ferret 0.9.6 → 0.10.0
Sign up to get free protection for your applications and to get access to all the features.
- data/MIT-LICENSE +1 -1
- data/README +12 -24
- data/Rakefile +38 -54
- data/TODO +14 -17
- data/ext/analysis.c +982 -823
- data/ext/analysis.h +133 -76
- data/ext/array.c +96 -58
- data/ext/array.h +40 -13
- data/ext/bitvector.c +476 -118
- data/ext/bitvector.h +264 -22
- data/ext/compound_io.c +217 -229
- data/ext/defines.h +49 -0
- data/ext/document.c +107 -317
- data/ext/document.h +31 -65
- data/ext/except.c +81 -36
- data/ext/except.h +117 -55
- data/ext/extconf.rb +2 -9
- data/ext/ferret.c +211 -104
- data/ext/ferret.h +22 -11
- data/ext/filter.c +97 -82
- data/ext/fs_store.c +348 -367
- data/ext/global.c +226 -188
- data/ext/global.h +44 -26
- data/ext/hash.c +474 -391
- data/ext/hash.h +441 -68
- data/ext/hashset.c +124 -96
- data/ext/hashset.h +169 -20
- data/ext/helper.c +56 -5
- data/ext/helper.h +7 -0
- data/ext/inc/lang.h +29 -49
- data/ext/inc/threading.h +31 -0
- data/ext/ind.c +288 -278
- data/ext/ind.h +68 -0
- data/ext/index.c +5688 -0
- data/ext/index.h +663 -616
- data/ext/lang.h +29 -49
- data/ext/libstemmer.c +3 -3
- data/ext/mem_pool.c +84 -0
- data/ext/mem_pool.h +35 -0
- data/ext/posh.c +1006 -0
- data/ext/posh.h +1007 -0
- data/ext/priorityqueue.c +117 -194
- data/ext/priorityqueue.h +135 -39
- data/ext/q_boolean.c +1305 -1108
- data/ext/q_const_score.c +106 -93
- data/ext/q_filtered_query.c +138 -135
- data/ext/q_fuzzy.c +206 -242
- data/ext/q_match_all.c +94 -80
- data/ext/q_multi_term.c +663 -0
- data/ext/q_parser.c +667 -593
- data/ext/q_phrase.c +992 -555
- data/ext/q_prefix.c +72 -61
- data/ext/q_range.c +235 -210
- data/ext/q_span.c +1480 -1166
- data/ext/q_term.c +273 -246
- data/ext/q_wildcard.c +127 -114
- data/ext/r_analysis.c +1720 -711
- data/ext/r_index.c +3049 -0
- data/ext/r_qparser.c +433 -146
- data/ext/r_search.c +2934 -1993
- data/ext/r_store.c +372 -143
- data/ext/r_utils.c +941 -0
- data/ext/ram_store.c +330 -326
- data/ext/search.c +1291 -668
- data/ext/search.h +403 -702
- data/ext/similarity.c +91 -113
- data/ext/similarity.h +45 -30
- data/ext/sort.c +721 -484
- data/ext/stopwords.c +361 -273
- data/ext/store.c +556 -58
- data/ext/store.h +706 -126
- data/ext/tags +3578 -2780
- data/ext/term_vectors.c +352 -0
- data/ext/threading.h +31 -0
- data/ext/win32.h +54 -0
- data/lib/ferret.rb +5 -17
- data/lib/ferret/document.rb +130 -2
- data/lib/ferret/index.rb +577 -26
- data/lib/ferret/number_tools.rb +157 -0
- data/lib/ferret_version.rb +3 -0
- data/test/test_helper.rb +5 -13
- data/test/unit/analysis/tc_analyzer.rb +513 -1
- data/test/unit/analysis/{ctc_tokenstream.rb → tc_token_stream.rb} +23 -0
- data/test/unit/index/tc_index.rb +183 -240
- data/test/unit/index/tc_index_reader.rb +312 -479
- data/test/unit/index/tc_index_writer.rb +397 -13
- data/test/unit/index/th_doc.rb +269 -206
- data/test/unit/query_parser/tc_query_parser.rb +40 -33
- data/test/unit/search/tc_filter.rb +59 -71
- data/test/unit/search/tc_fuzzy_query.rb +24 -16
- data/test/unit/search/tc_index_searcher.rb +23 -201
- data/test/unit/search/tc_multi_searcher.rb +78 -226
- data/test/unit/search/tc_search_and_sort.rb +93 -81
- data/test/unit/search/tc_sort.rb +23 -23
- data/test/unit/search/tc_sort_field.rb +7 -7
- data/test/unit/search/tc_spans.rb +51 -47
- data/test/unit/search/tm_searcher.rb +339 -0
- data/test/unit/store/tc_fs_store.rb +1 -1
- data/test/unit/store/tm_store_lock.rb +3 -3
- data/test/unit/tc_document.rb +81 -0
- data/test/unit/ts_analysis.rb +1 -1
- data/test/unit/ts_utils.rb +1 -1
- data/test/unit/utils/tc_bit_vector.rb +288 -0
- data/test/unit/utils/tc_number_tools.rb +117 -0
- data/test/unit/utils/tc_priority_queue.rb +106 -0
- metadata +140 -301
- data/CHANGELOG +0 -9
- data/ext/dummy.exe +0 -0
- data/ext/field.c +0 -408
- data/ext/frtio.h +0 -13
- data/ext/inc/except.h +0 -90
- data/ext/index_io.c +0 -382
- data/ext/index_rw.c +0 -2658
- data/ext/lang.c +0 -41
- data/ext/nix_io.c +0 -134
- data/ext/q_multi_phrase.c +0 -380
- data/ext/r_doc.c +0 -582
- data/ext/r_index_io.c +0 -1021
- data/ext/r_term.c +0 -219
- data/ext/term.c +0 -820
- data/ext/termdocs.c +0 -611
- data/ext/vector.c +0 -637
- data/ext/w32_io.c +0 -150
- data/lib/ferret/analysis.rb +0 -11
- data/lib/ferret/analysis/analyzers.rb +0 -112
- data/lib/ferret/analysis/standard_tokenizer.rb +0 -71
- data/lib/ferret/analysis/token.rb +0 -100
- data/lib/ferret/analysis/token_filters.rb +0 -86
- data/lib/ferret/analysis/token_stream.rb +0 -26
- data/lib/ferret/analysis/tokenizers.rb +0 -112
- data/lib/ferret/analysis/word_list_loader.rb +0 -27
- data/lib/ferret/document/document.rb +0 -152
- data/lib/ferret/document/field.rb +0 -312
- data/lib/ferret/index/compound_file_io.rb +0 -338
- data/lib/ferret/index/document_writer.rb +0 -289
- data/lib/ferret/index/field_infos.rb +0 -279
- data/lib/ferret/index/fields_io.rb +0 -181
- data/lib/ferret/index/index.rb +0 -675
- data/lib/ferret/index/index_file_names.rb +0 -33
- data/lib/ferret/index/index_reader.rb +0 -503
- data/lib/ferret/index/index_writer.rb +0 -534
- data/lib/ferret/index/multi_reader.rb +0 -377
- data/lib/ferret/index/multiple_term_doc_pos_enum.rb +0 -98
- data/lib/ferret/index/segment_infos.rb +0 -130
- data/lib/ferret/index/segment_merge_info.rb +0 -49
- data/lib/ferret/index/segment_merge_queue.rb +0 -16
- data/lib/ferret/index/segment_merger.rb +0 -358
- data/lib/ferret/index/segment_reader.rb +0 -412
- data/lib/ferret/index/segment_term_enum.rb +0 -169
- data/lib/ferret/index/segment_term_vector.rb +0 -58
- data/lib/ferret/index/term.rb +0 -53
- data/lib/ferret/index/term_buffer.rb +0 -83
- data/lib/ferret/index/term_doc_enum.rb +0 -291
- data/lib/ferret/index/term_enum.rb +0 -52
- data/lib/ferret/index/term_info.rb +0 -37
- data/lib/ferret/index/term_infos_io.rb +0 -321
- data/lib/ferret/index/term_vector_offset_info.rb +0 -20
- data/lib/ferret/index/term_vectors_io.rb +0 -553
- data/lib/ferret/query_parser.rb +0 -312
- data/lib/ferret/query_parser/query_parser.tab.rb +0 -928
- data/lib/ferret/search.rb +0 -50
- data/lib/ferret/search/boolean_clause.rb +0 -100
- data/lib/ferret/search/boolean_query.rb +0 -299
- data/lib/ferret/search/boolean_scorer.rb +0 -294
- data/lib/ferret/search/caching_wrapper_filter.rb +0 -40
- data/lib/ferret/search/conjunction_scorer.rb +0 -99
- data/lib/ferret/search/disjunction_sum_scorer.rb +0 -205
- data/lib/ferret/search/exact_phrase_scorer.rb +0 -32
- data/lib/ferret/search/explanation.rb +0 -41
- data/lib/ferret/search/field_cache.rb +0 -215
- data/lib/ferret/search/field_doc.rb +0 -31
- data/lib/ferret/search/field_sorted_hit_queue.rb +0 -184
- data/lib/ferret/search/filter.rb +0 -11
- data/lib/ferret/search/filtered_query.rb +0 -130
- data/lib/ferret/search/filtered_term_enum.rb +0 -79
- data/lib/ferret/search/fuzzy_query.rb +0 -154
- data/lib/ferret/search/fuzzy_term_enum.rb +0 -247
- data/lib/ferret/search/hit_collector.rb +0 -34
- data/lib/ferret/search/hit_queue.rb +0 -11
- data/lib/ferret/search/index_searcher.rb +0 -200
- data/lib/ferret/search/match_all_query.rb +0 -104
- data/lib/ferret/search/multi_phrase_query.rb +0 -216
- data/lib/ferret/search/multi_searcher.rb +0 -261
- data/lib/ferret/search/multi_term_query.rb +0 -65
- data/lib/ferret/search/non_matching_scorer.rb +0 -22
- data/lib/ferret/search/phrase_positions.rb +0 -55
- data/lib/ferret/search/phrase_query.rb +0 -214
- data/lib/ferret/search/phrase_scorer.rb +0 -152
- data/lib/ferret/search/prefix_query.rb +0 -54
- data/lib/ferret/search/query.rb +0 -140
- data/lib/ferret/search/query_filter.rb +0 -51
- data/lib/ferret/search/range_filter.rb +0 -103
- data/lib/ferret/search/range_query.rb +0 -139
- data/lib/ferret/search/req_excl_scorer.rb +0 -125
- data/lib/ferret/search/req_opt_sum_scorer.rb +0 -70
- data/lib/ferret/search/score_doc.rb +0 -38
- data/lib/ferret/search/score_doc_comparator.rb +0 -114
- data/lib/ferret/search/scorer.rb +0 -91
- data/lib/ferret/search/similarity.rb +0 -278
- data/lib/ferret/search/sloppy_phrase_scorer.rb +0 -47
- data/lib/ferret/search/sort.rb +0 -112
- data/lib/ferret/search/sort_comparator.rb +0 -60
- data/lib/ferret/search/sort_field.rb +0 -91
- data/lib/ferret/search/spans.rb +0 -12
- data/lib/ferret/search/spans/near_spans_enum.rb +0 -304
- data/lib/ferret/search/spans/span_first_query.rb +0 -79
- data/lib/ferret/search/spans/span_near_query.rb +0 -108
- data/lib/ferret/search/spans/span_not_query.rb +0 -130
- data/lib/ferret/search/spans/span_or_query.rb +0 -176
- data/lib/ferret/search/spans/span_query.rb +0 -25
- data/lib/ferret/search/spans/span_scorer.rb +0 -74
- data/lib/ferret/search/spans/span_term_query.rb +0 -105
- data/lib/ferret/search/spans/span_weight.rb +0 -84
- data/lib/ferret/search/spans/spans_enum.rb +0 -44
- data/lib/ferret/search/term_query.rb +0 -128
- data/lib/ferret/search/term_scorer.rb +0 -183
- data/lib/ferret/search/top_docs.rb +0 -36
- data/lib/ferret/search/top_field_docs.rb +0 -17
- data/lib/ferret/search/weight.rb +0 -54
- data/lib/ferret/search/wildcard_query.rb +0 -26
- data/lib/ferret/search/wildcard_term_enum.rb +0 -61
- data/lib/ferret/stemmers.rb +0 -1
- data/lib/ferret/stemmers/porter_stemmer.rb +0 -218
- data/lib/ferret/store.rb +0 -5
- data/lib/ferret/store/buffered_index_io.rb +0 -190
- data/lib/ferret/store/directory.rb +0 -141
- data/lib/ferret/store/fs_store.rb +0 -381
- data/lib/ferret/store/index_io.rb +0 -245
- data/lib/ferret/store/ram_store.rb +0 -286
- data/lib/ferret/utils.rb +0 -8
- data/lib/ferret/utils/bit_vector.rb +0 -123
- data/lib/ferret/utils/date_tools.rb +0 -138
- data/lib/ferret/utils/number_tools.rb +0 -91
- data/lib/ferret/utils/parameter.rb +0 -41
- data/lib/ferret/utils/priority_queue.rb +0 -120
- data/lib/ferret/utils/string_helper.rb +0 -47
- data/lib/ferret/utils/thread_local.rb +0 -28
- data/lib/ferret/utils/weak_key_hash.rb +0 -60
- data/lib/rferret.rb +0 -37
- data/rake_utils/code_statistics.rb +0 -106
- data/test/benchmark/tb_ram_store.rb +0 -76
- data/test/benchmark/tb_rw_vint.rb +0 -26
- data/test/functional/thread_safety_index_test.rb +0 -81
- data/test/functional/thread_safety_test.rb +0 -137
- data/test/longrunning/tc_numbertools.rb +0 -60
- data/test/longrunning/tm_store.rb +0 -19
- data/test/unit/analysis/ctc_analyzer.rb +0 -532
- data/test/unit/analysis/data/wordfile +0 -6
- data/test/unit/analysis/rtc_letter_tokenizer.rb +0 -20
- data/test/unit/analysis/rtc_lower_case_filter.rb +0 -20
- data/test/unit/analysis/rtc_lower_case_tokenizer.rb +0 -27
- data/test/unit/analysis/rtc_per_field_analyzer_wrapper.rb +0 -39
- data/test/unit/analysis/rtc_porter_stem_filter.rb +0 -16
- data/test/unit/analysis/rtc_standard_analyzer.rb +0 -20
- data/test/unit/analysis/rtc_standard_tokenizer.rb +0 -20
- data/test/unit/analysis/rtc_stop_analyzer.rb +0 -20
- data/test/unit/analysis/rtc_stop_filter.rb +0 -14
- data/test/unit/analysis/rtc_white_space_analyzer.rb +0 -21
- data/test/unit/analysis/rtc_white_space_tokenizer.rb +0 -20
- data/test/unit/analysis/rtc_word_list_loader.rb +0 -32
- data/test/unit/analysis/tc_token.rb +0 -25
- data/test/unit/document/rtc_field.rb +0 -28
- data/test/unit/document/tc_document.rb +0 -47
- data/test/unit/document/tc_field.rb +0 -98
- data/test/unit/index/rtc_compound_file_io.rb +0 -107
- data/test/unit/index/rtc_field_infos.rb +0 -127
- data/test/unit/index/rtc_fields_io.rb +0 -167
- data/test/unit/index/rtc_multiple_term_doc_pos_enum.rb +0 -83
- data/test/unit/index/rtc_segment_infos.rb +0 -74
- data/test/unit/index/rtc_segment_term_docs.rb +0 -17
- data/test/unit/index/rtc_segment_term_enum.rb +0 -60
- data/test/unit/index/rtc_segment_term_vector.rb +0 -71
- data/test/unit/index/rtc_term_buffer.rb +0 -57
- data/test/unit/index/rtc_term_info.rb +0 -19
- data/test/unit/index/rtc_term_infos_io.rb +0 -192
- data/test/unit/index/rtc_term_vectors_io.rb +0 -108
- data/test/unit/index/tc_term.rb +0 -27
- data/test/unit/index/tc_term_voi.rb +0 -18
- data/test/unit/search/rtc_similarity.rb +0 -37
- data/test/unit/search/rtc_sort_field.rb +0 -14
- data/test/unit/search/tc_multi_searcher2.rb +0 -126
- data/test/unit/store/rtc_fs_store.rb +0 -62
- data/test/unit/store/rtc_ram_store.rb +0 -15
- data/test/unit/store/rtm_store.rb +0 -150
- data/test/unit/store/rtm_store_lock.rb +0 -2
- data/test/unit/ts_document.rb +0 -2
- data/test/unit/utils/rtc_bit_vector.rb +0 -73
- data/test/unit/utils/rtc_date_tools.rb +0 -50
- data/test/unit/utils/rtc_number_tools.rb +0 -59
- data/test/unit/utils/rtc_parameter.rb +0 -40
- data/test/unit/utils/rtc_priority_queue.rb +0 -62
- data/test/unit/utils/rtc_string_helper.rb +0 -21
- data/test/unit/utils/rtc_thread.rb +0 -61
- data/test/unit/utils/rtc_weak_key_hash.rb +0 -25
- data/test/utils/number_to_spoken.rb +0 -132
data/ext/index_io.c
DELETED
@@ -1,382 +0,0 @@
|
|
1
|
-
#include <store.h>
|
2
|
-
#include <string.h>
|
3
|
-
|
4
|
-
static char * const STORE_EOF_ERROR_MSG = "EOF Error when trying to refill";
|
5
|
-
|
6
|
-
Buffer *buf_create()
|
7
|
-
{
|
8
|
-
Buffer *buf = ALLOC(Buffer);
|
9
|
-
buf->start = 0;
|
10
|
-
buf->pos = 0;
|
11
|
-
buf->len = 0;
|
12
|
-
return buf;
|
13
|
-
}
|
14
|
-
|
15
|
-
void buf_destroy(Buffer *buf)
|
16
|
-
{
|
17
|
-
free(buf);
|
18
|
-
}
|
19
|
-
|
20
|
-
OutStream *os_create()
|
21
|
-
{
|
22
|
-
OutStream *os = ALLOC(OutStream);
|
23
|
-
os->buf.start = 0;
|
24
|
-
os->buf.pos = 0;
|
25
|
-
os->buf.len = 0;
|
26
|
-
return os;
|
27
|
-
}
|
28
|
-
|
29
|
-
inline void os_flush(OutStream *os)
|
30
|
-
{
|
31
|
-
os->flush_internal(os, os->buf.buf, os->buf.pos);
|
32
|
-
os->buf.start += os->buf.pos;
|
33
|
-
os->buf.pos = 0;
|
34
|
-
}
|
35
|
-
|
36
|
-
void os_close(OutStream *os)
|
37
|
-
{
|
38
|
-
os_flush(os);
|
39
|
-
os->close_internal(os);
|
40
|
-
free(os);
|
41
|
-
}
|
42
|
-
|
43
|
-
int os_pos(OutStream *os)
|
44
|
-
{
|
45
|
-
return os->buf.start + os->buf.pos;
|
46
|
-
}
|
47
|
-
|
48
|
-
void os_seek(OutStream *os, int new_pos)
|
49
|
-
{
|
50
|
-
os_flush(os);
|
51
|
-
os->buf.start = new_pos;
|
52
|
-
os->seek_internal(os, new_pos);
|
53
|
-
}
|
54
|
-
|
55
|
-
#define write_byte(os, b) os->buf.buf[os->buf.pos++] = b
|
56
|
-
|
57
|
-
inline void os_write_byte(OutStream *os, uchar b)
|
58
|
-
{
|
59
|
-
if (os->buf.pos >= BUFFER_SIZE) {
|
60
|
-
os_flush(os);
|
61
|
-
}
|
62
|
-
write_byte(os, b);
|
63
|
-
}
|
64
|
-
|
65
|
-
void os_write_bytes(OutStream *os, uchar *b, int len)
|
66
|
-
{
|
67
|
-
if (os->buf.pos > 0) { /* flush buffer */
|
68
|
-
os_flush(os);
|
69
|
-
}
|
70
|
-
|
71
|
-
if (len < BUFFER_SIZE) {
|
72
|
-
os->flush_internal(os, b, len);
|
73
|
-
os->buf.start += len;
|
74
|
-
} else {
|
75
|
-
int pos = 0;
|
76
|
-
int size;
|
77
|
-
while (pos < len) {
|
78
|
-
if (len - pos < BUFFER_SIZE) {
|
79
|
-
size = len - pos;
|
80
|
-
} else {
|
81
|
-
size = BUFFER_SIZE;
|
82
|
-
}
|
83
|
-
os->flush_internal(os, b + pos, size);
|
84
|
-
pos += size;
|
85
|
-
os->buf.start += size;
|
86
|
-
}
|
87
|
-
}
|
88
|
-
}
|
89
|
-
|
90
|
-
InStream *is_create()
|
91
|
-
{
|
92
|
-
InStream *is = ALLOC(InStream);
|
93
|
-
is->buf.start = 0;
|
94
|
-
is->buf.pos = 0;
|
95
|
-
is->buf.len = 0;
|
96
|
-
return is;
|
97
|
-
}
|
98
|
-
|
99
|
-
void is_refill(InStream *is)
|
100
|
-
{
|
101
|
-
int start = is->buf.start + is->buf.pos;
|
102
|
-
int last = start + BUFFER_SIZE;
|
103
|
-
int flen = is->length_internal(is);
|
104
|
-
if (last > flen) { /* don't read past EOF */
|
105
|
-
last = flen;
|
106
|
-
}
|
107
|
-
|
108
|
-
is->buf.len = last - start;
