMPANY>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
21 |
_T("<EMAIL>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
22 |
_T("<HOST>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
23 |
_T("<NUM>"), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
24 |
_T("<CJK>") |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
25 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
26 |
const TCHAR** tokenImage = tokenImageArray; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
27 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
28 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
29 |
/* A bunch of shortcut macros, many of which make assumptions about variable |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
30 |
** names. These macros enhance readability, not just convenience! */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
31 |
#define EOS (ch==-1 || rd->Eos()) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
32 |
#define SPACE (_istspace((TCHAR)ch) != 0) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
33 |
#define ALPHA (_istalpha((TCHAR)ch) != 0) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
34 |
#define ALNUM (_istalnum(ch) != 0) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
35 |
#define DIGIT (_istdigit(ch) != 0) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
36 |
#define UNDERSCORE (ch == '_') |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
37 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
38 |
#define _CJK ( (ch>=0x3040 && ch<=0x318f) || \ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
39 |
(ch>=0x3300 && ch<=0x337f) || \ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
40 |
(ch>=0x3400 && ch<=0x3d2d) || \ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
41 |
(ch>=0x4e00 && ch<=0x9fff) || \ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
42 |
(ch>=0xf900 && ch<=0xfaff) || \ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
43 |
(ch>=0xac00 && ch<=0xd7af) ) //korean |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
44 |
|
10
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
45 |
// CHANGED Thai support -> |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
46 |
#define _THAI ( ch >= 0x0E00 && ch <= 0x0Eff ) |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
47 |
// CHANGED |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
48 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
49 |
#define DASH (ch == '-') |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
50 |
#define NEGATIVE_SIGN_ DASH |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
51 |
//#define POSITIVE_SIGN_ (ch == '+') |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
52 |
//#define SIGN (NEGATIVE_SIGN_ || POSITIVE_SIGN_) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
53 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
54 |
#define DOT (ch == '.') |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
55 |
#define DECIMAL DOT |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
56 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
57 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
58 |
//freebsd seems to have a problem with defines over multiple lines, so this has to be one long line |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
59 |
#define _CONSUME_AS_LONG_AS(conditionFails) while (true) { ch = readChar(); if (ch==-1 || (!(conditionFails) || str.len >= LUCENE_MAX_WORD_LEN)) { break; } str.appendChar(ch);} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
60 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
61 |
#define CONSUME_ALPHAS _CONSUME_AS_LONG_AS(ALPHA) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
62 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
63 |
#define CONSUME_DIGITS _CONSUME_AS_LONG_AS(DIGIT) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
64 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
65 |
/* otherMatches is a condition (possibly compound) under which a character |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
66 |
** that's not an ALNUM or UNDERSCORE can be considered not to break the |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
67 |
** span. Callers should pass false if only ALNUM/UNDERSCORE are acceptable. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
68 |
#define CONSUME_WORD _CONSUME_AS_LONG_AS(ALNUM || UNDERSCORE) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
69 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
70 |
/* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
71 |
** Consume CJK characters |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
72 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
73 |
#define CONSUME_CJK _CONSUME_AS_LONG_AS(_CJK) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
74 |
|
10
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
75 |
// CHANGED Thai support -> |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
76 |
/** |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
77 |
* Consume Thai |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
78 |
*/ |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
79 |
#define CONSUME_THAI _CONSUME_AS_LONG_AS(_THAI) |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
80 |
// CHANGED |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
81 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
82 |
/* It is considered that "nothing of value" has been read if: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
83 |
** a) The "read head" hasn't moved since specialCharPos was established. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
84 |
** or |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
85 |
** b) The "read head" has moved by one character, but that character was |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
86 |
** either whitespace or not among the characters found in the body of |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
87 |
** a token (deliberately doesn't include the likes of '@'/'&'). */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
88 |
#define CONSUMED_NOTHING_OF_VALUE (rdPos == specialCharPos || (rdPos == specialCharPos+1 && ( SPACE || !(ALNUM || DOT || DASH || UNDERSCORE) ))) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
89 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
90 |
#define RIGHTMOST(sb) (sb.getBuffer()[sb.len-1]) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
91 |
#define RIGHTMOST_IS(sb, c) (RIGHTMOST(sb) == c) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
92 |
/* To discard the last character in a StringBuffer, we decrement the buffer's |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
93 |
** length indicator and move the terminator back by one character. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
94 |