|
109
|
-
if (is->buf.len <= 0) {
|
110
|
-
RAISE(EOF_ERROR, STORE_EOF_ERROR_MSG);
|
111
|
-
}
|
112
|
-
|
113
|
-
is->read_internal(is, is->buf.buf, 0, is->buf.len);
|
114
|
-
|
115
|
-
is->buf.start = start;
|
116
|
-
is->buf.pos = 0;
|
117
|
-
}
|
118
|
-
|
119
|
-
#define read_byte(is) is->buf.buf[is->buf.pos++]
|
120
|
-
inline uchar is_read_byte(InStream *is)
|
121
|
-
{
|
122
|
-
if (is->buf.pos >= is->buf.len) {
|
123
|
-
is_refill(is);
|
124
|
-
}
|
125
|
-
|
126
|
-
return read_byte(is);
|
127
|
-
}
|
128
|
-
|
129
|
-
int is_pos(InStream *is)
|
130
|
-
{
|
131
|
-
return is->buf.start + is->buf.pos;
|
132
|
-
}
|
133
|
-
|
134
|
-
uchar *is_read_bytes(InStream *is, uchar *b, int offset, int len)
|
135
|
-
{
|
136
|
-
int i, start;
|
137
|
-
if ((offset + len) < BUFFER_SIZE) {
|
138
|
-
for (i = offset; i < offset + len; i++) {
|
139
|
-
b[i] = is_read_byte(is);
|
140
|
-
}
|
141
|
-
} else { // read all-at-once
|
142
|
-
start = is_pos(is);
|
143
|
-
is->seek_internal(is, start);
|
144
|
-
is->read_internal(is, b, offset, len);
|
145
|
-
|
146
|
-
is->buf.start = start + len; // adjust stream variables
|
147
|
-
is->buf.pos = 0;
|
148
|
-
is->buf.len = 0; // trigger refill on read
|
149
|
-
}
|
150
|
-
return b;
|
151
|
-
}
|
152
|
-
|
153
|
-
void is_seek(InStream *is, int pos)
|
154
|
-
{
|
155
|
-
if (pos >= is->buf.start && pos < (is->buf.start + is->buf.len)) {
|
156
|
-
is->buf.pos = pos - is->buf.start; // seek within buffer
|
157
|
-
} else {
|
158
|
-
is->buf.start = pos;
|
159
|
-
is->buf.pos = 0;
|
160
|
-
is->buf.len = 0; // trigger refill() on read()
|
161
|
-
is->seek_internal(is, pos);
|
162
|
-
}
|
163
|
-
}
|
164
|
-
|
165
|
-
void is_close(InStream *is)
|
166
|
-
{
|
167
|
-
is->close_internal(is);
|
168
|
-
free(is);
|
169
|
-
}
|
170
|
-
|
171
|
-
InStream *is_clone(InStream *is)
|
172
|
-
{
|
173
|
-
InStream *new_index_i = ALLOC(InStream);
|
174
|
-
memcpy(new_index_i, is, sizeof(InStream));
|
175
|
-
new_index_i->is_clone = true;
|
176
|
-
is->clone_internal(is, new_index_i);
|
177
|
-
return new_index_i;
|
178
|
-
}
|
179
|
-
|
180
|
-
int
|
181
|
-
is_read_int(InStream *is)
|
182
|
-
{
|
183
|
-
return ((int)is_read_byte(is) << 24) |
|
184
|
-
((int)is_read_byte(is) << 16) |
|
185
|
-
((int)is_read_byte(is) << 8) |
|
186
|
-
(int)is_read_byte(is);
|
187
|
-
}
|
188
|
-
|
189
|
-
llong
|
190
|
-
is_read_long(InStream *is)
|
191
|
-
{
|
192
|
-
return ((llong)is_read_byte(is) << 56) |
|
193
|
-
((llong)is_read_byte(is) << 48) |
|
194
|
-
((llong)is_read_byte(is) << 40) |
|
195
|
-
((llong)is_read_byte(is) << 32) |
|
196
|
-
((llong)is_read_byte(is) << 24) |
|
197
|
-
((llong)is_read_byte(is) << 16) |
|
198
|
-
((llong)is_read_byte(is) << 8) |
|
199
|
-
(llong)is_read_byte(is);
|
200
|
-
}
|
201
|
-
|
202
|
-
unsigned int
|
203
|
-
is_read_uint(InStream *is)
|
204
|
-
{
|
205
|
-
return ((unsigned int)is_read_byte(is) << 24) |
|
206
|
-
((unsigned int)is_read_byte(is) << 16) |
|
207
|
-
((unsigned int)is_read_byte(is) << 8) |
|
208
|
-
(unsigned int)is_read_byte(is);
|
209
|
-
}
|
210
|
-
|
211
|
-
ullong
|
212
|
-
is_read_ulong(InStream *is)
|
213
|
-
{
|
214
|
-
return ((ullong)is_read_byte(is) << 56) |
|
215
|
-
((ullong)is_read_byte(is) << 48) |
|
216
|
-
((ullong)is_read_byte(is) << 40) |
|
217
|
-
((ullong)is_read_byte(is) << 32) |
|
218
|
-
((ullong)is_read_byte(is) << 24) |
|
219
|
-
((ullong)is_read_byte(is) << 16) |
|
220
|
-
((ullong)is_read_byte(is) << 8) |
|
221
|
-
(ullong)is_read_byte(is);
|
222
|
-
}
|
223
|
-
|
224
|
-
/* optimized to use unchecked read_byte if there is definitely space */
|
225
|
-
inline ullong
|
226
|
-
is_read_vint(InStream *is)
|
227
|
-
{
|
228
|
-
register ullong res, b;
|
229
|
-
register int shift = 7;
|
230
|
-
|
231
|
-
if (is->buf.pos > (is->buf.len - VINT_MAX_LEN)) {
|
232
|
-
b = is_read_byte(is);
|
233
|
-
res = b & 0x7F; // 0x7F = 0b01111111
|
234
|
-
|
235
|
-
while ((b & 0x80) != 0) {// 0x80 = 0b10000000
|
236
|
-
b = is_read_byte(is);
|
237
|
-
res |= (b & 0x7F) << shift;
|
238
|
-
shift += 7;
|
239
|
-
}
|
240
|
-
} else { // unchecked
|
241
|
-
b = read_byte(is);
|
242
|
-
res = b & 0x7F; // 0x7F = 0b01111111
|
243
|
-
|
244
|
-
while ((b & 0x80) != 0) {// 0x80 = 0b10000000
|
245
|
-
b = read_byte(is);
|
246
|
-
res |= (b & 0x7F) << shift;
|
247
|
-
shift += 7;
|
248
|
-
}
|
249
|
-
}
|
250
|
-
|
251
|
-
return res;
|
252
|
-
}
|
253
|
-
|
254
|
-
inline void
|
255
|
-
is_skip_vints(InStream *is, register int cnt)
|
256
|
-
{
|
257
|
-
for (; cnt > 0; cnt--) {
|
258
|
-
while ((is_read_byte(is) & 0x80) != 0) {
|
259
|
-
}
|
260
|
-
}
|
261
|
-
}
|
262
|
-
|
263
|
-
inline void
|
264
|
-
is_read_chars(InStream *is, char* buffer, int off, int len)
|
265
|
-
{
|
266
|
-
int end, i;
|
267
|
-
|
268
|
-
end = off + len;
|
269
|
-
|
270
|
-
for(i = off; i < end; i++) {
|
271
|
-
buffer[i] = is_read_byte(is);
|
272
|
-
}
|
273
|
-
}
|
274
|
-
|
275
|
-
char *
|
276
|
-
is_read_string(InStream *is)
|
277
|
-
{
|
278
|
-
register int length = (int)is_read_vint(is);
|
279
|
-
char *str = ALLOC_N(char, length + 1);
|
280
|
-
str[length] = '\0';
|
281
|
-
|
282
|
-
if (is->buf.pos > (is->buf.len - length)) {
|
283
|
-
register int i;
|
284
|
-
for(i = 0; i < length; i++) {
|
285
|
-
str[i] = is_read_byte(is);
|
286
|
-
}
|
287
|
-
} else { // unchecked
|
288
|
-
memcpy(str, is->buf.buf + is->buf.pos, length);
|
289
|
-
is->buf.pos += length;
|
290
|
-
}
|
291
|
-
//is_read_chars(is, str, 0, length);
|
292
|
-
|
293
|
-
return str;
|
294
|
-
}
|
295
|
-
|
296
|
-
void
|
297
|
-
os_write_int(OutStream *os, int l)
|
298
|
-
{
|
299
|
-
os_write_byte(os, (uchar)((l >> 24) & 0xFF));
|
300
|
-
os_write_byte(os, (uchar)((l >> 16) & 0xFF));
|
301
|
-
os_write_byte(os, (uchar)((l >> 8) & 0xFF));
|
302
|
-
os_write_byte(os, (uchar)(l & 0xFF));
|
303
|
-
}
|
304
|
-
|
305
|
-
void
|
306
|
-
os_write_long(OutStream *os, llong l)
|
307
|
-
{
|
308
|
-
os_write_byte(os, (uchar)((l >> 56) & 0xFF));
|
309
|
-
os_write_byte(os, (uchar)((l >> 48) & 0xFF));
|
310
|
-
os_write_byte(os, (uchar)((l >> 40) & 0xFF));
|
311
|
-
os_write_byte(os, (uchar)((l >> 32) & 0xFF));
|
312
|
-
os_write_byte(os, (uchar)((l >> 24) & 0xFF));
|
313
|
-
os_write_byte(os, (uchar)((l >> 16) & 0xFF));
|
314
|
-
os_write_byte(os, (uchar)((l >> 8) & 0xFF));
|
315
|
-
os_write_byte(os, (uchar)(l & 0xFF));
|
316
|
-
}
|
317
|
-
|
318
|
-
void
|
319
|
-
os_write_uint(OutStream *os, unsigned int l)
|
320
|
-
{
|
321
|
-
os_write_byte(os, (uchar)((l >> 24) & 0xFF));
|
322
|
-
os_write_byte(os, (uchar)((l >> 16) & 0xFF));
|
323
|
-
os_write_byte(os, (uchar)((l >> 8) & 0xFF));
|
324
|
-
os_write_byte(os, (uchar)(l & 0xFF));
|
325
|
-
}
|
326
|
-
|
327
|
-
void
|
328
|
-
os_write_ulong(OutStream *os, ullong l)
|
329
|
-
{
|
330
|
-
os_write_byte(os, (uchar)((l >> 56) & 0xFF));
|
331
|
-
os_write_byte(os, (uchar)((l >> 48) & 0xFF));
|
332
|
-
os_write_byte(os, (uchar)((l >> 40) & 0xFF));
|
333
|
-
os_write_byte(os, (uchar)((l >> 32) & 0xFF));
|
334
|
-
os_write_byte(os, (uchar)((l >> 24) & 0xFF));
|
335
|
-
os_write_byte(os, (uchar)((l >> 16) & 0xFF));
|
336
|
-
os_write_byte(os, (uchar)((l >> 8) & 0xFF));
|
337
|
-
os_write_byte(os, (uchar)(l & 0xFF));
|
338
|
-
}
|
339
|
-
|
340
|
-
/* optimized to use an unchecked write if there is space */
|
341
|
-
inline void
|
342
|
-
os_write_vint(OutStream *os, register ullong i)
|
343
|
-
{
|
344
|
-
if (os->buf.pos > VINT_END) {
|
345
|
-
while (i > 127) {
|
346
|
-
os_write_byte(os, (uchar)((i & 0x7f) | 0x80));
|
347
|
-
i >>= 7;
|
348
|
-
}
|
349
|
-
os_write_byte(os, (uchar)(i));
|
350
|
-
} else {
|
351
|
-
while (i > 127) {
|
352
|
-
write_byte(os, (uchar)((i & 0x7f) | 0x80));
|
353
|
-
i >>= 7;
|
354
|
-
}
|
355
|
-
write_byte(os, (uchar)(i));
|
356
|
-
}
|
357
|
-
}
|
358
|
-
|
359
|
-
void
|
360
|
-
os_write_chars(OutStream *os, char *buf, int start, int length)
|
361
|
-
{
|
362
|
-
int i;
|
363
|
-
|
364
|
-
for (i = start; i < start + length; i++) {
|
365
|
-
os_write_byte(os, buf[i]);
|
366
|
-
}
|
367
|
-
}
|
368
|
-
|
369
|
-
void
|
370
|
-
os_write_string(OutStream *os, char *str)
|
371
|
-
{
|
372
|
-
int len = (int)strlen(str);
|
373
|
-
os_write_vint(os, len);
|
374
|
-
|
375
|
-
os_write_chars(os, str, 0, len);
|
376
|
-
}
|
377
|
-
|
378
|
-
int file_is_lock(char *filename)
|
379
|
-
{
|
380
|
-
int start = (int)strlen(filename) - 4;
|
381
|
-
return ((start > 0) && (strcmp(".lck", &filename[start]) == 0));
|
382
|
-
}
|
data/ext/index_rw.c
DELETED
@@ -1,2658 +0,0 @@
|
|
1
|
-
#include "index.h"
|
2
|
-
#include <stdlib.h>
|
3
|
-
#include <string.h>
|
4
|
-
#include <array.h>
|
5
|
-
static char * const FORMAT_VERSION_ERROR_MSG = "Unknown format version";
|
6
|
-
static char * const WRITE_LOCK_ERROR_MSG = "Could not obtain write lock when trying to write index";
|
7
|
-
static char * const COMMIT_LOCK_ERROR_MSG = "Could not obtain commit lock when trying to write index";
|
8
|
-
static char * const DELETED_DOC_ERROR_MSG = "Tried to get doc that has already been deleted";
|
9
|
-
static char * const INVALID_FIELD_TYPE_MSG = "Invalid field-type";
|
10
|
-
static char * const DOC_ORDER_ERROR_MSG = "docs out of order curent";
|
11
|
-
static char * const STALE_READER_ERROR_MSG = "IndexReader out of date and no longer valid for delete, undelete, or set_norm operations";
|
12
|
-
|
13
|
-
const char *INDEX_EXTENSIONS[] = {
|
14
|
-
"cfs", "fnm", "fdx", "fdt", "tii", "tis", "frq", "prx", "del",
|
15
|
-
"tvx", "tvd", "tvf", "tvp"
|
16
|
-
};
|
17
|
-
|
18
|
-
const char *COMPOUND_EXTENSIONS[] = {
|
19
|
-
"fnm", "frq", "prx", "fdx", "fdt", "tii", "tis"
|
20
|
-
};
|
21
|
-
|
22
|
-
const char *VECTOR_EXTENSIONS[] = {
|
23
|
-
"tvx", "tvd", "tvf"
|
24
|
-
};
|
25
|
-
|
26
|
-
FerretConfig config = {
|
27
|
-
10, /* default merge_factor */
|
28
|
-
10, /* default min_merge_docs */
|
29
|
-
INT_MAX, /* default max_merge_docs */
|
30
|
-
10000, /* default max_field_length */
|
31
|
-
128 /* default term_index_interval */
|
32
|
-
};
|
33
|
-
|
34
|
-
/***************************************************************************
|
35
|
-
*
|
36
|
-
* CacheObject
|
37
|
-
*
|
38
|
-
***************************************************************************/
|
39
|
-
|
40
|
-
unsigned int co_hash(const void *key)
|
41
|
-
{
|
42
|
-
return (unsigned int)key;
|
43
|
-
}
|
44
|
-
|
45
|
-
int co_eq(const void *key1, const void *key2)
|
46
|
-
{
|
47
|
-
return (key1 == key2);
|
48
|
-
}
|
49
|
-
|
50
|
-
void co_destroy(CacheObject *self)
|
51
|
-
{
|
52
|
-
h_rem(self->ref_tab1, self->ref2, false);
|
53
|
-
h_rem(self->ref_tab2, self->ref1, false);
|
54
|
-
self->destroy(self->obj);
|
55
|
-
free(self);
|
56
|
-
}
|
57
|
-
|
58
|
-
CacheObject *co_create(HshTable *ref_tab1, HshTable *ref_tab2,
|
59
|
-
void *ref1, void *ref2, free_ft destroy, void *obj)
|
60
|
-
{
|
61
|
-
CacheObject *self = ALLOC(CacheObject);
|
62
|
-
h_set(ref_tab1, ref2, self);
|
63
|
-
h_set(ref_tab2, ref1, self);
|
64
|
-
self->ref_tab1 = ref_tab1;
|
65
|
-
self->ref_tab2 = ref_tab2;
|
66
|
-
self->ref1 = ref1;
|
67
|
-
self->ref2 = ref2;
|
68
|
-
self->destroy = destroy;
|
69
|
-
self->obj = obj;
|
70
|
-
return self;
|
71
|
-
}
|
72
|
-
|
73
|
-
HshTable *co_hsh_create()
|
74
|
-
{
|
75
|
-
return h_new(&co_hash, &co_eq, (free_ft)NULL, (free_ft)&co_destroy);
|
76
|
-
}
|
77
|
-
|
78
|
-
/***************************************************************************
|
79
|
-
*
|
80
|
-
* Posting
|
81
|
-
*
|
82
|
-
***************************************************************************/
|
83
|
-
|
84
|
-
Posting *p_create(Term *term, int position, TVOffsetInfo *offset)
|
85
|
-
{
|
86
|
-
Posting *self = ALLOC(Posting);
|
87
|
-
self->freq = 1;
|
88
|
-
self->size = 1;
|
89
|
-
self->term = term;
|
90
|
-
self->positions = ALLOC(int);
|
91
|
-
self->positions[0] = position;
|
92
|
-
self->offsets = ALLOC(TVOffsetInfo *);
|
93
|
-
self->offsets[0] = offset;
|
94
|
-
return self;
|
95
|
-
}
|
96
|
-
|
97
|
-
void p_destroy(Posting *self)
|
98
|
-
{
|
99
|
-
/* the positions and offsets will be put in a TVTerm so no need to free */
|
100
|
-
int i;
|
101
|
-
free(self->positions);
|
102
|
-
for (i = 0; i < self->freq; i++)
|
103
|
-
tvoi_destroy(self->offsets[i]);
|
104
|
-
free(self->offsets);
|
105
|
-
free(self);
|
106
|
-
}
|
107
|
-
|
108
|
-
void p_add_occurance(Posting *self, int position, TVOffsetInfo *offset)
|
109
|
-
{
|
110
|
-
if (self->freq >= self->size) {
|
111
|
-
self->size *= 2;
|
112
|
-
REALLOC_N(self->positions, int, self->size);
|
113
|
-
REALLOC_N(self->offsets, TVOffsetInfo *, self->size);
|
114
|
-
}
|
115
|
-
self->positions[self->freq] = position;
|
116
|
-
self->offsets[self->freq] = offset;
|
117
|
-
self->freq++;
|
118
|
-
}
|
119
|
-
|
120
|
-
inline int p_cmp(const void *const p1, const void *const p2)
|
121
|
-
{
|
122
|
-
Term *t1 = (*(Posting **)p1)->term;
|
123
|
-
Term *t2 = (*(Posting **)p2)->term;
|
124
|
-
int res = strcmp(t1->field, t2->field);
|
125
|
-
if (res != 0) {
|
126
|
-
return res;
|
127
|
-
} else {
|
128
|
-
return strcmp(t1->text, t2->text);
|
129
|
-
}
|
130
|
-
}
|
131
|
-
|
132
|
-
DocumentWriter *dw_open(Store *store,
|
133
|
-
Analyzer *analyzer,
|
134
|
-
Similarity *similarity,
|
135
|
-
int max_field_length,
|
136
|
-
int term_index_interval)
|
137
|
-
{
|
138
|
-
DocumentWriter *self = ALLOC(DocumentWriter);
|
139
|
-
self->store = store;
|
140
|
-
self->analyzer = analyzer;
|
141
|
-
self->similarity = similarity;
|
142
|
-
self->fis = NULL;
|
143
|
-
self->postingtable = h_new(&term_hash, &term_eq,
|
144
|
-
(free_ft)&term_destroy,
|
145
|
-
(free_ft)&p_destroy);
|
146
|
-
self->max_field_length = max_field_length;
|
147
|
-
self->term_index_interval = term_index_interval;
|
148
|
-
return self;
|
149
|
-
}
|
150
|
-
|
151
|
-
void dw_close(DocumentWriter *self)
|
152
|
-
{
|
153
|
-
if (self->fis) fis_destroy(self->fis);
|
154
|
-
h_destroy(self->postingtable);
|
155
|
-
free(self);
|
156
|
-
}
|
157
|
-
|
158
|
-
void dw_add_position(DocumentWriter *self, char *field, char *text,
|
159
|
-
int position, TVOffsetInfo *offset)
|
160
|
-
{
|
161
|
-
Term termbuf = {field, text}, *term;
|
162
|
-
Posting *p = (Posting *)h_get(self->postingtable, &termbuf);
|
163
|
-
|
164
|
-
if (p) { /* word seen before */
|
165
|
-
if (p->freq >= p->size) {
|
166
|
-
/* double size of posting to make room for more posts. */
|
167
|
-
p->size <<= 1;
|
168
|
-
REALLOC_N(p->positions, int, p->size);
|
169
|
-
p->offsets = REALLOC_N(p->offsets, TVOffsetInfo *, p->size);
|
170
|
-
}
|
171
|
-
p->positions[p->freq] = position; /* add new position */
|
172
|
-
p->offsets[p->freq] = offset; /* add new offset */
|
173
|
-
p->freq++; /* update frequency */
|
174
|
-
} else { /* word not seen before */
|
175
|
-
term = term_create(field, text);
|
176
|
-
h_set(self->postingtable, term, p_create(term, position, offset));
|
177
|
-
}
|
178
|
-
}
|
179
|
-
|
180
|
-
void dw_invert_doc(DocumentWriter *self, Document *doc)
|
181
|
-
{
|
182
|
-
int i;
|
183
|
-
int dfcnt = doc->dfcnt;
|
184
|
-
char *field_name, *text;
|
185
|
-
int field_number, length, position, offset, slen;
|
186
|
-
TokenStream *stream;
|
187
|
-
Token *token;
|
188
|
-
FieldInfo *fi;
|
189
|
-
char text_buf[MAX_WORD_SIZE];
|
190
|
-
text_buf[MAX_WORD_SIZE - 1] = '\0';
|
191
|
-
|
192
|
-
DocField **fields = doc->df_arr, *field;
|
193
|
-
for (i = 0; i < dfcnt; i++) {
|
194
|
-
field = fields[i];
|
195
|
-
field_name = field->name;
|
196
|
-
fi = ((FieldInfo *)ht_get(self->fis->by_name, field_name));
|
197
|
-
field_number = fi->number;
|
198
|
-
|
199
|
-
length = self->field_lengths[field_number];
|
200
|
-
offset = self->field_offsets[field_number];
|
201
|
-
position = self->field_positions[field_number];
|
202
|
-
|
203
|
-
if (fi->is_indexed) {
|
204
|
-
if (!field->is_tokenized) { /* un-tokenized field */
|
205
|
-
text = field->data;
|
206
|
-
slen = (int)strlen(text);
|
207
|
-
if (slen >= MAX_WORD_SIZE) {
|
208
|
-
slen = MAX_WORD_SIZE - 1;
|
209
|
-
text = strncpy(text_buf, text, MAX_WORD_SIZE - 1);
|
210
|
-
}
|
211
|
-
if (fi->store_offset) {
|
212
|
-
dw_add_position(self, field_name, text, position,
|
213
|
-
tvoi_create(offset, offset+slen));
|
214
|
-
} else {
|
215
|
-
dw_add_position(self, field_name, text, position, NULL);
|
216
|
-
}
|
217
|
-
offset += slen;
|
218
|
-
length++;
|
219
|
-
} else {
|
220
|
-
|
221
|
-
/* Tokenize field and add to posting_table */
|
222
|
-
stream = a_get_ts(self->analyzer, field_name, field->data);
|
223
|
-
|
224
|
-
while ((token = ts_next(stream)) != NULL) {
|
225
|
-
position += (token->pos_inc - 1);
|
226
|
-
|
227
|
-
if (fi->store_offset) {
|
228
|
-
dw_add_position(self,
|
229
|
-
field_name,
|
230
|
-
token->text,
|
231
|
-
position,
|
232
|
-
tvoi_create(offset + token->start, offset + token->end));
|
233
|
-
position++;
|
234
|
-
} else {
|
235
|
-
dw_add_position(self, field_name, token->text, position, NULL);
|
236
|
-
position++;
|
237
|
-
}
|
238
|
-
|
239
|
-
length++;
|
240
|
-
/* stop if we reach the max field length */
|
241
|
-
if (length > self->max_field_length) {
|
242
|
-
break;
|
243
|
-
}
|
244
|
-
}
|
245
|
-
|
246
|
-
if (token) {
|
247
|
-
offset += token->end + 1;
|
248
|
-
}
|
249
|
-
}
|
250
|
-
self->field_lengths[field_number] = length;
|
251
|
-
self->field_offsets[field_number] = offset;
|
252
|
-
self->field_positions[field_number] = position;
|
253
|
-
self->field_boosts[field_number] *= field->boost;
|
254
|
-
}
|
255
|
-
}
|
256
|
-
}
|
257
|
-
|
258
|
-
Posting **dw_sort_posting_table(DocumentWriter *self)
|
259
|
-
{
|
260
|
-
HshTable *ht = self->postingtable;
|
261
|
-
HshEntry *he = ht->table;
|
262
|
-
Posting **postings;
|
263
|
-
int i;
|
264
|
-
|
265
|
-
self->pcnt = i = ht->used;
|
266
|
-
postings = ALLOC_N(Posting *, i);
|
267
|
-
|
268
|
-
while (i > 0) {
|
269
|
-
if (he->value != NULL) {
|
270
|
-
i--;
|
271
|
-
postings[i] = (Posting *)he->value;
|
272
|
-
}
|
273
|
-
he++;
|
274
|
-
}
|
275
|
-
qsort(postings, self->pcnt, sizeof(Posting *), &p_cmp);
|
276
|
-
return postings;
|
277
|
-
}
|
278
|
-
|
279
|
-
void dw_write_postings(DocumentWriter *self, Posting **postings, char *segment)
|
280
|
-
{
|
281
|
-
OutStream * volatile freq_out = NULL, * volatile prox_out = NULL;
|
282
|
-
TermInfosWriter * volatile tiw = NULL;
|
283
|
-
TermVectorsWriter * volatile tvw = NULL;
|
284
|
-
Store *store = self->store;
|
285
|
-
TermInfo * volatile ti = NULL;
|
286
|
-
Posting *posting;
|
287
|
-
int i, j, posting_freq, position, last_position;
|
288
|
-
char fname[SEGMENT_NAME_MAX_LENGTH], *curr_field = NULL, *term_field;
|
289
|
-
strcpy(fname, segment);
|
290
|
-
|
291
|
-
TRY
|
292
|
-
/* open files for inverse index storage */
|
293
|
-
sprintf(fname, "%s.frq", segment);
|
294
|
-
freq_out = store->create_output(store, fname);
|
295
|
-
sprintf(fname, "%s.prx", segment);
|
296
|
-
prox_out = store->create_output(store, fname);
|
297
|
-
tiw = tiw_open(store, segment, self->fis, self->term_index_interval);
|
298
|
-
ti = ti_create(0, 0, 0, 0);
|
299
|
-
|
300
|
-
for (i = 0; i < self->pcnt; i++) {
|
301
|
-
posting = postings[i];
|
302
|
-
|
303
|
-
/* add an entry to dictionary with pointers to prox and freq_out files */
|
304
|
-
ti_set(ti, 1, os_pos(freq_out), os_pos(prox_out), -1);
|
305
|
-
tiw_add(tiw, posting->term, ti);
|
306
|
-
|
307
|
-
/* add an entry to the freq_out file */
|
308
|
-
posting_freq = posting->freq;
|
309
|
-
if (posting_freq == 1) { /* optimize freq=1 */
|
310
|
-
os_write_vint(freq_out, 1); /* set low bit of doc num */
|
311
|
-
} else {
|
312
|
-
os_write_vint(freq_out, 0); /* the doc number */
|
313
|
-
os_write_vint(freq_out, posting_freq); /* frequency in doc */
|
314
|
-
}
|
315
|
-
|
316
|
-
last_position = 0; /* write positions */
|
317
|
-
|
318
|
-
for (j = 0; j < posting_freq; j++) {
|
319
|
-
position = posting->positions[j];
|
320
|
-
os_write_vint(prox_out, position - last_position);
|
321
|
-
last_position = position;
|
322
|
-
}
|
323
|
-
|
324
|
-
/* check to see if we switched to a new field */
|
325
|
-
term_field = posting->term->field;
|
326
|
-
if (curr_field != term_field) {
|
327
|
-
FieldInfo *fi;
|
328
|
-
/* changing field - see if there is something to save */
|
329
|
-
curr_field = term_field;
|
330
|
-
fi = (FieldInfo *)ht_get(self->fis->by_name, curr_field);
|
331
|
-
if (fi->store_tv) {
|
332
|
-
if (tvw == NULL) {
|
333
|
-
tvw = tvw_open(store, segment, self->fis);
|
334
|
-
tvw_open_doc(tvw);
|
335
|
-
}
|
336
|
-
tvw_open_field(tvw, curr_field);
|
337
|
-
|
338
|
-
} else if (tvw != NULL) {
|
339
|
-
tvw_close_field(tvw);
|
340
|
-
}
|
341
|
-
}
|
342
|
-
/* tvw->curr_field != NULL implies field is still open */
|
343
|
-
if (tvw != NULL && tvw->curr_field != NULL) {
|
344
|
-
tvw_add_term(tvw, posting->term->text, posting_freq, posting->positions, posting->offsets);
|
345
|
-
}
|
346
|
-
}
|
347
|
-
XFINALLY
|
348
|
-
if (tvw) {
|
349
|
-
tvw_close_doc(tvw);
|
350
|
-
tvw_close(tvw);
|
351
|
-
}
|
352
|
-
/* make an effort to close all streams we can but remember and re-raise
|
353
|
-
* the last exception encountered in this process */
|
354
|
-
if (freq_out) os_close(freq_out);
|
355
|
-
if (prox_out) os_close(prox_out);
|
356
|
-
if (tiw) tiw_close(tiw);
|
357
|
-
if (ti) ti_destroy(ti);
|
358
|
-
XENDTRY
|
359
|
-
}
|
360
|
-
|
361
|
-
void dw_write_norms(DocumentWriter *self, char *segment)
|
362
|
-
{
|
363
|
-
int i;
|
364
|
-
float norm;
|
365
|
-
OutStream *norms_out;
|
366
|
-
char fname[SEGMENT_NAME_MAX_LENGTH];
|
367
|
-
FieldInfos *fis = self->fis;
|
368
|
-
FieldInfo *fi;
|
369
|
-
|
370
|
-
for (i = 0; i < fis->fcnt; i++) {
|
371
|
-
fi = fis->by_number[i];
|
372
|
-
|
373
|
-
if (fi->is_indexed && !fi->omit_norms) {
|
374
|
-
norm = self->field_boosts[i] *
|
375
|
-
sim_length_norm(self->similarity, fi->name, self->field_lengths[i]);
|
376
|
-
sprintf(fname, "%s.f%d", segment, i);
|
377
|
-
norms_out = self->store->create_output(self->store, fname);
|
378
|
-
TRY
|
379
|
-
os_write_byte(norms_out, sim_encode_norm(self->similarity, norm));
|
380
|
-
XFINALLY
|
381
|
-
os_close(norms_out);
|
382
|
-
XENDTRY
|
383
|
-
}
|
384
|
-
}
|
385
|
-
}
|
386
|
-
|
387
|
-
void dw_add_doc(DocumentWriter *self, char *segment, Document *doc)
|
388
|
-
{
|
389
|
-
Posting **postings;
|
390
|
-
FieldsWriter *fw;
|
391
|
-
int i;
|
392
|
-
|
393
|
-
/* write field names */
|
394
|
-
self->fis = fis_create();
|
395
|
-
fis_add_doc(self->fis, doc);
|
396
|
-
fis_write(self->fis, self->store, segment, ".fnm");
|
397
|
-
|
398
|
-
/* write field values */
|
399
|
-
fw = fw_open(self->store, segment, self->fis);
|
400
|
-
TRY
|
401
|
-
fw_add_doc(fw, doc);
|
402
|
-
XFINALLY
|
403
|
-
fw_close(fw);
|
404
|
-
XENDTRY
|
405
|
-
|
406
|
-
/* invert doc into posting_table */
|
407
|
-
|
408
|
-
h_clear(self->postingtable); /* clear posting_table */
|
409
|
-
|
410
|
-
self->field_boosts = ALLOC_N(float, self->fis->fcnt);
|
411
|
-
self->field_lengths = ALLOC_AND_ZERO_N(int, self->fis->fcnt);
|
412
|
-
self->field_offsets = ALLOC_AND_ZERO_N(int, self->fis->fcnt);
|
413
|
-
self->field_positions = ALLOC_AND_ZERO_N(int, self->fis->fcnt);
|
414
|
-
|
415
|
-
for (i = 0; i < self->fis->fcnt; i++) {
|
416
|
-
self->field_boosts[i] = doc->boost;
|
417
|
-
}
|
418
|
-
|
419
|
-
dw_invert_doc(self, doc);
|
420
|
-
|
421
|
-
/* sort posting_table into an array */
|
422
|
-
postings = dw_sort_posting_table(self);
|
423
|
-
|
424
|
-
/* write postings */
|
425
|
-
dw_write_postings(self, postings, segment);
|
426
|
-
free(postings);
|
427
|
-
|
428
|
-
/* write norms of indexed fields */
|
429
|
-
dw_write_norms(self, segment);
|
430
|
-
|
431
|
-
free(self->field_boosts);
|
432
|
-
free(self->field_lengths);
|
433
|
-
free(self->field_offsets);
|
434
|
-
free(self->field_positions);
|
435
|
-
}
|
436
|
-
|
437
|
-
/****************************************************************************
|
438
|
-
*
|
439
|
-
* SegmentInfo
|
440
|
-
*
|
441
|
-
****************************************************************************/
|
442
|
-
|
443
|
-
SegmentInfo *si_create(char *name, int doc_cnt, Store *store)
|
444
|
-
{
|
445
|
-
SegmentInfo *si = ALLOC(SegmentInfo);
|
446
|
-
si->name = name;
|
447
|
-
si->doc_cnt = doc_cnt;
|
448
|
-
si->store = store;
|
449
|
-
return si;
|
450
|
-
}
|
451
|
-
|
452
|
-
void si_destroy(SegmentInfo *si)
|
453
|
-
{
|
454
|
-
free(si->name);
|
455
|
-
free(si);
|
456
|
-
}
|
457
|
-
|
458
|
-
bool si_has_deletions(SegmentInfo *si)
|
459
|
-
{
|
460
|
-
char del_file_name[SEGMENT_NAME_MAX_LENGTH];
|
461
|
-
sprintf(del_file_name, "%s.del", si->name);
|
462
|
-
return si->store->exists(si->store, del_file_name);
|
463
|
-
}
|
464
|
-
|
465
|
-
bool si_uses_compound_file(SegmentInfo *si)
|
466
|
-
{
|
467
|
-
char compound_file_name[SEGMENT_NAME_MAX_LENGTH];
|
468
|
-
sprintf(compound_file_name, "%s.cfs", si->name);
|
469
|
-
return si->store->exists(si->store, compound_file_name);
|
470
|
-
}
|
471
|
-
|
472
|
-
struct NormTester {
|
473
|
-
bool has_norm_file;
|
474
|
-
char *segment_name;
|
475
|
-
};
|
476
|
-
void is_norm_file(char *fname, void *arg)
|
477
|
-
{
|
478
|
-
struct NormTester *nt = (struct NormTester *)arg;
|
479
|
-
char norm_file_pattern[SEGMENT_NAME_MAX_LENGTH];
|
480
|
-
sprintf(norm_file_pattern, "%s.s", nt->segment_name);
|
481
|
-
if (strncmp(fname, norm_file_pattern, strlen(norm_file_pattern)) == 0) {
|
482
|
-
nt->has_norm_file = true;
|
483
|
-
}
|
484
|
-
}
|
485
|
-
|
486
|
-
bool si_has_separate_norms(SegmentInfo *si)
|
487
|
-
{
|
488
|
-
struct NormTester nt;
|
489
|
-
nt.segment_name = si->name;
|
490
|
-
nt.has_norm_file = false;
|
491
|
-
si->store->each(si->store, &is_norm_file, &nt);
|
492
|
-
|
493
|
-
return nt.has_norm_file;
|
494
|
-
}
|
495
|
-
|
496
|
-
|
497
|
-
/****************************************************************************
|
498
|
-
*
|
499
|
-
* SegmentInfos
|
500
|
-
*
|
501
|
-
****************************************************************************/
|
502
|
-
|
503
|
-
#include <time.h>
|
504
|
-
#define FORMAT -1
|
505
|
-
#define SEGMENT_FILENAME "segments"
|
506
|
-
#define TEMPORARY_SEGMENT_FILENAME "segments.new"
|
507
|
-
|
508
|
-
SegmentInfos *sis_create()
|
509
|
-
{
|
510
|
-
SegmentInfos *sis = ALLOC(SegmentInfos);
|
511
|
-
sis->format = FORMAT;
|
512
|
-
sis->version = (unsigned int)time(NULL);
|
513
|
-
sis->scnt = 0;
|
514
|
-
sis->counter = 0;
|
515
|
-
sis->size = 4;
|
516
|
-
sis->segs = ALLOC_N(SegmentInfo *, sis->size);
|
517
|
-
return sis;
|
518
|
-
}
|
519
|
-
|
520
|
-
void sis_destroy_not_infos(SegmentInfos *sis)
|
521
|
-
{
|
522
|
-
free(sis->segs);
|
523
|
-
free(sis);
|
524
|
-
}
|
525
|
-
|
526
|
-
void sis_destroy(SegmentInfos *sis)
|
527
|
-
{
|
528
|
-
int i;
|
529
|
-
for (i = 0; i < sis->scnt; i++)
|
530
|
-
si_destroy(sis->segs[i]);
|
531
|
-
free(sis->segs);
|
532
|
-
free(sis);
|
533
|
-
}
|
534
|
-
|
535
|
-
void sis_add_si(SegmentInfos *sis, SegmentInfo *si)
|
536
|
-
{
|
537
|
-
if (sis->scnt >= sis->size) {
|
538
|
-
sis->size = sis->scnt * 2;
|
539
|
-
REALLOC_N(sis->segs, SegmentInfo *, sis->size);
|
540
|
-
}
|
541
|
-
sis->segs[sis->scnt] = si;
|
542
|
-
sis->scnt++;
|
543
|
-
}
|
544
|
-
|
545
|
-
void sis_del_at(SegmentInfos *sis, int at)
|
546
|
-
{
|
547
|
-
int i;
|
548
|
-
si_destroy(sis->segs[at]);
|
549
|
-
sis->scnt--;
|
550
|
-
for (i = at; i < sis->scnt; i++) {
|
551
|
-
sis->segs[i] = sis->segs[i+1];
|
552
|
-
}
|
553
|
-
}
|
554
|
-
|
555
|
-
void sis_del_from_to(SegmentInfos *sis, int from, int to)
|
556
|
-
{
|
557
|
-
int i, num_to_del = to - from;
|
558
|
-
sis->scnt -= num_to_del;
|
559
|
-
for (i = from; i < to; i++) {
|
560
|
-
si_destroy(sis->segs[i]);
|
561
|
-
}
|
562
|
-
for (i = from; i < sis->scnt; i++) {
|
563
|
-
sis->segs[i] = sis->segs[i+num_to_del];
|
564
|
-
}
|
565
|
-
}
|
566
|
-
|
567
|
-
void sis_clear(SegmentInfos *sis)
|
568
|
-
{
|
569
|
-
int i;
|
570
|
-
for (i = 0; i < sis->scnt; i++) {
|
571
|
-
si_destroy(sis->segs[i]);
|
572
|
-
}
|
573
|
-
sis->scnt = 0;
|
574
|
-
}
|
575
|
-
|
576
|
-
void sis_read(SegmentInfos *sis, Store *store)
|
577
|
-
{
|
578
|
-
int doc_cnt;
|
579
|
-
int seg_count;
|
580
|
-
int i;
|
581
|
-
char *name;
|
582
|
-
InStream *is = store->open_input(store, SEGMENT_FILENAME);
|
583
|
-
|
584
|
-
TRY
|
585
|
-
|
586
|
-
sis->format = is_read_int(is);
|
587
|
-
if (sis->format < 0) { /* file contains explicit format info */
|
588
|
-
/* check that it is a format we can understand */
|
589
|
-
if (sis->format < FORMAT)
|
590
|
-
RAISE(EXCEPTION, FORMAT_VERSION_ERROR_MSG);
|
591
|
-
sis->version = (uint)is_read_long(is);
|
592
|
-
sis->counter = (int)is_read_int(is);
|
593
|
-
} else { /* file is in old format without explicit format info */
|
594
|
-
sis->counter = sis->format;
|
595
|
-
}
|
596
|
-
|
597
|
-
seg_count = is_read_int(is);
|
598
|
-
for (i = 0; i < seg_count; i++) {
|
599
|
-
name = is_read_string(is);
|
600
|
-
doc_cnt = is_read_int(is);
|
601
|
-
sis_add_si(sis, si_create(name, doc_cnt, store));
|
602
|
-
}
|
603
|
-
|
604
|
-
if (sis->format >= 0) {
|
605
|
-
/* in old format the version number may be at the end of the file */
|
606
|
-
if (is_pos(is) >= is_length(is)) {
|
607
|
-
sis->version = 0; /* old file format without version number */
|
608
|
-
} else {
|
609
|
-
sis->version = (int)is_read_long(is); /* read version */
|
610
|
-
}
|
611
|
-
}
|
612
|
-
XFINALLY
|
613
|
-
is_close(is);
|
614
|
-
XENDTRY
|
615
|
-
}
|
616
|
-
|
617
|
-
void sis_write(SegmentInfos *sis, Store *store)
|
618
|
-
{
|
619
|
-
int i;
|
620
|
-
SegmentInfo *si;
|
621
|
-
OutStream *os = store->create_output(store, TEMPORARY_SEGMENT_FILENAME);
|
622
|
-
TRY
|
623
|
-
os_write_int(os, FORMAT);
|
624
|
-
os_write_long(os, ++(sis->version)); /* every write changes the index */
|
625
|
-
os_write_int(os, sis->counter);
|
626
|
-
os_write_int(os, sis->scnt);
|
627
|
-
for (i = 0; i < sis->scnt; i++) {
|
628
|
-
si = sis->segs[i];
|
629
|
-
os_write_string(os, si->name);
|
630
|
-
os_write_int(os, si->doc_cnt);
|
631
|
-
}
|
632
|
-
|
633
|
-
XFINALLY
|
634
|
-
os_close(os);
|
635
|
-
XENDTRY
|
636
|
-
|
637
|
-
/* install new segment info */
|
638
|
-
store->rename(store, TEMPORARY_SEGMENT_FILENAME, SEGMENT_FILENAME);
|
639
|
-
}
|
640
|
-
|
641
|
-
int sis_read_current_version(Store *store)
|
642
|
-
{
|
643
|
-
InStream *is;
|
644
|
-
SegmentInfos *sis;
|
645
|
-
int format = 0;
|
646
|
-
int version = 0;
|
647
|
-
|
648
|
-
if (!store->exists(store, SEGMENT_FILENAME))
|
649
|
-
return 0;
|
650
|
-
is = store->open_input(store, SEGMENT_FILENAME);
|
651
|
-
|
652
|
-
TRY
|
653
|
-
format = is_read_int(is);
|
654
|
-
if (format < 0) {
|
655
|
-
if (format < FORMAT)
|
656
|
-
RAISE(EXCEPTION, FORMAT_VERSION_ERROR_MSG);
|
657
|
-
version = (int)is_read_long(is);
|
658
|
-
}
|
659
|
-
XFINALLY
|
660
|
-
is_close(is);
|
661
|
-
XENDTRY
|
662
|
-
|
663
|
-
if (format < 0)
|
664
|
-
return version;
|
665
|
-
|
666
|
-
/* We cannot be sure about the format of the file.