#define SHAVE_RIGHTMOST(sb) (sb.getBuffer()[--sb.len] = '\0') |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
95 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
96 |
//#define REMOVE_TRAILING_CHARS(sb, charMatchesCondition) { TCHAR* sbBuf = sb.getBuffer(); for (int32_t i = sb.len-1; i >= 0; i--) { TCHAR c = sbBuf[i]; if (charMatchesCondition) { sbBuf[--sb.len] = '\0'; } else {break;}}} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
97 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
98 |
/* Does StringBuffer sb contain any of the characters in string ofThese? */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
99 |
#define CONTAINS_ANY(sb, ofThese) (_tcscspn(sb.getBuffer(), _T(ofThese)) != static_cast<size_t>(sb.len)) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
100 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
101 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
102 |
StandardTokenizer::StandardTokenizer(Reader* reader): |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
103 |
/* rdPos is zero-based. It starts at -1, and will advance to the first |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
104 |
** position when readChar() is first called. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
105 |
rdPos(-1), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
106 |
tokenStart(-1), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
107 |
rd(_CLNEW FastCharStream(reader)) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
108 |
{ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
109 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
110 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
111 |
StandardTokenizer::~StandardTokenizer() { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
112 |
_CLDELETE(rd); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
113 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
114 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
115 |
int StandardTokenizer::readChar() { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
116 |
/* Increment by 1 because we're speaking in terms of characters, not |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
117 |
** necessarily bytes: */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
118 |
rdPos++; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
119 |
return rd->GetNext(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
120 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
121 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
122 |
void StandardTokenizer::unReadChar() { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
123 |
rd->UnGet(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
124 |
rdPos--; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
125 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
126 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
127 |
inline bool StandardTokenizer::setToken(Token* t, StringBuffer* sb, TokenTypes tokenCode) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
128 |
t->setStartOffset(tokenStart); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
129 |
t->setEndOffset(tokenStart+sb->length()); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
130 |
t->setType(tokenImage[tokenCode]); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
131 |
sb->getBuffer(); //null terminates the buffer |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
132 |
t->resetTermTextLen(); |
10
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
133 |
t->setPositionIncrement(1); |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
134 |
return true; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
135 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
136 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
137 |
bool StandardTokenizer::next(Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
138 |
int ch=0; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
139 |
while (!EOS) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
140 |
ch = readChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
141 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
142 |
if ( ch == 0 || ch == -1 ){ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
143 |
continue; |
10
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
144 |
// CHANGED Thai Support -> |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
145 |
} else if ( _THAI ) { |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
146 |
if ( ReadThai(ch,t) ) |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
147 |
return true; |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
148 |
// CHANGED |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
149 |
} else if (SPACE) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
150 |
continue; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
151 |
} else if (ALPHA || UNDERSCORE) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
152 |
tokenStart = rdPos; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
153 |
return ReadAlphaNum(ch,t); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
154 |
} else if (DIGIT || NEGATIVE_SIGN_ || DECIMAL) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
155 |
tokenStart = rdPos; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
156 |
/* ReadNumber returns NULL if it fails to extract a valid number; in |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
157 |
** that case, we just continue. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
158 |
if (ReadNumber(NULL, ch,t)) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
159 |
return true; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
160 |
} else if ( _CJK ){ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
161 |
if ( ReadCJK(ch,t) ) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
162 |
return true; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
163 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
164 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
165 |
return false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
166 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
167 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
168 |
bool StandardTokenizer::ReadNumber(const TCHAR* previousNumber, const TCHAR prev,Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
169 |
/* previousNumber is only non-NULL if this function already read a complete |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
170 |
** number in a previous recursion, yet has been asked to read additional |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
171 |
** numeric segments. For example, in the HOST "192.168.1.3", "192.168" is |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
172 |
** a complete number, but this function will recurse to read the "1.3", |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
173 |