|
667
|
-
* Therefore we have to read the whole file and cannot simply
|
668
|
-
* seek to the version entry. */
|
669
|
-
|
670
|
-
sis = sis_create();
|
671
|
-
sis_read(sis, store);
|
672
|
-
version = sis->version;
|
673
|
-
sis_destroy(sis);
|
674
|
-
return version;
|
675
|
-
}
|
676
|
-
|
677
|
-
/****************************************************************************
|
678
|
-
*
|
679
|
-
* IndexWriter
|
680
|
-
*
|
681
|
-
****************************************************************************/
|
682
|
-
|
683
|
-
/**
|
684
|
-
* Deletes the analyzer by default but leaves the store by default
|
685
|
-
*/
|
686
|
-
IndexWriter *iw_open(Store *store, Analyzer *analyzer, bool create)
|
687
|
-
{
|
688
|
-
IndexWriter *iw = ALLOC(IndexWriter);
|
689
|
-
if (create)
|
690
|
-
store->clear_all(store);
|
691
|
-
mutex_init(&iw->mutex, NULL);
|
692
|
-
iw->merge_factor = config.merge_factor;
|
693
|
-
iw->min_merge_docs = config.min_merge_docs;
|
694
|
-
iw->max_merge_docs = config.max_merge_docs;
|
695
|
-
iw->max_field_length = config.max_field_length;
|
696
|
-
iw->term_index_interval = config.term_index_interval;
|
697
|
-
iw->use_compound_file = true;
|
698
|
-
iw->store = store;
|
699
|
-
ref(store);
|
700
|
-
iw->analyzer = analyzer;
|
701
|
-
iw->sis = sis_create();
|
702
|
-
iw->similarity = sim_create_default();
|
703
|
-
iw->ram_store = open_ram_store();
|
704
|
-
|
705
|
-
mutex_lock(&store->mutex);
|
706
|
-
/* keep the write_lock obtained until the IndexWriter is closed. */
|
707
|
-
iw->write_lock = store->open_lock(store, WRITE_LOCK_NAME);
|
708
|
-
if (!iw->write_lock->obtain(iw->write_lock)) {
|
709
|
-
RAISE(STATE_ERROR, WRITE_LOCK_ERROR_MSG);
|
710
|
-
}
|
711
|
-
|
712
|
-
if (create) {
|
713
|
-
Lock *commit_lock = store->open_lock(store, COMMIT_LOCK_NAME);
|
714
|
-
if (!commit_lock->obtain(commit_lock)) {
|
715
|
-
store->close_lock(commit_lock);
|
716
|
-
RAISE(STATE_ERROR, COMMIT_LOCK_ERROR_MSG);
|
717
|
-
}
|
718
|
-
TRY
|
719
|
-
/* commit the index */
|
720
|
-
store->clear(store);
|
721
|
-
sis_write(iw->sis, store);
|
722
|
-
XFINALLY
|
723
|
-
commit_lock->release(commit_lock);
|
724
|
-
store->close_lock(commit_lock);
|
725
|
-
XENDTRY
|
726
|
-
} else {
|
727
|
-
sis_read(iw->sis, store);
|
728
|
-
}
|
729
|
-
mutex_unlock(&store->mutex);
|
730
|
-
return iw;
|
731
|
-
}
|
732
|
-
|
733
|
-
const char base36_digitmap[] = "0123456789abcdefghijklmnopqrstuvwxyz";
|
734
|
-
|
735
|
-
char *new_segment_name(int counter)
|
736
|
-
{
|
737
|
-
char buf[SEGMENT_NAME_MAX_LENGTH];
|
738
|
-
int i;
|
739
|
-
|
740
|
-
buf[SEGMENT_NAME_MAX_LENGTH - 1] = '\0';
|
741
|
-
for (i = SEGMENT_NAME_MAX_LENGTH - 2; ; i--) {
|
742
|
-
buf[i] = base36_digitmap[counter%36];
|
743
|
-
counter /= 36;
|
744
|
-
if (counter == 0) break;
|
745
|
-
}
|
746
|
-
i--;
|
747
|
-
buf[i] = '_';
|
748
|
-
return estrdup(&buf[i]);
|
749
|
-
}
|
750
|
-
|
751
|
-
int iw_doc_count(IndexWriter *iw)
|
752
|
-
{
|
753
|
-
int i, doc_cnt = 0;
|
754
|
-
mutex_lock(&iw->mutex);
|
755
|
-
for (i = 0; i < iw->sis->scnt; i++)
|
756
|
-
doc_cnt += iw->sis->segs[i]->doc_cnt;
|
757
|
-
mutex_unlock(&iw->mutex);
|
758
|
-
return doc_cnt;
|
759
|
-
}
|
760
|
-
|
761
|
-
void delete_files(Array *file_names, Store *store)
|
762
|
-
{
|
763
|
-
int i;
|
764
|
-
for (i = 0; i < file_names->size; i++) {
|
765
|
-
store->remove(store, (char *)file_names->elems[i]);
|
766
|
-
}
|
767
|
-
ary_destroy(file_names);
|
768
|
-
}
|
769
|
-
|
770
|
-
|
771
|
-
Array *sr_file_names(IndexReader *ir);
|
772
|
-
void iw_delete_segments(IndexWriter *iw, IndexReader **segment_readers, int del_cnt)
|
773
|
-
{
|
774
|
-
/* The java version keeps a record of files that it couldn't delete. This
|
775
|
-
* shouldn't be a problem on linux I hope. */
|
776
|
-
IndexReader *ir;
|
777
|
-
int i;
|
778
|
-
for (i = 0; i < del_cnt; i++) {
|
779
|
-
ir = segment_readers[i];
|
780
|
-
delete_files(sr_file_names(ir), ir->store);
|
781
|
-
}
|
782
|
-
}
|
783
|
-
|
784
|
-
void make_compound_file(IndexWriter *iw, char *merged_name, SegmentMerger *merger)
|
785
|
-
{
|
786
|
-
Array *files_to_delete;
|
787
|
-
Lock *commit_lock;
|
788
|
-
char merged_tmp[SEGMENT_NAME_MAX_LENGTH], merged_cfs[SEGMENT_NAME_MAX_LENGTH];
|
789
|
-
|
790
|
-
mutex_lock(&iw->store->mutex);
|
791
|
-
sprintf(merged_tmp, "%s.tmp", merged_name);
|
792
|
-
sprintf(merged_cfs, "%s.cfs", merged_name);
|
793
|
-
|
794
|
-
files_to_delete = sm_create_compound_file(merger, merged_tmp);
|
795
|
-
commit_lock = iw->store->open_lock(iw->store, COMMIT_LOCK_NAME);
|
796
|
-
|
797
|
-
if (!commit_lock->obtain(commit_lock)) {
|
798
|
-
RAISE(STATE_ERROR, COMMIT_LOCK_ERROR_MSG);
|
799
|
-
}
|
800
|
-
|
801
|
-
/* make compound file visible for SegmentReaders */
|
802
|
-
iw->store->rename(iw->store, merged_tmp, merged_cfs);
|
803
|
-
|
804
|
-
/* delete now unused files of segment */
|
805
|
-
delete_files(files_to_delete, iw->store);
|
806
|
-
|
807
|
-
commit_lock->release(commit_lock);
|
808
|
-
iw->store->close_lock(commit_lock);
|
809
|
-
mutex_unlock(&iw->store->mutex);
|
810
|
-
}
|
811
|
-
|
812
|
-
void iw_merge_segments_with_max(IndexWriter *iw, int min_segment, int max_segment)
|
813
|
-
{
|
814
|
-
int i;
|
815
|
-
int merged_doc_count;
|
816
|
-
Lock *commit_lock;
|
817
|
-
IndexReader **segments_to_delete = ALLOC_N(IndexReader *, max_segment - min_segment);
|
818
|
-
int del_cnt = 0;
|
819
|
-
|
820
|
-
char *merged_name = new_segment_name(iw->sis->counter++);
|
821
|
-
|
822
|
-
SegmentMerger *merger = sm_create(iw->store, merged_name, iw->term_index_interval);
|
823
|
-
IndexReader *reader;
|
824
|
-
|
825
|
-
|
826
|
-
for (i = min_segment; i < max_segment; i++) {
|
827
|
-
reader = sr_open(iw->sis, i, false);
|
828
|
-
sm_add(merger, reader);
|
829
|
-
if ((reader->store == iw->store) || /* if we own the directory */
|
830
|
-
(reader->store == iw->ram_store)) {
|
831
|
-
segments_to_delete[del_cnt++] = reader; /* queue segment for deletion */
|
832
|
-
}
|
833
|
-
}
|
834
|
-
|
835
|
-
merged_doc_count = sm_merge(merger);
|
836
|
-
|
837
|
-
sis_del_from_to(iw->sis, min_segment, max_segment);
|
838
|
-
|
839
|
-
sis_add_si(iw->sis, si_create(merged_name, merged_doc_count, iw->store));
|
840
|
-
|
841
|
-
/* close readers before we attempt to delete now-obsolete segments */
|
842
|
-
|
843
|
-
mutex_lock(&iw->store->mutex);
|
844
|
-
commit_lock = iw->store->open_lock(iw->store, COMMIT_LOCK_NAME);
|
845
|
-
if (!commit_lock->obtain(commit_lock)) {
|
846
|
-
RAISE(STATE_ERROR, COMMIT_LOCK_ERROR_MSG);
|
847
|
-
}
|
848
|
-
/* commit the index */
|
849
|
-
sis_write(iw->sis, iw->store);
|
850
|
-
iw_delete_segments(iw, segments_to_delete, del_cnt);
|
851
|
-
|
852
|
-
commit_lock->release(commit_lock);
|
853
|
-
iw->store->close_lock(commit_lock);
|
854
|
-
mutex_unlock(&iw->store->mutex);
|
855
|
-
|
856
|
-
if (iw->use_compound_file) {
|
857
|
-
make_compound_file(iw, merged_name, merger);
|
858
|
-
}
|
859
|
-
|
860
|
-
free(segments_to_delete);
|
861
|
-
sm_destroy(merger);
|
862
|
-
}
|
863
|
-
|
864
|
-
void iw_merge_segments(IndexWriter *iw, int min_segment)
|
865
|
-
{
|
866
|
-
iw_merge_segments_with_max(iw, min_segment, iw->sis->scnt);
|
867
|
-
}
|
868
|
-
|
869
|
-
void iw_maybe_merge_segments(IndexWriter *iw)
|
870
|
-
{
|
871
|
-
int target_merge_docs = iw->min_merge_docs;
|
872
|
-
int min_segment, merge_docs;
|
873
|
-
SegmentInfo *si;
|
874
|
-
|
875
|
-
while (target_merge_docs <= iw->max_merge_docs) {
|
876
|
-
/* find segments smaller than current target size */
|
877
|
-
min_segment = iw->sis->scnt - 1;
|
878
|
-
merge_docs = 0;
|
879
|
-
while (min_segment >= 0) {
|
880
|
-
si = iw->sis->segs[min_segment];
|
881
|
-
if (si->doc_cnt >= target_merge_docs) {
|
882
|
-
break;
|
883
|
-
}
|
884
|
-
merge_docs += si->doc_cnt;
|
885
|
-
min_segment -= 1;
|
886
|
-
}
|
887
|
-
|
888
|
-
if (merge_docs >= target_merge_docs) { /* found a merge to do */
|
889
|
-
iw_merge_segments(iw, min_segment + 1);
|
890
|
-
} else {
|
891
|
-
break;
|
892
|
-
}
|
893
|
-
|
894
|
-
target_merge_docs *= iw->merge_factor; /* increase target size */
|
895
|
-
}
|
896
|
-
}
|
897
|
-
|
898
|
-
void iw_flush_ram_segments(IndexWriter *iw)
|
899
|
-
{
|
900
|
-
int min_segment = iw->sis->scnt-1;
|
901
|
-
int doc_count = 0;
|
902
|
-
SegmentInfo **segs = iw->sis->segs;
|
903
|
-
while ((min_segment >= 0) &&
|
904
|
-
(segs[min_segment]->store == iw->ram_store)) {
|
905
|
-
doc_count += segs[min_segment]->doc_cnt;
|
906
|
-
min_segment--;
|
907
|
-
}
|
908
|
-
/* the following if statement is actually incrementing for different
|
909
|
-
* reasons. If min_segment < 0 then we must increment as we searched
|
910
|
-
* off the end. If the top segment is not ram_store there are no
|
911
|
-
* ram segments to flush so we increment so the next check will return
|
912
|
-
* us from this function. Lastly, the min_segment stopped at a segment
|
913
|
-
* that wasn't the ram segment. But if it fit's in with the merge
|
914
|
-
* factor, why not merge it. Otherwise we leave it and increment min_seg
|
915
|
-
*/
|
916
|
-
if ((min_segment < 0) || /* add one FS segment? */
|
917
|
-
((doc_count + segs[min_segment]->doc_cnt) > iw->merge_factor) ||
|
918
|
-
(segs[iw->sis->scnt - 1]->store != iw->ram_store)) {
|
919
|
-
min_segment++;
|
920
|
-
}
|
921
|
-
if (min_segment >= iw->sis->scnt) {
|
922
|
-
return;
|
923
|
-
}
|
924
|
-
iw_merge_segments(iw, min_segment);
|
925
|
-
}
|
926
|
-
|
927
|
-
void iw_add_doc(IndexWriter *iw, Document *doc)
|
928
|
-
{
|
929
|
-
DocumentWriter *dw;
|
930
|
-
char *segment_name;
|
931
|
-
|
932
|
-
mutex_lock(&iw->mutex);
|
933
|
-
dw = dw_open(iw->ram_store,
|
934
|
-
iw->analyzer,
|
935
|
-
iw->similarity,
|
936
|
-
iw->max_field_length,
|
937
|
-
iw->term_index_interval);
|
938
|
-
segment_name = new_segment_name(iw->sis->counter++);
|
939
|
-
dw_add_doc(dw, segment_name, doc);
|
940
|
-
dw_close(dw);
|
941
|
-
sis_add_si(iw->sis, si_create(segment_name, 1, iw->ram_store));
|
942
|
-
iw_maybe_merge_segments(iw);
|
943
|
-
mutex_unlock(&iw->mutex);
|
944
|
-
}
|
945
|
-
|
946
|
-
static inline void iw_optimize_internal(IndexWriter *iw)
|
947
|
-
{
|
948
|
-
int min_segment;
|
949
|
-
iw_flush_ram_segments(iw);
|
950
|
-
while (iw->sis->scnt > 1 ||
|
951
|
-
(iw->sis->scnt == 1 &&
|
952
|
-
( si_has_deletions(iw->sis->segs[0]) ||
|
953
|
-
(iw->sis->segs[0]->store != iw->store) ||
|
954
|
-
(iw->use_compound_file &&
|
955
|
-
(!si_uses_compound_file(iw->sis->segs[0]) ||
|
956
|
-
si_has_separate_norms(iw->sis->segs[0])))))) {
|
957
|
-
min_segment = iw->sis->scnt - iw->merge_factor;
|
958
|
-
iw_merge_segments(iw, min_segment < 0 ? 0 : min_segment);
|
959
|
-
}
|
960
|
-
}
|
961
|
-
void iw_optimize(IndexWriter *iw)
|
962
|
-
{
|
963
|
-
mutex_lock(&iw->mutex);
|
964
|
-
iw_optimize_internal(iw);
|
965
|
-
mutex_unlock(&iw->mutex);
|
966
|
-
}
|
967
|
-
|
968
|
-
void iw_close(IndexWriter *iw)
|
969
|
-
{
|
970
|
-
mutex_lock(&iw->mutex);
|
971
|
-
iw_flush_ram_segments(iw);
|
972
|
-
store_deref(iw->ram_store);
|
973
|
-
sis_destroy(iw->sis);
|
974
|
-
|
975
|
-
sim_destroy(iw->similarity);
|
976
|
-
a_deref(iw->analyzer);
|
977
|
-
|
978
|
-
iw->write_lock->release(iw->write_lock);
|
979
|
-
iw->store->close_lock(iw->write_lock);
|
980
|
-
|
981
|
-
store_deref(iw->store);
|
982
|
-
mutex_destroy(&iw->mutex);
|
983
|
-
free(iw);
|
984
|
-
}
|
985
|
-
|
986
|
-
void iw_add_indexes(IndexWriter *iw, Store **stores, int cnt)
|
987
|
-
{
|
988
|
-
int i, j, end, start;
|
989
|
-
|
990
|
-
mutex_lock(&iw->mutex);
|
991
|
-
iw_optimize_internal(iw); /* start with zero or 1 seg */
|
992
|
-
|
993
|
-
start = iw->sis->scnt;
|
994
|
-
|
995
|
-
for (i = 0; i < cnt; i++) {
|
996
|
-
Store *store = stores[i];
|
997
|
-
SegmentInfos *sis = sis_create(); /* read infos from dir */
|
998
|
-
sis_read(sis, store);
|
999
|
-
|
1000
|
-
for (j = 0; j < sis->scnt; j++) {
|
1001
|
-
SegmentInfo *si = sis->segs[j];
|
1002
|
-
sis_add_si(iw->sis, si);
|
1003
|
-
}
|
1004
|
-
sis_destroy_not_infos(sis);
|
1005
|
-
}
|
1006
|
-
|
1007
|
-
/* merge newly added segments in log(n) passes */
|
1008
|
-
while (iw->sis->scnt > start + iw->merge_factor) {
|
1009
|
-
for (i = start + 1; i < iw->sis->scnt; i++) {
|
1010
|
-
end = MIN(iw->sis->scnt, i + iw->merge_factor);
|
1011
|
-
if (end - i > 1) {
|
1012
|
-
iw_merge_segments_with_max(iw, i, end);
|
1013
|
-
}
|
1014
|
-
}
|
1015
|
-
}
|
1016
|
-
|
1017
|
-
/* final cleanup */
|
1018
|
-
iw_optimize_internal(iw);
|
1019
|
-
mutex_unlock(&iw->mutex);
|
1020
|
-
}
|
1021
|
-
|
1022
|
-
|
1023
|
-
/**
|
1024
|
-
* This adds an array of readers to the index leaving the added readers open.
|
1025
|
-
*/
|
1026
|
-
void iw_add_readers(IndexWriter *iw, IndexReader **irs, int cnt)
|
1027
|
-
{
|
1028
|
-
IndexReader *ir = NULL;
|
1029
|
-
int i, del_cnt = 0;
|
1030
|
-
int doc_count;
|
1031
|
-
char *merged_name;
|
1032
|
-
SegmentMerger *merger;
|
1033
|
-
Lock *commit_lock;
|
1034
|
-
|
1035
|
-
mutex_lock(&iw->mutex);
|
1036
|
-
iw_optimize_internal(iw); /* start with zero or 1 seg */
|
1037
|
-
|
1038
|
-
merged_name = new_segment_name(iw->sis->counter++);
|
1039
|
-
|
1040
|
-
merger = sm_create(iw->store, merged_name, iw->term_index_interval);
|
1041
|
-
merger->readers->free_elem = NULL; /* don't close readers */
|
1042
|
-
|
1043
|
-
if (iw->sis->scnt == 1) { /* add existing index, if any */
|
1044
|
-
ir = sr_open_si(iw->sis->segs[0]);
|
1045
|
-
sm_add(merger, ir);
|
1046
|
-
del_cnt = 1;
|
1047
|
-
}
|
1048
|
-
|
1049
|
-
for (i = 0; i < cnt; i++) {
|
1050
|
-
sm_add(merger, irs[i]);
|
1051
|
-
}
|
1052
|
-
|
1053
|
-
doc_count = sm_merge(merger); /* merge 'em */
|
1054
|
-
|
1055
|
-
/* pop old infos and add new ones. */
|
1056
|
-
sis_clear(iw->sis);
|
1057
|
-
sis_add_si(iw->sis, si_create(merged_name, doc_count, iw->store));
|
1058
|
-
|
1059
|
-
|
1060
|
-
commit_lock = iw->store->open_lock(iw->store, COMMIT_LOCK_NAME);
|
1061
|
-
if (!commit_lock->obtain(commit_lock)) { /* obtain write lock */
|
1062
|
-
RAISE(STATE_ERROR, COMMIT_LOCK_ERROR_MSG);
|
1063
|
-
}
|
1064
|
-
|
1065
|
-
sis_write(iw->sis, iw->store); /* commit changes */
|
1066
|
-
iw_delete_segments(iw, &ir, del_cnt);
|
1067
|
-
if (ir) ir_close(ir);
|
1068
|
-
|
1069
|
-
commit_lock->release(commit_lock);
|
1070
|
-
iw->store->close_lock(commit_lock);
|
1071
|
-
|
1072
|
-
if (iw->use_compound_file) {
|
1073
|
-
make_compound_file(iw, merged_name, merger);
|
1074
|
-
}
|
1075
|
-
|
1076
|
-
iw_optimize_internal(iw);
|
1077
|
-
sm_destroy(merger);
|
1078
|
-
|
1079
|
-
mutex_unlock(&iw->mutex);
|
1080
|
-
}
|
1081
|
-
|
1082
|
-
/****************************************************************************
|
1083
|
-
*
|
1084
|
-
* Norm
|
1085
|
-
*
|
1086
|
-
****************************************************************************/
|
1087
|
-
|
1088
|
-
Norm *norm_create(InStream *is, int field_num)
|
1089
|
-
{
|
1090
|
-
Norm *norm = ALLOC(Norm);
|
1091
|
-
norm->is = is;
|
1092
|
-
norm->field_num = field_num;
|
1093
|
-
norm->bytes = NULL;
|
1094
|
-
norm->is_dirty = false;
|
1095
|
-
return norm;
|
1096
|
-
}
|
1097
|
-
|
1098
|
-
void norm_destroy(Norm *norm)
|
1099
|
-
{
|
1100
|
-
is_close(norm->is);
|
1101
|
-
if (norm->bytes != NULL) {
|
1102
|
-
free(norm->bytes);
|
1103
|
-
}
|
1104
|
-
free(norm);
|
1105
|
-
}
|
1106
|
-
|
1107
|
-
void norm_rewrite(Norm *norm, Store *store, char *segment,
|
1108
|
-
int doc_count, Store *cfs_store)
|
1109
|
-
{
|
1110
|
-
OutStream *os;
|
1111
|
-
char tmp_fname[SEGMENT_NAME_MAX_LENGTH];
|
1112
|
-
char norm_fname[SEGMENT_NAME_MAX_LENGTH];
|
1113
|
-
|
1114
|
-
if (norm->bytes == NULL) {
|
1115
|
-
return; /* These norms do not need to be rewritten */
|
1116
|
-
}
|
1117
|
-
|
1118
|
-
sprintf(tmp_fname, "%s.tmp", segment);
|
1119
|
-
os = store->create_output(store, tmp_fname);
|
1120
|
-
TRY
|
1121
|
-
os_write_bytes(os, norm->bytes, doc_count);
|
1122
|
-
XFINALLY
|
1123
|
-
os_close(os);
|
1124
|
-
XENDTRY
|
1125
|
-
if (cfs_store) {
|
1126
|
-
sprintf(norm_fname, "%s.s%d", segment, norm->field_num);
|
1127
|
-
} else {
|
1128
|
-
sprintf(norm_fname, "%s.f%d", segment, norm->field_num);
|
1129
|
-
}
|
1130
|
-
store->rename(store, tmp_fname, norm_fname);
|
1131
|
-
norm->is_dirty = false;
|
1132
|
-
}
|
1133
|
-
|
1134
|
-
/****************************************************************************
|
1135
|
-
*
|
1136
|
-
* SegmentReader
|
1137
|
-
*
|
1138
|
-
****************************************************************************/
|
1139
|
-
|
1140
|
-
#define GET_SR SegmentReader *sr = (SegmentReader *)ir->data
|
1141
|
-
|
1142
|
-
int sr_max_doc(IndexReader *ir)
|
1143
|
-
{
|
1144
|
-
return ((SegmentReader *)ir->data)->fr->len;
|
1145
|
-
}
|
1146
|
-
|
1147
|
-
static inline void sr_close_norms(SegmentReader *sr)
|
1148
|
-
{
|
1149
|
-
h_destroy(sr->norms);
|
1150
|
-
}
|
1151
|
-
|
1152
|
-
static inline TermVectorsReader *sr_tvr(SegmentReader *sr)
|
1153
|
-
{
|
1154
|
-
TermVectorsReader *tvr;
|
1155
|
-
if ((tvr = thread_getspecific(sr->thread_tvr)) == NULL) {
|
1156
|
-
tvr = tvr_clone(sr->orig_tvr);
|
1157
|
-
if (tvr == NULL) printf("scuk\n");
|
1158
|
-
ary_append(sr->tvr_bucket, tvr);
|
1159
|
-
thread_setspecific(sr->thread_tvr, tvr);
|
1160
|
-
}
|
1161
|
-
return tvr;
|
1162
|
-
}
|
1163
|
-
|
1164
|
-
void sr_close(IndexReader *ir)
|
1165
|
-
{
|
1166
|
-
GET_SR;
|
1167
|
-
fr_close(sr->fr);
|
1168
|
-
tir_close(sr->tir);
|
1169
|
-
|
1170
|
-
if (sr->freq_in) is_close(sr->freq_in);
|
1171
|
-
if (sr->prox_in) is_close(sr->prox_in);
|
1172
|
-
|
1173
|
-
fis_destroy(sr->fis);
|
1174
|
-
sr_close_norms(sr);
|
1175
|
-
|
1176
|
-
if (sr->orig_tvr) {
|
1177
|
-
tvr_close(sr->orig_tvr);
|
1178
|
-
thread_key_delete(sr->thread_tvr);
|
1179
|
-
ary_destroy(sr->tvr_bucket);
|
1180
|
-
}
|
1181
|
-
if (sr->deleted_docs) bv_destroy(sr->deleted_docs);
|
1182
|
-
if (sr->cfs_store) store_deref(sr->cfs_store);
|
1183
|
-
if (sr->fake_norms) free(sr->fake_norms);
|
1184
|
-
free(sr->segment);
|
1185
|
-
free(sr);
|
1186
|
-
}
|
1187
|
-
|
1188
|
-
void sr_delete_doc(IndexReader *ir, int doc_num)
|
1189
|
-
{
|
1190
|
-
GET_SR;
|
1191
|
-
if (sr->deleted_docs == NULL)
|
1192
|
-
sr->deleted_docs = bv_create();
|
1193
|
-
|
1194
|
-
sr->deleted_docs_dirty = true;
|
1195
|
-
sr->undelete_all = false;
|
1196
|
-
bv_set(sr->deleted_docs, doc_num);
|
1197
|
-
}
|
1198
|
-
|
1199
|
-
static inline bool sr_is_deleted_internal(IndexReader *ir, int doc_num)
|
1200
|
-
{
|
1201
|
-
GET_SR;
|
1202
|
-
return (sr->deleted_docs != NULL && bv_get(sr->deleted_docs, doc_num));
|
1203
|
-
}
|
1204
|
-
|
1205
|
-
bool sr_is_deleted(IndexReader *ir, int doc_num)
|
1206
|
-
{
|
1207
|
-
bool is_del;
|
1208
|
-
|
1209
|
-
mutex_lock(&ir->mutex);
|
1210
|
-
is_del = sr_is_deleted_internal(ir, doc_num);
|
1211
|
-
mutex_unlock(&ir->mutex);
|
1212
|
-
|
1213
|
-
return is_del;
|
1214
|
-
}
|
1215
|
-
|
1216
|
-
bool sr_has_norms(IndexReader *ir, char *field)
|
1217
|
-
{
|
1218
|
-
GET_SR;
|
1219
|
-
bool has_norms;
|
1220
|
-
mutex_lock(&ir->mutex);
|
1221
|
-
has_norms = h_has_key(sr->norms, field);
|
1222
|
-
mutex_unlock(&ir->mutex);
|
1223
|
-
|
1224
|
-
return has_norms;
|
1225
|
-
}
|
1226
|
-
|
1227
|
-
bool sr_has_deletions(IndexReader *ir)
|
1228
|
-
{
|
1229
|
-
GET_SR;
|
1230
|
-
return (sr->deleted_docs != NULL);
|
1231
|
-
}
|
1232
|
-
|
1233
|
-
void sr_undelete_all(IndexReader *ir)
|
1234
|
-
{
|
1235
|
-
GET_SR;
|
1236
|
-
sr->undelete_all = true;
|
1237
|
-
sr->deleted_docs_dirty = false;
|
1238
|
-
if (sr->deleted_docs != NULL) bv_destroy(sr->deleted_docs);
|
1239
|
-
sr->deleted_docs = NULL;
|
1240
|
-
}
|
1241
|
-
|
1242
|
-
TermEnum *sr_terms(IndexReader *ir)
|
1243
|
-
{
|
1244
|
-
TermEnum *te = ((SegmentReader *)ir->data)->tir->orig_te;
|
1245
|
-
return te->clone(te);
|
1246
|
-
}
|
1247
|
-
|
1248
|
-
TermEnum *sr_terms_from(IndexReader *ir, Term *term)
|
1249
|
-
{
|
1250
|
-
TermEnum *te = ((SegmentReader *)ir->data)->tir->orig_te;
|
1251
|
-
TermEnum *ret_te = te->clone(te);
|
1252
|
-
te_skip_to(ret_te, term);
|
1253
|
-
return ret_te;
|
1254
|
-
}
|
1255
|
-
|
1256
|
-
Document *sr_get_doc(IndexReader *ir, int doc_num)
|
1257
|
-
{
|
1258
|
-
GET_SR;
|
1259
|
-
Document *doc;
|
1260
|
-
mutex_lock(&ir->mutex);
|
1261
|
-
if (sr_is_deleted_internal(ir, doc_num)) {
|
1262
|
-
mutex_unlock(&ir->mutex);
|
1263
|
-
RAISE(STATE_ERROR, DELETED_DOC_ERROR_MSG);
|
1264
|
-
}
|
1265
|
-
doc = fr_get_doc(sr->fr, doc_num);
|
1266
|
-
mutex_unlock(&ir->mutex);
|
1267
|
-
return doc;
|
1268
|
-
}
|
1269
|
-
|
1270
|
-
static inline void
|
1271
|
-
sr_get_norms_into_internal(IndexReader *ir, char *field, uchar *buf, int offset)
|
1272
|
-
{
|
1273
|
-
GET_SR;
|
1274
|
-
Norm *norm = h_get(sr->norms, field);
|
1275
|
-
if (norm == NULL) {
|
1276
|
-
memset(buf + offset*sizeof(uchar), 0, sr_max_doc(ir)*sizeof(uchar));
|
1277
|
-
} else if (norm->bytes != NULL) { /* can copy from cache */
|
1278
|
-
memcpy(buf + offset*sizeof(uchar), norm->bytes, sr_max_doc(ir)*sizeof(uchar));
|
1279
|
-
} else {
|
1280
|
-
InStream *norm_in = is_clone(norm->is);
|
1281
|
-
/* read from disk */
|
1282
|
-
is_seek(norm_in, 0);
|
1283
|
-
is_read_bytes(norm_in, buf, offset, sr_max_doc(ir));
|
1284
|
-
is_close(norm_in);
|
1285
|
-
}
|
1286
|
-
}
|
1287
|
-
|
1288
|
-
void sr_get_norms_into(IndexReader *ir, char *field, uchar *buf, int offset)
|
1289
|
-
{
|
1290
|
-
mutex_lock(&ir->mutex);
|
1291
|
-
sr_get_norms_into_internal(ir, field, buf, offset);
|
1292
|
-
mutex_unlock(&ir->mutex);
|
1293
|
-
}
|
1294
|
-
|
1295
|
-
static inline uchar *sr_get_norms_internal(IndexReader *ir, char *field)
|
1296
|
-
{
|
1297
|
-
GET_SR;
|
1298
|
-
Norm *norm = h_get(sr->norms, field);
|
1299
|
-
if (norm == NULL) { /* not an indexed field */
|
1300
|
-
return NULL;
|
1301
|
-
}
|
1302
|
-
|
1303
|
-
if (norm->bytes == NULL) { /* value not yet read */
|
1304
|
-
uchar *bytes = ALLOC_N(uchar, ir->max_doc(ir));
|
1305
|
-
sr_get_norms_into_internal(ir, field, bytes, 0);
|
1306
|
-
norm->bytes = bytes; /* cache it */
|
1307
|
-
}
|
1308
|
-
return norm->bytes;
|
1309
|
-
}
|
1310
|
-
|
1311
|
-
uchar *sr_get_norms(IndexReader *ir, char *field)
|
1312
|
-
{
|
1313
|
-
uchar *norms;
|
1314
|
-
mutex_lock(&ir->mutex);
|
1315
|
-
norms = sr_get_norms_internal(ir, field);
|
1316
|
-
mutex_unlock(&ir->mutex);
|
1317
|
-
return norms;
|
1318
|
-
}
|
1319
|
-
|
1320
|
-
static inline uchar *sr_get_norms_always(IndexReader *ir, char *field)
|
1321
|
-
{
|
1322
|
-
GET_SR;
|
1323
|
-
uchar *bytes;
|
1324
|
-
mutex_lock(&ir->mutex);
|
1325
|
-
|
1326
|
-
bytes = sr_get_norms_internal(ir, field);
|
1327
|
-
if (bytes == NULL) {
|
1328
|
-
if (sr->fake_norms) {
|
1329
|
-
bytes = sr->fake_norms;
|
1330
|
-
} else {
|
1331
|
-
int len = ir->max_doc(ir);
|
1332
|
-
sr->fake_norms = bytes = ALLOC_N(uchar, len);
|
1333
|
-
memset(bytes, 0, len);
|
1334
|
-
}
|
1335
|
-
}
|
1336
|
-
mutex_unlock(&ir->mutex);
|
1337
|
-
return bytes;
|
1338
|
-
}
|
1339
|
-
|
1340
|
-
void sr_set_norm(IndexReader *ir, int doc_num, char *field, uchar val)
|
1341
|
-
{
|
1342
|
-
GET_SR;
|
1343
|
-
Norm *norm;
|
1344
|
-
|
1345
|
-
norm = h_get(sr->norms, field);
|
1346
|
-
if (norm != NULL) { /* an indexed field */
|
1347
|
-
norm->is_dirty = true; /* mark it dirty */
|
1348
|
-
sr->norms_dirty = true;
|
1349
|
-
|
1350
|
-
sr_get_norms_internal(ir, field)[doc_num] = val;
|
1351
|
-
}
|
1352
|
-
}
|
1353
|
-
|
1354
|
-
int sr_doc_freq(IndexReader *ir, Term *t)
|
1355
|
-
{
|
1356
|
-
GET_SR;
|
1357
|
-
TermInfo *ti = tir_get_ti(sr->tir, t);
|
1358
|
-
if (ti != NULL) {
|
1359
|
-
int df = ti->doc_freq;
|
1360
|
-
ti_destroy(ti);
|
1361
|
-
return df;
|
1362
|
-
} else {
|
1363
|
-
return 0;
|
1364
|
-
}
|
1365
|
-
}
|
1366
|
-
|
1367
|
-
Array *sr_file_names(IndexReader *ir)
|
1368
|
-
{
|
1369
|
-
GET_SR;
|
1370
|
-
Array *file_names = ary_create(0, &free);
|
1371
|
-
FieldInfo *fi;
|
1372
|
-
int i;
|
1373
|
-
char fname[SEGMENT_NAME_MAX_LENGTH];
|
1374
|
-
|
1375
|
-
for (i = 0; i < NELEMS(INDEX_EXTENSIONS); i++) {
|
1376
|
-
sprintf(fname, "%s.%s", sr->segment, INDEX_EXTENSIONS[i]);
|
1377
|
-
if (ir->store->exists(ir->store, fname))
|
1378
|
-
ary_append(file_names, estrdup(fname));
|
1379
|
-
}
|
1380
|
-
|
1381
|
-
for (i = 0; i < sr->fis->fcnt; i++) {
|
1382
|
-
fi = sr->fis->by_number[i];
|
1383
|
-
if (fi->is_indexed && !fi->omit_norms) {
|
1384
|
-
if (sr->cfs_store) {
|
1385
|
-
sprintf(fname, "%s.s%d", sr->segment, i);
|
1386
|
-
} else {
|
1387
|
-
sprintf(fname, "%s.f%d", sr->segment, i);
|
1388
|
-
}
|
1389
|
-
if (ir->store->exists(ir->store, fname))
|
1390
|
-
ary_append(file_names, estrdup(fname));
|
1391
|
-
}
|
1392
|
-
}
|
1393
|
-
return file_names;
|
1394
|
-
}
|
1395
|
-
|
1396
|
-
HashSet *sr_get_field_names(IndexReader *ir, int field_type)
|
1397
|
-
{
|
1398
|
-
GET_SR;
|
1399
|
-
int i;
|
1400
|
-
HashSet *field_set = hs_str_create(NULL);
|
1401
|
-
FieldInfo *fi;
|
1402
|
-
for (i = 0; i < sr->fis->fcnt; i++) {
|
1403
|
-
fi = sr->fis->by_number[i];
|
1404
|
-
switch(field_type) {
|
1405
|
-
case IR_ALL:
|
1406
|
-
hs_add(field_set, fi->name);
|
1407
|
-
break;
|
1408
|
-
case IR_UNINDEXED:
|
1409
|
-
if (!fi->is_indexed) hs_add(field_set, fi->name);
|
1410
|
-
break;
|
1411
|
-
case IR_INDEXED:
|
1412
|
-
if (fi->is_indexed) hs_add(field_set, fi->name);
|
1413
|
-
break;
|
1414
|
-
case IR_INDEXED_NO_TERM_VECTOR:
|
1415
|
-
if (fi->is_indexed && !fi->store_tv) hs_add(field_set, fi->name);
|
1416
|
-
break;
|
1417
|
-
case IR_TERM_VECTOR:
|
1418
|
-
if (fi->store_tv && !fi->store_pos && !fi->store_offset)
|
1419
|
-
hs_add(field_set, fi->name);
|
1420
|
-
break;
|
1421
|
-
case IR_INDEXED_WITH_TERM_VECTOR:
|
1422
|
-
if (fi->is_indexed && fi->store_tv) hs_add(field_set, fi->name);
|
1423
|
-
break;
|
1424
|
-
case IR_TERM_VECTOR_WITH_POSITION:
|
1425
|
-