** generating a single HOST token "192.168.1.3". */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
174 |
t->growBuffer(LUCENE_MAX_WORD_LEN+1);//make sure token can hold the next word |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
175 |
StringBuffer str(t->_termText,t->bufferLength(),true); //use stringbuffer to read data onto the termText |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
176 |
TokenTypes tokenType; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
177 |
bool decExhausted; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
178 |
if (previousNumber != NULL) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
179 |
str.prepend(previousNumber); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
180 |
tokenType = CL_NS2(analysis,standard)::HOST; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
181 |
decExhausted = false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
182 |
} else { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
183 |
tokenType = CL_NS2(analysis,standard)::NUM; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
184 |
decExhausted = (prev == '.'); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
185 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
186 |
if ( str.len >= LUCENE_MAX_WORD_LEN ){ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
187 |
//if a number is too long, i would say there is no point |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
188 |
//storing it, because its going to be the wrong number anyway? |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
189 |
//what do people think? |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
190 |
return false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
191 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
192 |
str.appendChar(prev); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
193 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
194 |
const bool signExhausted = (prev == '-'); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
195 |
int ch = prev; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
196 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
197 |
CONSUME_DIGITS; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
198 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
199 |
if (str.len < 2 /* CONSUME_DIGITS didn't find any digits. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
200 |
&& ( |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
201 |
(signExhausted && !DECIMAL) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
202 |
|| (decExhausted /* && !DIGIT is implied, since CONSUME_DIGITS stopped on a non-digit. */) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
203 |
) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
204 |
) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
205 |
{ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
206 |
/* We have either: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
207 |
** a) a negative sign that's not followed by either digit(s) or a decimal |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
208 |
** b) a decimal that's not followed by digit(s) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
209 |
** so this is not a valid number. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
210 |
if (!EOS) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
211 |
/* Unread the character that stopped CONSUME_DIGITS: */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
212 |
unReadChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
213 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
214 |
return false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
215 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
216 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
217 |
/* We just read a group of digits. Is it followed by a decimal symbol, |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
218 |
** implying that there might be another group of digits available? */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
219 |
if (!EOS) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
220 |
if (DECIMAL) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
221 |
if ( str.len >= LUCENE_MAX_WORD_LEN ) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
222 |
return false; //read above for rationale |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
223 |
str.appendChar(ch); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
224 |
} else { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
225 |
unReadChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
226 |
goto SUCCESSFULLY_EXTRACTED_NUMBER; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
227 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
228 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
229 |
CONSUME_DIGITS; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
230 |
if (!DIGIT && !DECIMAL) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
231 |
unReadChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
232 |
} else if (!EOS && DECIMAL && _istdigit(rd->Peek())) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
233 |
/* We just read the fractional digit group, but it's also followed by |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
234 |
** a decimal symbol and at least one more digit, so this must be a |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
235 |
** HOST rather than a real number. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
236 |
return ReadNumber(str.getBuffer(), '.',t); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
237 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
238 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
239 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
240 |
SUCCESSFULLY_EXTRACTED_NUMBER: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
241 |
TCHAR rightmost = RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
242 |
/* Don't including a trailing decimal point. */ |
10
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
243 |
if(ALPHA){ |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
244 |
return ReadAlphaNum(prev,t); |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
245 |
} |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
246 |
if (rightmost == '.') { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
247 |
SHAVE_RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
248 |
unReadChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
249 |
rightmost = RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
250 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
251 |
/* If all we have left is a negative sign, it's not a valid number. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
252 |
if (rightmost == '-') { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
253 |