if (fi->store_pos && !fi->store_offset) hs_add(field_set, fi->name);
|
1426
|
-
break;
|
1427
|
-
case IR_TERM_VECTOR_WITH_OFFSET:
|
1428
|
-
if (!fi->store_pos && fi->store_offset) hs_add(field_set, fi->name);
|
1429
|
-
case IR_TERM_VECTOR_WITH_POSITION_OFFSET:
|
1430
|
-
if (fi->store_pos && fi->store_offset) hs_add(field_set, fi->name);
|
1431
|
-
break;
|
1432
|
-
default:
|
1433
|
-
RAISE(ARG_ERROR, INVALID_FIELD_TYPE_MSG);
|
1434
|
-
}
|
1435
|
-
}
|
1436
|
-
return field_set;
|
1437
|
-
}
|
1438
|
-
|
1439
|
-
int sr_num_docs(IndexReader *ir)
|
1440
|
-
{
|
1441
|
-
GET_SR;
|
1442
|
-
int num_docs;
|
1443
|
-
|
1444
|
-
mutex_lock(&ir->mutex);
|
1445
|
-
num_docs = sr_max_doc(ir);
|
1446
|
-
if (sr->deleted_docs != NULL)
|
1447
|
-
num_docs -= sr->deleted_docs->count;
|
1448
|
-
mutex_unlock(&ir->mutex);
|
1449
|
-
return num_docs;
|
1450
|
-
}
|
1451
|
-
|
1452
|
-
TermDocEnum *sr_term_docs(IndexReader *ir)
|
1453
|
-
{
|
1454
|
-
return stde_create(ir);
|
1455
|
-
}
|
1456
|
-
|
1457
|
-
TermDocEnum *sr_term_positions(IndexReader *ir)
|
1458
|
-
{
|
1459
|
-
return stpe_create(ir);
|
1460
|
-
}
|
1461
|
-
|
1462
|
-
void sr_open_norms(IndexReader *ir, Store *cfs_store)
|
1463
|
-
{
|
1464
|
-
GET_SR;
|
1465
|
-
int i;
|
1466
|
-
FieldInfo *fi;
|
1467
|
-
Store *tmp_store;
|
1468
|
-
char fname[SEGMENT_NAME_MAX_LENGTH];
|
1469
|
-
for (i = 0; i < sr->fis->fcnt; i++) {
|
1470
|
-
tmp_store = ir->store;
|
1471
|
-
fi = sr->fis->by_number[i];
|
1472
|
-
if (fi->is_indexed && !fi->omit_norms) {
|
1473
|
-
sprintf(fname, "%s.s%d", sr->segment, fi->number);
|
1474
|
-
if (! tmp_store->exists(tmp_store, fname)) {
|
1475
|
-
sprintf(fname, "%s.f%d", sr->segment, fi->number);
|
1476
|
-
tmp_store = cfs_store;
|
1477
|
-
}
|
1478
|
-
h_set(sr->norms, fi->name,
|
1479
|
-
norm_create(tmp_store->open_input(tmp_store, fname), fi->number));
|
1480
|
-
}
|
1481
|
-
}
|
1482
|
-
sr->norms_dirty = false;
|
1483
|
-
}
|
1484
|
-
|
1485
|
-
TermVector *sr_get_term_vector(IndexReader *ir, int doc_num, char *field)
|
1486
|
-
{
|
1487
|
-
GET_SR;
|
1488
|
-
FieldInfo *fi = (FieldInfo *)ht_get(sr->fis->by_name, field);
|
1489
|
-
TermVectorsReader *tvr;
|
1490
|
-
|
1491
|
-
if (fi == NULL || !fi->store_tv || !sr->orig_tvr || !(tvr = sr_tvr(sr))) {
|
1492
|
-
return NULL;
|
1493
|
-
}
|
1494
|
-
|
1495
|
-
return tvr_get_field_tv(tvr, doc_num, field);
|
1496
|
-
}
|
1497
|
-
|
1498
|
-
Array *sr_get_term_vectors(IndexReader *ir, int doc_num)
|
1499
|
-
{
|
1500
|
-
GET_SR;
|
1501
|
-
TermVectorsReader *tvr;
|
1502
|
-
if (sr->orig_tvr == NULL || (tvr = sr_tvr(sr)) == NULL) {
|
1503
|
-
return NULL;
|
1504
|
-
}
|
1505
|
-
|
1506
|
-
return tvr_get_tv(tvr, doc_num);
|
1507
|
-
}
|
1508
|
-
|
1509
|
-
void sr_commit(IndexReader *ir)
|
1510
|
-
{
|
1511
|
-
GET_SR;
|
1512
|
-
char tmp_fname[SEGMENT_NAME_MAX_LENGTH];
|
1513
|
-
char del_fname[SEGMENT_NAME_MAX_LENGTH];
|
1514
|
-
|
1515
|
-
sprintf(del_fname, "%s.del", sr->segment);
|
1516
|
-
|
1517
|
-
if (sr->deleted_docs_dirty) { /* re-write deleted */
|
1518
|
-
sprintf(tmp_fname, "%s.tmp", sr->segment);
|
1519
|
-
bv_write(sr->deleted_docs, ir->store, tmp_fname);
|
1520
|
-
ir->store->rename(ir->store, tmp_fname, del_fname);
|
1521
|
-
}
|
1522
|
-
if (sr->undelete_all && ir->store->exists(ir->store, del_fname))
|
1523
|
-
ir->store->remove(ir->store, del_fname);
|
1524
|
-
if (sr->norms_dirty) {/* re-write norms */
|
1525
|
-
int i;
|
1526
|
-
FieldInfo *fi;
|
1527
|
-
for (i = 0; i < sr->fis->fcnt; i++) {
|
1528
|
-
fi = sr->fis->by_number[i];
|
1529
|
-
if (fi->is_indexed) {
|
1530
|
-
norm_rewrite((Norm *)h_get(sr->norms, fi->name), ir->store,
|
1531
|
-
sr->segment, sr_max_doc(ir), sr->cfs_store);
|
1532
|
-
}
|
1533
|
-
}
|
1534
|
-
}
|
1535
|
-
sr->deleted_docs_dirty = false;
|
1536
|
-
sr->norms_dirty = false;
|
1537
|
-
sr->undelete_all = false;
|
1538
|
-
}
|
1539
|
-
|
1540
|
-
IndexReader *sr_open_internal(IndexReader *ir, SegmentInfo *si)
|
1541
|
-
{
|
1542
|
-
Store *store = si->store;
|
1543
|
-
SegmentReader *sr = ALLOC(SegmentReader);
|
1544
|
-
char fname[SEGMENT_NAME_MAX_LENGTH];
|
1545
|
-
|
1546
|
-
ir->get_term_vector = &sr_get_term_vector;
|
1547
|
-
ir->get_term_vectors = &sr_get_term_vectors;
|
1548
|
-
ir->num_docs = &sr_num_docs;
|
1549
|
-
ir->max_doc = &sr_max_doc;
|
1550
|
-
ir->get_doc = &sr_get_doc;
|
1551
|
-
ir->get_norms_into = &sr_get_norms_into;
|
1552
|
-
ir->get_norms = &sr_get_norms;
|
1553
|
-
ir->get_norms_always = &sr_get_norms_always;
|
1554
|
-
ir->do_set_norm = &sr_set_norm;
|
1555
|
-
ir->terms = &sr_terms;
|
1556
|
-
ir->terms_from = &sr_terms_from;
|
1557
|
-
ir->doc_freq = &sr_doc_freq;
|
1558
|
-
ir->term_docs = &sr_term_docs;
|
1559
|
-
ir->term_positions = &sr_term_positions;
|
1560
|
-
ir->do_delete_doc = &sr_delete_doc;
|
1561
|
-
ir->is_deleted = &sr_is_deleted;
|
1562
|
-
ir->has_norms = &sr_has_norms;
|
1563
|
-
ir->has_deletions = &sr_has_deletions;
|
1564
|
-
ir->do_undelete_all = &sr_undelete_all;
|
1565
|
-
ir->get_field_names = &sr_get_field_names;
|
1566
|
-
ir->do_commit = &sr_commit;
|
1567
|
-
ir->do_close = &sr_close;
|
1568
|
-
ir->data = sr;
|
1569
|
-
sr->segment = estrdup(si->name);
|
1570
|
-
sr->cfs_store = NULL;
|
1571
|
-
sr->fake_norms = NULL;
|
1572
|
-
sprintf(fname, "%s.cfs", sr->segment);
|
1573
|
-
if (store->exists(store, fname)) {
|
1574
|
-
sr->cfs_store = open_cmpd_store(store, fname);
|
1575
|
-
store = sr->cfs_store;
|
1576
|
-
}
|
1577
|
-
|
1578
|
-
sprintf(fname, "%s.fnm", sr->segment);
|
1579
|
-
|
1580
|
-
sr->fis = fis_open(store, fname);
|
1581
|
-
sr->fr = fr_open(store, sr->segment, sr->fis);
|
1582
|
-
|
1583
|
-
sr->tir = tir_open(store, sr->segment, sr->fis);
|
1584
|
-
sr->deleted_docs = NULL;
|
1585
|
-
sr->deleted_docs_dirty = false;
|
1586
|
-
sr->undelete_all = false;
|
1587
|
-
if (si_has_deletions(si)) {
|
1588
|
-
sprintf(fname, "%s.del", sr->segment);
|
1589
|
-
sr->deleted_docs = bv_read(si->store, fname);
|
1590
|
-
}
|
1591
|
-
|
1592
|
-
sprintf(fname, "%s.frq", sr->segment);
|
1593
|
-
sr->freq_in = store->open_input(store, fname);
|
1594
|
-
sprintf(fname, "%s.prx", sr->segment);
|
1595
|
-
sr->prox_in = store->open_input(store, fname);
|
1596
|
-
sr->norms = h_new_str((free_ft)NULL, (free_ft)&norm_destroy);
|
1597
|
-
sr_open_norms(ir, store);
|
1598
|
-
|
1599
|
-
if (fis_has_vectors(sr->fis)) {
|
1600
|
-
sr->orig_tvr = tvr_open(store, sr->segment, sr->fis);
|
1601
|
-
thread_key_create(&sr->thread_tvr, NULL);
|
1602
|
-
sr->tvr_bucket = ary_create(1, (free_ft)&tvr_close);
|
1603
|
-
} else {
|
1604
|
-
sr->orig_tvr = NULL;
|
1605
|
-
}
|
1606
|
-
return ir;
|
1607
|
-
}
|
1608
|
-
|
1609
|
-
IndexReader *sr_open_si(SegmentInfo *si)
|
1610
|
-
{
|
1611
|
-
IndexReader *ir = ir_create(si->store, NULL, false);
|
1612
|
-
ref(si->store);
|
1613
|
-
return sr_open_internal(ir, si);
|
1614
|
-
}
|
1615
|
-
|
1616
|
-
IndexReader *sr_open(SegmentInfos *sis, int si_num, bool is_owner)
|
1617
|
-
{
|
1618
|
-
SegmentInfo *si = sis->segs[si_num];
|
1619
|
-
IndexReader *ir = ir_create(si->store, sis, is_owner);
|
1620
|
-
ref(si->store);
|
1621
|
-
return sr_open_internal(ir, si);
|
1622
|
-
}
|
1623
|
-
|
1624
|
-
/****************************************************************************
|
1625
|
-
*
|
1626
|
-
* MultiReader
|
1627
|
-
*
|
1628
|
-
****************************************************************************/
|
1629
|
-
|
1630
|
-
#define GET_MR MultiReader *mr = (MultiReader *)ir->data
|
1631
|
-
#define GET_READER(doc_num) MultiReader *mr = (MultiReader *)ir->data;\
|
1632
|
-
int i = mr_reader_index(mr, doc_num);\
|
1633
|
-
IndexReader *reader = mr->sub_readers[i]
|
1634
|
-
|
1635
|
-
|
1636
|
-
|
1637
|
-
int mr_reader_index(MultiReader *mr, int doc_num)
|
1638
|
-
{
|
1639
|
-
int lo = 0; /* search @starts array */
|
1640
|
-
int hi = mr->rcnt - 1; /* for first element less */
|
1641
|
-
int mid;
|
1642
|
-
int mid_value;
|
1643
|
-
|
1644
|
-
while (hi >= lo) {
|
1645
|
-
mid = (lo + hi) >> 1;
|
1646
|
-
mid_value = mr->starts[mid];
|
1647
|
-
if (doc_num < mid_value) {
|
1648
|
-
hi = mid - 1;
|
1649
|
-
} else if (doc_num > mid_value) {
|
1650
|
-
lo = mid + 1;
|
1651
|
-
} else { /* found a match */
|
1652
|
-
while ((mid+1 < mr->rcnt) && (mr->starts[mid+1] == mid_value))
|
1653
|
-
mid += 1; /* scan to last match in case we have empty segments */
|
1654
|
-
return mid;
|
1655
|
-
}
|
1656
|
-
}
|
1657
|
-
return hi;
|
1658
|
-
}
|
1659
|
-
|
1660
|
-
TermVector *mr_get_term_vector(IndexReader *ir, int doc_num, char *field)
|
1661
|
-
{
|
1662
|
-
GET_READER(doc_num);
|
1663
|
-
return reader->get_term_vector(reader, doc_num - mr->starts[i], field);
|
1664
|
-
}
|
1665
|
-
|
1666
|
-
Array *mr_get_term_vectors(IndexReader *ir, int doc_num)
|
1667
|
-
{
|
1668
|
-
GET_READER(doc_num);
|
1669
|
-
return reader->get_term_vectors(reader, doc_num - mr->starts[i]);
|
1670
|
-
}
|
1671
|
-
|
1672
|
-
int mr_num_docs(IndexReader *ir)
|
1673
|
-
{
|
1674
|
-
int i, num_docs;
|
1675
|
-
GET_MR;
|
1676
|
-
mutex_lock(&ir->mutex);
|
1677
|
-
if (mr->num_docs_cache == -1) {
|
1678
|
-
IndexReader *reader;
|
1679
|
-
mr->num_docs_cache = 0;
|
1680
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1681
|
-
reader = mr->sub_readers[i];
|
1682
|
-
mr->num_docs_cache += reader->num_docs(reader);
|
1683
|
-
}
|
1684
|
-
}
|
1685
|
-
num_docs = mr->num_docs_cache;
|
1686
|
-
mutex_unlock(&ir->mutex);
|
1687
|
-
|
1688
|
-
return num_docs;
|
1689
|
-
}
|
1690
|
-
|
1691
|
-
int mr_max_doc(IndexReader *ir)
|
1692
|
-
{
|
1693
|
-
GET_MR;
|
1694
|
-
return mr->max_doc;
|
1695
|
-
}
|
1696
|
-
|
1697
|
-
Document *mr_get_doc(IndexReader *ir, int doc_num)
|
1698
|
-
{
|
1699
|
-
GET_READER(doc_num);
|
1700
|
-
return reader->get_doc(reader, doc_num - mr->starts[i]);
|
1701
|
-
}
|
1702
|
-
|
1703
|
-
void mr_get_norms_into(IndexReader *ir, char *field, uchar *buf, int offset)
|
1704
|
-
{
|
1705
|
-
int i;
|
1706
|
-
uchar *bytes;
|
1707
|
-
GET_MR;
|
1708
|
-
|
1709
|
-
mutex_lock(&ir->mutex);
|
1710
|
-
bytes = h_get(mr->norms_cache, field);
|
1711
|
-
if (bytes != NULL) {
|
1712
|
-
memcpy(buf + offset, bytes, mr->max_doc);
|
1713
|
-
} else {
|
1714
|
-
IndexReader *reader;
|
1715
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1716
|
-
reader = mr->sub_readers[i];
|
1717
|
-
reader->get_norms_into(reader, field, buf, offset + mr->starts[i]);
|
1718
|
-
}
|
1719
|
-
}
|
1720
|
-
mutex_unlock(&ir->mutex);
|
1721
|
-
}
|
1722
|
-
|
1723
|
-
uchar *mr_get_norms(IndexReader *ir, char *field)
|
1724
|
-
{
|
1725
|
-
int i;
|
1726
|
-
GET_MR;
|
1727
|
-
uchar *bytes;
|
1728
|
-
IndexReader *reader;
|
1729
|
-
|
1730
|
-
mutex_lock(&ir->mutex);
|
1731
|
-
bytes = h_get(mr->norms_cache, field);
|
1732
|
-
if (bytes == NULL) {
|
1733
|
-
bytes = ALLOC_N(uchar, mr->max_doc);
|
1734
|
-
|
1735
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1736
|
-
reader = mr->sub_readers[i];
|
1737
|
-
reader->get_norms_into(reader, field, bytes, mr->starts[i]);
|
1738
|
-
}
|
1739
|
-
h_set(mr->norms_cache, field, bytes); /* update cache */
|
1740
|
-
}
|
1741
|
-
mutex_unlock(&ir->mutex);
|
1742
|
-
|
1743
|
-
return bytes;
|
1744
|
-
}
|
1745
|
-
|
1746
|
-
void mr_set_norm(IndexReader *ir, int doc_num, char *field, uchar val)
|
1747
|
-
{
|
1748
|
-
GET_READER(doc_num);
|
1749
|
-
h_del(mr->norms_cache, field); /* clear cache */
|
1750
|
-
ir_set_norm(reader, doc_num - mr->starts[i], field, val);
|
1751
|
-
}
|
1752
|
-
|
1753
|
-
TermEnum *mr_terms(IndexReader *ir)
|
1754
|
-
{
|
1755
|
-
GET_MR;
|
1756
|
-
return mte_create(mr->sub_readers, mr->starts, mr->rcnt, NULL);
|
1757
|
-
}
|
1758
|
-
|
1759
|
-
TermEnum *mr_terms_from(IndexReader *ir, Term *term)
|
1760
|
-
{
|
1761
|
-
GET_MR;
|
1762
|
-
return mte_create(mr->sub_readers, mr->starts, mr->rcnt, term);
|
1763
|
-
}
|
1764
|
-
|
1765
|
-
int mr_doc_freq(IndexReader *ir, Term *t)
|
1766
|
-
{
|
1767
|
-
int total = 0, i; /* sum freqs in segments */
|
1768
|
-
GET_MR;
|
1769
|
-
|
1770
|
-
IndexReader *reader;
|
1771
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1772
|
-
reader = mr->sub_readers[i];
|
1773
|
-
total += reader->doc_freq(reader, t);
|
1774
|
-
}
|
1775
|
-
return total;
|
1776
|
-
}
|
1777
|
-
|
1778
|
-
TermDocEnum *mr_term_docs(IndexReader *ir)
|
1779
|
-
{
|
1780
|
-
GET_MR;
|
1781
|
-
return mtde_create(mr->sub_readers, mr->starts, mr->rcnt);
|
1782
|
-
}
|
1783
|
-
|
1784
|
-
TermDocEnum *mr_term_positions(IndexReader *ir)
|
1785
|
-
{
|
1786
|
-
GET_MR;
|
1787
|
-
return mtpe_create(mr->sub_readers, mr->starts, mr->rcnt);
|
1788
|
-
}
|
1789
|
-
|
1790
|
-
void mr_delete_doc(IndexReader *ir, int doc_num)
|
1791
|
-
{
|
1792
|
-
GET_READER(doc_num);
|
1793
|
-
mr->num_docs_cache = -1; /* invalidate cache */
|
1794
|
-
|
1795
|
-
/* dispatch to segment reader */
|
1796
|
-
reader->do_delete_doc(reader, doc_num - mr->starts[i]);
|
1797
|
-
mr->has_deletions = true;
|
1798
|
-
}
|
1799
|
-
|
1800
|
-
bool mr_is_deleted(IndexReader *ir, int doc_num)
|
1801
|
-
{
|
1802
|
-
GET_READER(doc_num);
|
1803
|
-
return reader->is_deleted(reader, doc_num - mr->starts[i]);
|
1804
|
-
}
|
1805
|
-
|
1806
|
-
bool mr_has_norms(IndexReader *ir, char *field)
|
1807
|
-
{
|
1808
|
-
bool has_norms = false;
|
1809
|
-
int i;
|
1810
|
-
GET_MR;
|
1811
|
-
|
1812
|
-
IndexReader *reader;
|
1813
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1814
|
-
reader = mr->sub_readers[i];
|
1815
|
-
if (reader->has_norms(reader, field)) {
|
1816
|
-
has_norms = true;
|
1817
|
-
break;
|
1818
|
-
}
|
1819
|
-
}
|
1820
|
-
|
1821
|
-
return has_norms;
|
1822
|
-
}
|
1823
|
-
|
1824
|
-
bool mr_has_deletions(IndexReader *ir)
|
1825
|
-
{
|
1826
|
-
GET_MR;
|
1827
|
-
return mr->has_deletions;
|
1828
|
-
}
|
1829
|
-
|
1830
|
-
void mr_undelete_all(IndexReader *ir)
|
1831
|
-
{
|
1832
|
-
int i;
|
1833
|
-
GET_MR;
|
1834
|
-
IndexReader *reader;
|
1835
|
-
|
1836
|
-
mr->num_docs_cache = -1; /* invalidate cache */
|
1837
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1838
|
-
reader = mr->sub_readers[i];
|
1839
|
-
reader->do_undelete_all(reader);
|
1840
|
-
}
|
1841
|
-
mr->has_deletions = false;
|
1842
|
-
}
|
1843
|
-
|
1844
|
-
HashSet *mr_get_field_names(IndexReader *ir, int field_type)
|
1845
|
-
{
|
1846
|
-
int i;
|
1847
|
-
GET_MR;
|
1848
|
-
HashSet *field_set = hs_str_create(NULL);
|
1849
|
-
IndexReader *reader;
|
1850
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1851
|
-
reader = mr->sub_readers[i];
|
1852
|
-
hs_merge(field_set, reader->get_field_names(reader, field_type));
|
1853
|
-
}
|
1854
|
-
return field_set;
|
1855
|
-
}
|
1856
|
-
|
1857
|
-
void mr_commit(IndexReader *ir)
|
1858
|
-
{
|
1859
|
-
GET_MR;
|
1860
|
-
int i;
|
1861
|
-
IndexReader *reader;
|
1862
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1863
|
-
reader = mr->sub_readers[i];
|
1864
|
-
reader->do_commit(reader);
|
1865
|
-
}
|
1866
|
-
}
|
1867
|
-
|
1868
|
-
void mr_close(IndexReader *ir)
|
1869
|
-
{
|
1870
|
-
GET_MR;
|
1871
|
-
int i;
|
1872
|
-
IndexReader *reader;
|
1873
|
-
for (i = 0; i < mr->rcnt; i++) {
|
1874
|
-
reader = mr->sub_readers[i];
|
1875
|
-
ir_close(reader);
|
1876
|
-
}
|
1877
|
-
free(mr->sub_readers);
|
1878
|
-
h_destroy(mr->norms_cache);
|
1879
|
-
free(mr->starts);
|
1880
|
-
free(mr);
|
1881
|
-
}
|
1882
|
-
|
1883
|
-
IndexReader *mr_open(Store *store,
|
1884
|
-
SegmentInfos *sis,
|
1885
|
-
IndexReader **sub_readers,
|
1886
|
-
int rcnt)
|
1887
|
-
{
|
1888
|
-
int i;
|
1889
|
-
MultiReader *mr = ALLOC(MultiReader);
|
1890
|
-
IndexReader *sub_reader;
|
1891
|
-
IndexReader *ir;
|
1892
|
-
mr->sub_readers = sub_readers;
|
1893
|
-
mr->rcnt = rcnt;
|
1894
|
-
|
1895
|
-
mr->max_doc = 0;
|
1896
|
-
mr->num_docs_cache = -1;
|
1897
|
-
mr->has_deletions = false;
|
1898
|
-
|
1899
|
-
mr->starts = ALLOC_N(int, (rcnt+1));
|
1900
|
-
for (i = 0; i < rcnt; i++) {
|
1901
|
-
sub_reader = sub_readers[i];
|
1902
|
-
mr->starts[i] = mr->max_doc;
|
1903
|
-
mr->max_doc += sub_reader->max_doc(sub_reader); /* compute max_docs */
|
1904
|
-
|
1905
|
-
if (sub_reader->has_deletions(sub_reader)) {
|
1906
|
-
mr->has_deletions = true;
|
1907
|
-
}
|
1908
|
-
}
|
1909
|
-
mr->starts[rcnt] = mr->max_doc;
|
1910
|
-
mr->norms_cache = h_new_str(NULL, &free);
|
1911
|
-
|
1912
|
-
ir = ir_create(store, sis, true);
|
1913
|
-
ir->get_term_vector = &mr_get_term_vector;
|
1914
|
-
ir->get_term_vectors = &mr_get_term_vectors;
|
1915
|
-
ir->num_docs = &mr_num_docs;
|
1916
|
-
ir->max_doc = &mr_max_doc;
|
1917
|
-
ir->get_doc = &mr_get_doc;
|
1918
|
-
ir->get_norms_into = &mr_get_norms_into;
|
1919
|
-
ir->get_norms = &mr_get_norms;
|
1920
|
-
ir->get_norms_always = &mr_get_norms;
|
1921
|
-
ir->do_set_norm = &mr_set_norm;
|
1922
|
-
ir->terms = &mr_terms;
|
1923
|
-
ir->terms_from = &mr_terms_from;
|
1924
|
-
ir->doc_freq = &mr_doc_freq;
|
1925
|
-
ir->term_docs = &mr_term_docs;
|
1926
|
-
ir->term_positions = &mr_term_positions;
|
1927
|
-
ir->do_delete_doc = &mr_delete_doc;
|
1928
|
-
ir->is_deleted = &mr_is_deleted;
|
1929
|
-
ir->has_norms = &mr_has_norms;
|
1930
|
-
ir->has_deletions = &mr_has_deletions;
|
1931
|
-
ir->do_undelete_all = &mr_undelete_all;
|
1932
|
-
ir->get_field_names = &mr_get_field_names;
|
1933
|
-
ir->do_commit = &mr_commit;
|
1934
|
-
ir->do_close = &mr_close;
|
1935
|
-
ir->data = mr;
|
1936
|
-
|
1937
|
-
return ir;
|
1938
|
-
}
|
1939
|
-
|
1940
|
-
/****************************************************************************
|
1941
|
-
*
|
1942
|
-
* SegmentMergeInfo
|
1943
|
-
*
|
1944
|
-
****************************************************************************/
|
1945
|
-
|
1946
|
-
bool smi_lt(SegmentMergeInfo *smi1, SegmentMergeInfo *smi2)
|
1947
|
-
{
|
1948
|
-
int cmpres = tb_cmp(smi1->tb, smi2->tb);
|
1949
|
-
if (cmpres == 0) {
|
1950
|
-
return smi1->base < smi2->base;
|
1951
|
-
} else {
|
1952
|
-
return cmpres < 0;
|
1953
|
-
}
|
1954
|
-
}
|
1955
|
-
|
1956
|
-
int *smi_load_doc_map(SegmentMergeInfo *smi)
|
1957
|
-
{
|
1958
|
-
IndexReader *ir = smi->ir;
|
1959
|
-
if (ir->has_deletions(ir) && (smi->doc_map == NULL)) {
|
1960
|
-
int max_doc = ir->max_doc(ir);
|
1961
|
-
int j = 0, i;
|
1962
|
-
|
1963
|
-
smi->doc_map = ALLOC_N(int, max_doc);
|
1964
|
-
for (i = 0; i < max_doc; i++) {
|
1965
|
-
if (ir->is_deleted(ir, i)) {
|
1966
|
-
smi->doc_map[i] = -1;
|
1967
|
-
} else {
|
1968
|
-
smi->doc_map[i] = j++;
|
1969
|
-
}
|
1970
|
-
}
|
1971
|
-
}
|
1972
|
-
return smi->doc_map;
|
1973
|
-
}
|
1974
|
-
|
1975
|
-
SegmentMergeInfo *smi_create(int base, TermEnum *te, IndexReader *ir)
|
1976
|
-
{
|
1977
|
-
SegmentMergeInfo *smi = ALLOC(SegmentMergeInfo);
|
1978
|
-
smi->base = base;
|
1979
|
-
smi->ir = ir;
|
1980
|
-
smi->te = te;
|
1981
|
-
smi->tb = te->tb_curr;
|
1982
|
-
smi->postings = ir->term_positions(ir);
|
1983
|
-
smi->doc_map = NULL;
|
1984
|
-
return smi;
|
1985
|
-
}
|
1986
|
-
|
1987
|
-
void smi_destroy(SegmentMergeInfo *smi)
|
1988
|
-
{
|
1989
|
-
smi->postings->close(smi->postings);
|
1990
|
-
smi->te->close(smi->te);
|
1991
|
-
if (smi->doc_map != NULL)
|
1992
|
-
free(smi->doc_map);
|
1993
|
-
free(smi);
|
1994
|
-
}
|
1995
|
-
|
1996
|
-
TermBuffer *smi_next(SegmentMergeInfo *smi)
|
1997
|
-
{
|
1998
|
-
return (smi->tb = smi->te->next(smi->te));
|
1999
|
-
}
|
2000
|
-
|
2001
|
-
/****************************************************************************
|
2002
|
-
*
|
2003
|
-
* SegmentMerger
|
2004
|
-
*
|
2005
|
-
****************************************************************************/
|
2006
|
-
|
2007
|
-
SegmentMerger *sm_create(Store *store, char *name, int term_index_interval)
|
2008
|
-
{
|
2009
|
-
SegmentMerger *sm = ALLOC(SegmentMerger);
|
2010
|
-
sm->store = store;
|
2011
|
-
sm->name = estrdup(name);
|
2012
|
-
sm->readers = ary_create(config.merge_factor, (free_ft)&ir_close);
|
2013
|
-
sm->fis = NULL;
|
2014
|
-
sm->freq_out = NULL;
|
2015
|
-
sm->prox_out = NULL;
|
2016
|
-
sm->tiw = NULL;
|
2017
|
-
sm->queue = NULL;
|
2018
|
-
sm->ti = ti_create(0, 0, 0, 0);
|
2019
|
-
sm->term_index_interval = term_index_interval;
|
2020
|
-
sm->skip_buffer = ram_create_buffer();
|
2021
|
-
sm->skip_interval = -1;
|
2022
|
-
return sm;
|
2023
|
-
}
|
2024
|
-
|
2025
|
-
void sm_close(SegmentMerger *sm)
|
2026
|
-
{
|
2027
|
-
int i;
|
2028
|
-
if (sm->freq_out != NULL) os_close(sm->freq_out);
|
2029
|
-
if (sm->prox_out != NULL) os_close(sm->prox_out);
|
2030
|
-
if (sm->tiw != NULL) {
|
2031
|
-
for (i = 0; i < sm->terms_buf_size; i++) {
|
2032
|
-
free(sm->terms_buf[i].text);
|
2033
|
-
}
|
2034
|
-
free(sm->terms_buf);
|
2035
|
-
tiw_close(sm->tiw);
|
2036
|
-
}
|
2037
|
-
if (sm->queue != NULL) pq_destroy(sm->queue);
|
2038
|
-
sm->freq_out = NULL;
|
2039
|
-
sm->prox_out = NULL;
|
2040
|
-
sm->tiw = NULL;
|
2041
|
-
sm->queue = NULL;
|
2042
|
-
}
|
2043
|
-
|
2044
|
-
void sm_destroy(SegmentMerger *sm)
|
2045
|
-
{
|
2046
|
-
if (sm->fis != NULL) fis_destroy(sm->fis);
|
2047
|
-
ary_destroy(sm->readers);
|
2048
|
-
sm_close(sm);
|
2049
|
-
free(sm->name);
|
2050
|
-
ti_destroy(sm->ti);
|
2051
|
-
ram_destroy_buffer(sm->skip_buffer);
|
2052
|
-
free(sm);
|
2053
|
-
}
|
2054
|
-
|
2055
|
-
void sm_add(SegmentMerger *sm, IndexReader *ir)
|
2056
|
-
{
|
2057
|
-
ary_append(sm->readers, ir);
|
2058
|
-
}
|
2059
|
-
|
2060
|
-
static inline void sm_add_indexed(IndexReader *ir,
|
2061
|
-
FieldInfos *fis,
|
2062
|
-
HashSet *fields,
|
2063
|
-
bool store_tv,
|
2064
|
-
bool store_pos,
|
2065
|
-
bool store_offset)
|
2066
|
-
{
|
2067
|
-
int i;
|
2068
|
-
char *field;
|
2069
|
-
for (i = 0; i < fields->size; i++) {
|
2070
|
-
field = (char *)fields->elems[i];
|
2071
|
-
fis_add(fis, field, true, store_tv, store_pos, store_offset,
|
2072
|
-
!ir->has_norms(ir, field));
|
2073
|
-
}
|
2074
|
-
hs_destroy(fields);
|
2075
|
-
}
|
2076
|
-
|
2077
|
-
int sm_merge_fields(SegmentMerger *sm)
|
2078
|
-
{
|
2079
|
-
int i, j, maxdoc;
|
2080
|
-
FieldInfos *fis = sm->fis = fis_create();
|
2081
|
-
int doc_count = 0;
|
2082
|
-
Document *doc;
|
2083
|
-
FieldsWriter *fw;
|
2084
|
-
|
2085
|
-
for (i = 0; i < sm->readers->size; i++) {
|
2086
|
-
IndexReader *ir = sm->readers->elems[i];
|
2087
|
-
|
2088
|
-
sm_add_indexed(ir, fis,
|
2089
|
-
ir->get_field_names(ir, IR_TERM_VECTOR_WITH_POSITION_OFFSET),
|
2090
|
-
true, true, true);
|
2091
|
-
sm_add_indexed(ir, fis,
|
2092
|
-
ir->get_field_names(ir, IR_TERM_VECTOR_WITH_POSITION),
|
2093
|
-
true, true, false);
|
2094
|
-
sm_add_indexed(ir, fis,
|
2095
|
-
ir->get_field_names(ir, IR_TERM_VECTOR_WITH_OFFSET),
|
2096
|
-
true, false, true);
|
2097
|
-
sm_add_indexed(ir, fis, ir->get_field_names(ir, IR_TERM_VECTOR),
|
2098
|
-
true, false, false);
|
2099
|
-
sm_add_indexed(ir, fis, ir->get_field_names(ir, IR_INDEXED),
|
2100
|
-
false, false, false);
|
2101
|
-
fis_add_fields(fis, ir->get_field_names(ir, IR_UNINDEXED),
|
2102
|
-
false, false, false, false, false);
|
2103
|
-
}
|
2104
|
-
fis_write(fis, sm->store, sm->name, ".fnm");
|
2105
|
-
|
2106
|
-
/* merge field values */
|
2107
|
-
fw = fw_open(sm->store, sm->name, fis);
|
2108
|
-
|
2109
|
-
TRY
|
2110
|
-
for (i = 0; i < sm->readers->size; i++) {
|
2111
|
-
IndexReader *ir = sm->readers->elems[i];
|
2112
|
-
maxdoc = ir->max_doc(ir);
|
2113
|
-
for (j = 0; j < maxdoc; j++) {
|
2114
|
-
if (!ir->is_deleted(ir, j)) { /* skip deleted docs */
|
2115
|
-
doc = ir->get_doc(ir, j);
|
2116
|
-
fw_add_doc(fw, doc);
|
2117
|
-
doc_destroy(doc);
|
2118
|
-
doc_count++;
|
2119
|
-
}
|
2120
|
-
}
|
2121
|
-
}
|
2122
|
-
XFINALLY
|
2123
|
-
fw_close(fw);
|
2124
|
-
XENDTRY
|
2125
|
-
return doc_count;
|
2126
|
-
}
|
2127
|
-
|
2128
|
-
void sm_reset_skip(SegmentMerger *sm)
|
2129
|
-
{
|
2130
|
-
ramo_reset(sm->skip_buffer);
|
2131
|
-
sm->last_skip_doc = 0;
|
2132
|
-
sm->last_skip_freq_pointer = os_pos(sm->freq_out);
|
2133
|
-
sm->last_skip_prox_pointer = os_pos(sm->prox_out);
|
2134
|
-
}
|
2135
|
-
|
2136
|
-
inline void sm_buffer_skip(SegmentMerger *sm, int doc)
|
2137
|
-
{
|
2138
|
-
int freq_pointer = os_pos(sm->freq_out);
|
2139
|
-
int prox_pointer = os_pos(sm->prox_out);
|
2140
|
-
|
2141
|
-
os_write_vint(sm->skip_buffer, doc - sm->last_skip_doc);
|
2142
|
-
os_write_vint(sm->skip_buffer, freq_pointer - sm->last_skip_freq_pointer);
|
2143
|
-
os_write_vint(sm->skip_buffer, prox_pointer - sm->last_skip_prox_pointer);
|
2144
|
-
|
2145
|
-
sm->last_skip_doc = doc;
|
2146
|
-
sm->last_skip_freq_pointer = freq_pointer;
|
2147
|
-
sm->last_skip_prox_pointer = prox_pointer;
|
2148
|
-
}
|
2149
|
-
|
2150
|
-
int sm_append_postings(SegmentMerger *sm, SegmentMergeInfo **smis, int cnt)
|
2151
|
-
{
|
2152
|
-
int i, j;
|
2153
|
-
int last_doc = 0, base, doc, doc_code, freq, last_position, position;
|
2154
|
-
int *doc_map = NULL;
|
2155
|
-
int df = 0; /* number of docs w/ term */
|
2156
|
-
TermDocEnum *postings;
|
2157
|
-
SegmentMergeInfo *smi;
|
2158
|
-
sm_reset_skip(sm);
|
2159
|
-
for (i = 0; i < cnt; i++) {
|
2160
|
-
smi = smis[i];
|
2161
|
-
postings = smi->postings;
|
2162
|
-
base = smi->base;
|
2163
|
-
doc_map = smi_load_doc_map(smi);
|
2164
|
-
|
2165
|
-
stde_seek_ti(postings, smi->te->ti_curr);
|
2166
|
-
while (postings->next(postings)) {
|
2167
|
-
doc = postings->doc_num(postings);
|
2168
|
-
if (doc_map != NULL) {
|
2169
|
-
doc = doc_map[doc]; /* work around deletions */
|
2170
|
-
}
|
2171
|
-
doc += base; /* convert to merged space */
|
2172
|
-
|
2173
|
-
if (doc < last_doc) {
|
2174
|
-
RAISE(STATE_ERROR, DOC_ORDER_ERROR_MSG);
|
2175
|
-
}
|
2176
|
-
|
2177
|
-
df++;
|
2178
|
-
|
2179
|
-
if ((df % sm->skip_interval) == 0) {
|
2180
|
-
sm_buffer_skip(sm, last_doc);
|
2181
|
-
}
|
2182
|
-
|
2183
|
-
doc_code = (doc - last_doc) << 1; /* use low bit to flag freq=1 */
|
2184
|
-
last_doc = doc;
|
2185
|
-
|
2186
|
-
freq = postings->freq(postings);
|
2187
|
-
if (freq == 1) {
|
2188
|
-
os_write_vint(sm->freq_out, doc_code | 1); /* write doc & freq=1 */
|
2189
|
-
} else {
|
2190
|
-
os_write_vint(sm->freq_out, doc_code); /* write doc */
|
2191
|
-
os_write_vint(sm->freq_out, freq); /* write freqency in doc */
|
2192
|
-
}
|
2193
|
-
|
2194
|
-
|
2195
|
-
last_position = 0; /* write position deltas */
|
2196
|
-
for (j = 0; j < freq; j++) {
|
2197
|
-
position = postings->next_position(postings);
|
2198
|
-
os_write_vint(sm->prox_out, position - last_position);
|
2199
|
-
last_position = position;
|
2200
|
-
}
|
2201
|
-
}
|
2202
|
-
}
|
2203
|
-
return df;
|
2204
|
-
}
|
2205
|
-
|
2206
|
-