CND_PRECONDITION (str.len == 1, "Number is invalid"); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
254 |
return false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
255 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
256 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
257 |
return setToken(t,&str,tokenType); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
258 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
259 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
260 |
bool StandardTokenizer::ReadAlphaNum(const TCHAR prev, Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
261 |
t->growBuffer(LUCENE_MAX_WORD_LEN+1);//make sure token can hold the next word |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
262 |
StringBuffer str(t->_termText,t->bufferLength(),true); //use stringbuffer to read data onto the termText |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
263 |
if ( str.len < LUCENE_MAX_WORD_LEN ){ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
264 |
str.appendChar(prev); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
265 |
int ch = prev; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
266 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
267 |
CONSUME_WORD; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
268 |
if (!EOS && str.len < LUCENE_MAX_WORD_LEN-1 ) { //still have space for 1 more character? |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
269 |
switch(ch) { /* What follows the first alphanum segment? */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
270 |
case '.': |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
271 |
str.appendChar('.'); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
272 |
return ReadDotted(&str, CL_NS2(analysis,standard)::UNKNOWN,t); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
273 |
case '\'': |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
274 |
str.appendChar('\''); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
275 |
return ReadApostrophe(&str,t); |
2
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
276 |
// case '@': |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
277 |
// str.appendChar('@'); |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
278 |
// return ReadAt(&str,t); |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
279 |
case '&': |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
280 |
str.appendChar('&'); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
281 |
return ReadCompany(&str,t); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
282 |
/* default: fall through to end of this function. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
283 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
284 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
285 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
286 |
return setToken(t,&str,CL_NS2(analysis,standard)::ALPHANUM); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
287 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
288 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
289 |
bool StandardTokenizer::ReadCJK(const TCHAR prev, Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
290 |
t->growBuffer(LUCENE_MAX_WORD_LEN+1);//make sure token can hold the next word |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
291 |
StringBuffer str(t->_termText,t->bufferLength(),true); //use stringbuffer to read data onto the termText |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
292 |
if ( str.len < LUCENE_MAX_WORD_LEN ){ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
293 |
str.appendChar(prev); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
294 |
int ch = prev; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
295 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
296 |
CONSUME_CJK; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
297 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
298 |
return setToken(t,&str,CL_NS2(analysis,standard)::CJK); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
299 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
300 |
|
10
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
301 |
// CHANGED Thai Support |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
302 |
bool StandardTokenizer::ReadThai(const TCHAR prev, Token* t) { |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
303 |
t->growBuffer(LUCENE_MAX_WORD_LEN+1);//make sure token can hold the next word |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
304 |
StringBuffer str(t->_termText,t->bufferLength(),true); //use stringbuffer to read data onto the termText |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
305 |
if ( str.len < LUCENE_MAX_WORD_LEN ){ |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
306 |
str.appendChar(prev); |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
307 |
int ch = prev; |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
308 |
|
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
309 |
CONSUME_THAI; |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
310 |
} |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
311 |
return setToken(t,&str,CL_NS2(analysis,standard)::Thai); |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
312 |
} |
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
313 |
// CHANGED |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
314 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
315 |
bool StandardTokenizer::ReadDotted(StringBuffer* _str, TokenTypes forcedType, Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
316 |
const int32_t specialCharPos = rdPos; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
317 |
StringBuffer& str=*_str; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
318 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
319 |
/* A segment of a "dotted" is not allowed to begin with another dot or a dash. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
320 |
** Even though hosts, e-mail addresses, etc., could have a dotted-segment |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
321 |
** that begins with a dot or a dash, it's far more common in source text |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
322 |
** for a pattern like "abc.--def" to be intended as two tokens. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
323 |
int ch = rd->Peek(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
324 |
if (!(DOT || DASH)) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
325 |
bool prevWasDot; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
326 |
bool prevWasDash; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
327 |
if (str.len == 0) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
328 |
prevWasDot = false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
329 |
prevWasDash = false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
330 |
} else { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
331 |
prevWasDot = RIGHTMOST(str) == '.'; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
332 |
prevWasDash = RIGHTMOST(str) == '-'; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
333 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
334 |
while (!EOS && str.len < LUCENE_MAX_WORD_LEN-1 ) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
335 |
ch = readChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
336 |
const bool dot = ch == '.'; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
337 |
const bool dash = ch == '-'; |
2
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
338 |
//const bool at = ch == '@'; |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
339 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
340 |
if (!(ALNUM || UNDERSCORE || dot || dash)) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
341 |
break; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
342 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
343 |
/* Multiple dots or dashes in succession end the token. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
344 |
** Consider the following inputs: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
345 |
** "Visit windowsupdate.microsoft.com--update today!" |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
346 |
** "In the U.S.A.--yes, even there!" */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
347 |
if ((dot || dash) && (prevWasDot || prevWasDash)) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
348 |
/* We're not going to append the character we just read, in any case. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
349 |
** As to the character before it (which is currently RIGHTMOST(str)): |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
350 |
** Unless RIGHTMOST(str) is a dot, in which we need to save it so the |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
351 |
** acronym-versus-host detection can work, we want to get rid of it. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
352 |
if (!prevWasDot) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
353 |
SHAVE_RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
354 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
355 |
break; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
356 |
} |
2
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
357 |
|
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
358 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
359 |
str.appendChar(ch); |
2
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
360 |
|
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
361 |
|
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
362 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
363 |
prevWasDot = dot; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
364 |
prevWasDash = dash; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
365 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
366 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
367 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
368 |
/* There's a potential StringBuffer.append call in the code above, which |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
369 |
** could cause str to reallocate its internal buffer. We must wait to |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
370 |
** obtain the optimization-oriented strBuf pointer until after the initial |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
371 |
** potentially realloc-triggering operations on str. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
372 |
** Because there can be other such ops much later in this function, strBuf |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
373 |
** is guarded within a block to prevent its use during or after the calls |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
374 |
** that would potentially invalidate it. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
375 |
{ /* Begin block-guard of strBuf */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
376 |
TCHAR* strBuf = str.getBuffer(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
377 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
378 |
bool rightmostIsDot = RIGHTMOST_IS(str, '.'); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
379 |
if (CONSUMED_NOTHING_OF_VALUE) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
380 |
/* No more alphanums available for this token; shave trailing dot, if any. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
381 |
if (rightmostIsDot) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
382 |
SHAVE_RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
383 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
384 |
/* If there are no dots remaining, this is a generic ALPHANUM. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
385 |
if (_tcschr(strBuf, '.') == NULL) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
386 |
forcedType = CL_NS2(analysis,standard)::ALPHANUM; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
387 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
388 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
389 |
/* Check the token to see if it's an acronym. An acronym must have a |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
390 |
** letter in every even slot and a dot in every odd slot, including the |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
391 |
** last slot (for example, "U.S.A."). */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
392 |
} else if (rightmostIsDot) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
393 |
bool isAcronym = true; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
394 |
const int32_t upperCheckLimit = str.len - 1; /* -1 b/c we already checked the last slot. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
395 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
396 |
for (int32_t i = 0; i < upperCheckLimit; i++) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
397 |
const bool even = (i % 2 == 0); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
398 |
ch = strBuf[i]; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
399 |
if ( (even && !ALPHA) || (!even && !DOT) ) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
400 |
isAcronym = false; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
401 |
break; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
402 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
403 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
404 |
if (isAcronym) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
405 |
forcedType = CL_NS2(analysis,standard)::ACRONYM; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
406 |
} else { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
407 |
/* If it's not an acronym, we don't want the trailing dot. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
408 |
SHAVE_RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
409 |
/* If there are no dots remaining, this is a generic ALPHANUM. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
410 |
if (_tcschr(strBuf, '.') == NULL) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
411 |
forcedType = CL_NS2(analysis,standard)::ALPHANUM; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
412 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
413 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
414 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
415 |
} /* End block-guard of strBuf */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
416 |
|
2
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
417 |
|
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
418 |
|
2
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
419 |
// if (!EOS) { |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
420 |
// if (ch == '@' && str.len < LUCENE_MAX_WORD_LEN-1) { |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
421 |
// str.appendChar('@'); |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
422 |
// return ReadAt(&str,t); |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
423 |
// } else { |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
424 |
// unReadChar(); |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
425 |
// } |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
426 |
|
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
427 |
if (!EOS) { |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
428 |
unReadChar(); |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
429 |
} |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
430 |
|
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
431 |
|
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
432 |
|
10
afe194b6b1cd
Revision: 201025
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
2
diff
changeset
|
433 |
return setToken(t,&str,forcedType); |
2
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
434 |
// return setToken(t,&str,CL_NS2(analysis,standard)::UNKNOWN |
6c1a2771f4b7
Revision: 201017
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
0
diff
changeset
|
435 |
// ? forcedType : CL_NS2(analysis,standard)::HOST); |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
436 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
437 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
438 |
bool StandardTokenizer::ReadApostrophe(StringBuffer* _str, Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
439 |
StringBuffer& str=*_str; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
440 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
441 |
TokenTypes tokenType = CL_NS2(analysis,standard)::APOSTROPHE; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
442 |
const int32_t specialCharPos = rdPos; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
443 |
int ch=0; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
444 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
445 |
CONSUME_ALPHAS; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
446 |
if (RIGHTMOST_IS(str, '\'') || CONSUMED_NOTHING_OF_VALUE) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
447 |
/* After the apostrophe, no more alphanums were available within this |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
448 |
** token; shave trailing apostrophe and revert to generic ALPHANUM. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
449 |
SHAVE_RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
450 |
tokenType = CL_NS2(analysis,standard)::ALPHANUM; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
451 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
452 |
if (!EOS) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
453 |
unReadChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
454 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
455 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
456 |
return setToken(t,&str,tokenType); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
457 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
458 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
459 |
bool StandardTokenizer::ReadAt(StringBuffer* str, Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
460 |
ReadDotted(str, CL_NS2(analysis,standard)::EMAIL,t); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
461 |
/* JLucene grammar indicates dots/digits not allowed in company name: */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
462 |
if (!CONTAINS_ANY((*str), ".0123456789")) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
463 |
setToken(t,str,CL_NS2(analysis,standard)::COMPANY); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
464 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
465 |
return true; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
466 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
467 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
468 |
bool StandardTokenizer::ReadCompany(StringBuffer* _str, Token* t) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
469 |
StringBuffer& str = *_str; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
470 |
const int32_t specialCharPos = rdPos; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
471 |
int ch=0; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
472 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
473 |
CONSUME_WORD; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
474 |
if (CONSUMED_NOTHING_OF_VALUE) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
475 |
/* After the ampersand, no more alphanums were available within this |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
476 |
** token; shave trailing ampersand and revert to ALPHANUM. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
477 |
CND_PRECONDITION(RIGHTMOST_IS(str, '&'),"ReadCompany failed"); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
478 |
SHAVE_RIGHTMOST(str); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
479 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
480 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
481 |
return setToken(t,&str,CL_NS2(analysis,standard)::ALPHANUM); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
482 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
483 |
if (!EOS) { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
484 |
unReadChar(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
485 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
486 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
487 |
return setToken(t,&str,CL_NS2(analysis,standard)::COMPANY); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
488 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
489 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
490 |
CL_NS_END2 |