int sm_write_skip(SegmentMerger *sm)
|
2207
|
-
{
|
2208
|
-
int skip_pointer = os_pos(sm->freq_out);
|
2209
|
-
ramo_write_to(sm->skip_buffer, sm->freq_out);
|
2210
|
-
return skip_pointer;
|
2211
|
-
}
|
2212
|
-
|
2213
|
-
Term *sm_tb_to_term(SegmentMerger *sm, TermBuffer *tb)
|
2214
|
-
{
|
2215
|
-
int index = sm->terms_buf_pointer % sm->terms_buf_size;
|
2216
|
-
sm->terms_buf_pointer++;
|
2217
|
-
sm->terms_buf[index].field = tb->field;
|
2218
|
-
strcpy(sm->terms_buf[index].text, tb->text);
|
2219
|
-
return &(sm->terms_buf[index]);
|
2220
|
-
}
|
2221
|
-
|
2222
|
-
void sm_merge_term_info(SegmentMerger *sm, SegmentMergeInfo **smis, int cnt)
|
2223
|
-
{
|
2224
|
-
int freq_pointer = os_pos(sm->freq_out);
|
2225
|
-
int prox_pointer = os_pos(sm->prox_out);
|
2226
|
-
|
2227
|
-
int df = sm_append_postings(sm, smis, cnt); /* append posting data */
|
2228
|
-
|
2229
|
-
int skip_pointer = sm_write_skip(sm);
|
2230
|
-
|
2231
|
-
if (df > 0) {
|
2232
|
-
/* add an entry to the dictionary with pointers to prox and freq files */
|
2233
|
-
ti_set(sm->ti, df, freq_pointer, prox_pointer, (skip_pointer - freq_pointer));
|
2234
|
-
tiw_add(sm->tiw, sm_tb_to_term(sm, smis[0]->tb), sm->ti);
|
2235
|
-
}
|
2236
|
-
}
|
2237
|
-
|
2238
|
-
void sm_merge_term_infos(SegmentMerger *sm)
|
2239
|
-
{
|
2240
|
-
int base = 0;
|
2241
|
-
int i, match_size;
|
2242
|
-
IndexReader *ir;
|
2243
|
-
TermEnum *te;
|
2244
|
-
SegmentMergeInfo *smi, *top, **match;
|
2245
|
-
TermBuffer *tb;
|
2246
|
-
|
2247
|
-
for (i = 0; i < sm->readers->size; i++) {
|
2248
|
-
ir = sm->readers->elems[i];
|
2249
|
-
te = ir->terms(ir);
|
2250
|
-
smi = smi_create(base, te, ir);
|
2251
|
-
base += ir->num_docs(ir);
|
2252
|
-
if (smi_next(smi) != NULL) {
|
2253
|
-
pq_push(sm->queue, smi); /* initialize @queue */
|
2254
|
-
} else {
|
2255
|
-
smi_destroy(smi);
|
2256
|
-
}
|
2257
|
-
}
|
2258
|
-
|
2259
|
-
match = ALLOC_N(SegmentMergeInfo *, sm->readers->size);
|
2260
|
-
|
2261
|
-
while (sm->queue->count > 0) {
|
2262
|
-
/*
|
2263
|
-
for (i = 1; i <= sm->queue->count; i++) {
|
2264
|
-
printf("<{%s:%s}>", ((SegmentMergeInfo *)sm->queue->heap[i])->tb->field,
|
2265
|
-
((SegmentMergeInfo *)sm->queue->heap[i])->tb->text);
|
2266
|
-
}printf("\n\n");
|
2267
|
-
*/
|
2268
|
-
match_size = 0; /* pop matching terms */
|
2269
|
-
match[match_size] = pq_pop(sm->queue);
|
2270
|
-
match_size++;
|
2271
|
-
tb = match[0]->tb;
|
2272
|
-
top = pq_top(sm->queue);
|
2273
|
-
while ((top != NULL) && (tb_cmp(tb, top->tb) == 0)) {
|
2274
|
-
match[match_size] = pq_pop(sm->queue);
|
2275
|
-
match_size++;
|
2276
|
-
top = pq_top(sm->queue);
|
2277
|
-
}
|
2278
|
-
|
2279
|
-
/* printf(">%s:%s<\n", match[0]->tb->field, match[0]->tb->text); */
|
2280
|
-
sm_merge_term_info(sm, match, match_size); /* add new TermInfo */
|
2281
|
-
|
2282
|
-
while (match_size > 0) {
|
2283
|
-
match_size--;
|
2284
|
-
smi = match[match_size];
|
2285
|
-
if (smi_next(smi) != NULL) {
|
2286
|
-
pq_push(sm->queue, smi); /* restore queue */
|
2287
|
-
} else {
|
2288
|
-
smi_destroy(smi); /* done with a segment */
|
2289
|
-
}
|
2290
|
-
}
|
2291
|
-
}
|
2292
|
-
free(match);
|
2293
|
-
}
|
2294
|
-
|
2295
|
-
void sm_merge_terms(SegmentMerger *sm)
|
2296
|
-
{
|
2297
|
-
int i;
|
2298
|
-
char fname[SEGMENT_NAME_MAX_LENGTH];
|
2299
|
-
|
2300
|
-
TRY
|
2301
|
-
sprintf(fname, "%s.frq", sm->name);
|
2302
|
-
sm->freq_out = sm->store->create_output(sm->store, fname);
|
2303
|
-
sprintf(fname, "%s.prx", sm->name);
|
2304
|
-
sm->prox_out = sm->store->create_output(sm->store, fname);
|
2305
|
-
sm->tiw = tiw_open(sm->store, sm->name, sm->fis, sm->term_index_interval);
|
2306
|
-
/* terms_buf_pointer holds a buffer of terms since the TermInfosWriter needs
|
2307
|
-
* to keep the last index_interval terms so that it can compare the last term
|
2308
|
-
* put in the index with the next one. So the size of the buffer must by
|
2309
|
-
* index_interval + 2. */
|
2310
|
-
sm->terms_buf_pointer = 0;
|
2311
|
-
sm->terms_buf_size = sm->tiw->index_interval + 2;
|
2312
|
-
sm->terms_buf = ALLOC_N(Term, sm->terms_buf_size);
|
2313
|
-
for (i = 0; i < sm->terms_buf_size; i++) {
|
2314
|
-
sm->terms_buf[i].field = NULL;
|
2315
|
-
sm->terms_buf[i].text = ALLOC_N(char, MAX_WORD_SIZE);
|
2316
|
-
}
|
2317
|
-
sm->skip_interval = sm->tiw->skip_interval;
|
2318
|
-
sm->queue = pq_create(sm->readers->size, (lt_ft)&smi_lt);
|
2319
|
-
|
2320
|
-
sm_merge_term_infos(sm);
|
2321
|
-
|
2322
|
-
XFINALLY
|
2323
|
-
sm_close(sm);
|
2324
|
-
XENDTRY
|
2325
|
-
}
|
2326
|
-
|
2327
|
-
void sm_merge_norms(SegmentMerger *sm)
|
2328
|
-
{
|
2329
|
-
int i, j, k, max_doc;
|
2330
|
-
uchar *norm_buf;
|
2331
|
-
FieldInfo *fi;
|
2332
|
-
OutStream *os;
|
2333
|
-
char fname[SEGMENT_NAME_MAX_LENGTH];
|
2334
|
-
IndexReader *ir;
|
2335
|
-
for (i = 0; i < sm->fis->fcnt; i++) {
|
2336
|
-
fi = sm->fis->by_number[i];
|
2337
|
-
if (fi->is_indexed && !fi->omit_norms) {
|
2338
|
-
sprintf(fname, "%s.f%d", sm->name, i);
|
2339
|
-
os = sm->store->create_output(sm->store, fname);
|
2340
|
-
TRY
|
2341
|
-
for (j = 0; j < sm->readers->size; j++) {
|
2342
|
-
ir = sm->readers->elems[j];
|
2343
|
-
max_doc = ir->max_doc(ir);
|
2344
|
-
norm_buf = ALLOC_N(uchar, max_doc);
|
2345
|
-
memset(norm_buf, 0, sizeof(uchar) * max_doc);
|
2346
|
-
ir->get_norms_into(ir, fi->name, norm_buf, 0);
|
2347
|
-
for (k = 0; k < max_doc; k++) {
|
2348
|
-
if (!ir->is_deleted(ir, k)) {
|
2349
|
-
os_write_byte(os, norm_buf[k]);
|
2350
|
-
}
|
2351
|
-
}
|
2352
|
-
free(norm_buf);
|
2353
|
-
}
|
2354
|
-
XFINALLY
|
2355
|
-
os_close(os);
|
2356
|
-
XENDTRY
|
2357
|
-
}
|
2358
|
-
}
|
2359
|
-
}
|
2360
|
-
|
2361
|
-
void sm_merge_vectors(SegmentMerger *sm)
|
2362
|
-
{
|
2363
|
-
int i, j, max_doc;
|
2364
|
-
TermVectorsWriter *tvw = tvw_open(sm->store, sm->name, sm->fis);
|
2365
|
-
IndexReader *ir;
|
2366
|
-
Array *tvs;
|
2367
|
-
TRY
|
2368
|
-
for (i = 0; i < sm->readers->size; i++) {
|
2369
|
-
ir = sm->readers->elems[i];
|
2370
|
-
max_doc = ir->max_doc(ir);
|
2371
|
-
for (j = 0; j < max_doc; j++) {
|
2372
|
-
/* skip deleted docs */
|
2373
|
-
if (! ir->is_deleted(ir, j)) {
|
2374
|
-
tvs = ir->get_term_vectors(ir, j);
|
2375
|
-
if (tvs) {
|
2376
|
-
tvw_add_all_doc_vectors(tvw, tvs);
|
2377
|
-
ary_destroy(tvs);
|
2378
|
-
}
|
2379
|
-
}
|
2380
|
-
}
|
2381
|
-
}
|
2382
|
-
XFINALLY
|
2383
|
-
tvw_close(tvw);
|
2384
|
-
XENDTRY
|
2385
|
-
}
|
2386
|
-
|
2387
|
-
int sm_merge(SegmentMerger *sm)
|
2388
|
-
{
|
2389
|
-
int doc_count = sm_merge_fields(sm);
|
2390
|
-
sm_merge_terms(sm);
|
2391
|
-
sm_merge_norms(sm);
|
2392
|
-
if (fis_has_vectors(sm->fis))
|
2393
|
-
sm_merge_vectors(sm);
|
2394
|
-
return doc_count;
|
2395
|
-
}
|
2396
|
-
|
2397
|
-
Array *sm_create_compound_file(SegmentMerger *sm, char *file_name)
|
2398
|
-
{
|
2399
|
-
Array *files = ary_create(0, &free);
|
2400
|
-
CompoundWriter *cw = open_cw(sm->store, file_name);
|
2401
|
-
FieldInfo *fi;
|
2402
|
-
char fname[SEGMENT_NAME_MAX_LENGTH];
|
2403
|
-
|
2404
|
-
int i;
|
2405
|
-
for (i = 0; i < NELEMS(COMPOUND_EXTENSIONS); i++) {
|
2406
|
-
sprintf(fname, "%s.%s", sm->name, COMPOUND_EXTENSIONS[i]);
|
2407
|
-
ary_append(files, estrdup(fname));
|
2408
|
-
}
|
2409
|
-
|
2410
|
-
/* Field norm files */
|
2411
|
-
for (i = 0; i < sm->fis->fcnt; i++) {
|
2412
|
-
fi = sm->fis->by_number[i];
|
2413
|
-
if (fi->is_indexed && !fi->omit_norms) {
|
2414
|
-
sprintf(fname, "%s.f%d", sm->name, i);
|
2415
|
-
ary_append(files, estrdup(fname));
|
2416
|
-
}
|
2417
|
-
}
|
2418
|
-
|
2419
|
-
/* Vector files */
|
2420
|
-
if (fis_has_vectors(sm->fis)) {
|
2421
|
-
for (i = 0; i < NELEMS(VECTOR_EXTENSIONS); i++) {
|
2422
|
-
sprintf(fname, "%s.%s", sm->name, VECTOR_EXTENSIONS[i]);
|
2423
|
-
ary_append(files, estrdup(fname));
|
2424
|
-
}
|
2425
|
-
}
|
2426
|
-
|
2427
|
-
/* Now merge all added files */
|
2428
|
-
for (i = 0; i < files->size; i++) {
|
2429
|
-
cw_add_file(cw, (char *)files->elems[i]);
|
2430
|
-
}
|
2431
|
-
|
2432
|
-
/* Perform the merge */
|
2433
|
-
cw_close(cw);
|
2434
|
-
|
2435
|
-
return files;
|
2436
|
-
}
|
2437
|
-
|
2438
|
-
/****************************************************************************
|
2439
|
-
*
|
2440
|
-
* IndexReader
|
2441
|
-
*
|
2442
|
-
****************************************************************************/
|
2443
|
-
|
2444
|
-
void ir_acquire_not_necessary(IndexReader *ir) {}
|
2445
|
-
void ir_acquire_write_lock(IndexReader *ir)
|
2446
|
-
{
|
2447
|
-
if (ir->is_stale)
|
2448
|
-
RAISE(STATE_ERROR, STALE_READER_ERROR_MSG);
|
2449
|
-
|
2450
|
-
if (ir->write_lock == NULL) {
|
2451
|
-
ir->write_lock = ir->store->open_lock(ir->store, WRITE_LOCK_NAME);
|
2452
|
-
if (!ir->write_lock->obtain(ir->write_lock)) /* obtain write lock */
|
2453
|
-
RAISE(STATE_ERROR, WRITE_LOCK_ERROR_MSG);
|
2454
|
-
|
2455
|
-
/* we have to check whether index has changed since this reader was opened.
|
2456
|
-
* if so, this reader is no longer valid for deletion */
|
2457
|
-
if (sis_read_current_version(ir->store) > ir->sis->version) {
|
2458
|
-
ir->is_stale = true;
|
2459
|
-
ir->write_lock->release(ir->write_lock);
|
2460
|
-
ir->store->close_lock(ir->write_lock);
|
2461
|
-
ir->write_lock = NULL;
|
2462
|
-
RAISE(STATE_ERROR, STALE_READER_ERROR_MSG);
|
2463
|
-
}
|
2464
|
-
}
|
2465
|
-
}
|
2466
|
-
|
2467
|
-
IndexReader *ir_create(Store *store, SegmentInfos *sis, int is_owner)
|
2468
|
-
{
|
2469
|
-
IndexReader *ir = ALLOC(IndexReader);
|
2470
|
-
|
2471
|
-
mutex_init(&ir->mutex, NULL);
|
2472
|
-
ir->is_owner = is_owner;
|
2473
|
-
if (is_owner) {
|
2474
|
-
ir->acquire_write_lock = &ir_acquire_write_lock;
|
2475
|
-
} else {
|
2476
|
-
ir->acquire_write_lock = &ir_acquire_not_necessary;
|
2477
|
-
}
|
2478
|
-
|
2479
|
-
ir->store = store;
|
2480
|
-
ir->sis = sis;
|
2481
|
-
ir->has_changes = false;
|
2482
|
-
ir->is_stale = false;
|
2483
|
-
ir->write_lock = NULL;
|
2484
|
-
ir->cache = NULL;
|
2485
|
-
ir->sort_cache = NULL;
|
2486
|
-
return ir;
|
2487
|
-
}
|
2488
|
-
|
2489
|
-
/**
|
2490
|
-
* Will keep a reference to the store. To let this method delete the store
|
2491
|
-
* make sure you deref the store that you pass to it
|
2492
|
-
*/
|
2493
|
-
IndexReader *ir_open(Store *store)
|
2494
|
-
{
|
2495
|
-
int i;
|
2496
|
-
IndexReader *ir;
|
2497
|
-
SegmentInfos *sis;
|
2498
|
-
|
2499
|
-
mutex_lock(&store->mutex);
|
2500
|
-
sis = sis_create();
|
2501
|
-
sis_read(sis, store);
|
2502
|
-
if (sis->scnt == 1) {
|
2503
|
-
ir = sr_open(sis, 0, true);
|
2504
|
-
} else {
|
2505
|
-
IndexReader **readers = ALLOC_N(IndexReader *, sis->scnt);
|
2506
|
-
for (i = 0; i < sis->scnt; i++) {
|
2507
|
-
readers[i] = sr_open(sis, i, false);
|
2508
|
-
}
|
2509
|
-
ref(store);
|
2510
|
-
ir = mr_open(store, sis, readers, sis->scnt);
|
2511
|
-
}
|
2512
|
-
mutex_unlock(&store->mutex);
|
2513
|
-
return ir;
|
2514
|
-
}
|
2515
|
-
|
2516
|
-
bool ir_index_exists(Store *store)
|
2517
|
-
{
|
2518
|
-
return store->exists(store, "segments");
|
2519
|
-
}
|
2520
|
-
|
2521
|
-
void ir_set_norm(IndexReader *ir, int doc_num, char *field, uchar val)
|
2522
|
-
{
|
2523
|
-
mutex_lock(&ir->mutex);
|
2524
|
-
ir->acquire_write_lock(ir);
|
2525
|
-
ir->do_set_norm(ir, doc_num, field, val);
|
2526
|
-
ir->has_changes = true;
|
2527
|
-
mutex_unlock(&ir->mutex);
|
2528
|
-
}
|
2529
|
-
|
2530
|
-
void ir_undelete_all(IndexReader *ir)
|
2531
|
-
{
|
2532
|
-
mutex_lock(&ir->mutex);
|
2533
|
-
ir->acquire_write_lock(ir);
|
2534
|
-
ir->do_undelete_all(ir);
|
2535
|
-
ir->has_changes = true;
|
2536
|
-
mutex_unlock(&ir->mutex);
|
2537
|
-
}
|
2538
|
-
|
2539
|
-
void ir_delete_doc(IndexReader *ir, int doc_num)
|
2540
|
-
{
|
2541
|
-
mutex_lock(&ir->mutex);
|
2542
|
-
ir->acquire_write_lock(ir);
|
2543
|
-
ir->do_delete_doc(ir, doc_num);
|
2544
|
-
ir->has_changes = true;
|
2545
|
-
mutex_unlock(&ir->mutex);
|
2546
|
-
}
|
2547
|
-
|
2548
|
-
Document *ir_get_doc_with_term(IndexReader *ir, Term *term)
|
2549
|
-
{
|
2550
|
-
TermDocEnum *tde = ir_term_docs_for(ir, term);
|
2551
|
-
Document *doc = NULL;
|
2552
|
-
|
2553
|
-
if (!tde) return NULL;
|
2554
|
-
|
2555
|
-
if (tde->next(tde)) {
|
2556
|
-
doc = ir->get_doc(ir, tde->doc_num(tde));
|
2557
|
-
}
|
2558
|
-
tde->close(tde);
|
2559
|
-
return doc;
|
2560
|
-
}
|
2561
|
-
|
2562
|
-
TermDocEnum *ir_term_docs_for(IndexReader *ir, Term *term)
|
2563
|
-
{
|
2564
|
-
TermDocEnum *tde = ir->term_docs(ir);
|
2565
|
-
tde->seek(tde, term);
|
2566
|
-
return tde;
|
2567
|
-
}
|
2568
|
-
|
2569
|
-
TermDocEnum *ir_term_positions_for(IndexReader *ir, Term *term)
|
2570
|
-
{
|
2571
|
-
TermDocEnum *tde = ir->term_positions(ir);
|
2572
|
-
tde->seek(tde, term);
|
2573
|
-
return tde;
|
2574
|
-
}
|
2575
|
-
|
2576
|
-
void ir_commit_internal(IndexReader *ir)
|
2577
|
-
{
|
2578
|
-
if (ir->has_changes) {
|
2579
|
-
if (ir->is_owner) {
|
2580
|
-
Lock *commit_lock;
|
2581
|
-
|
2582
|
-
mutex_lock(&ir->store->mutex);
|
2583
|
-
commit_lock = ir->store->open_lock(ir->store, COMMIT_LOCK_NAME);
|
2584
|
-
if (!commit_lock->obtain(commit_lock)) { /* obtain write lock */
|
2585
|
-
RAISE(STATE_ERROR, COMMIT_LOCK_ERROR_MSG);
|
2586
|
-
}
|
2587
|
-
|
2588
|
-
ir->do_commit(ir);
|
2589
|
-
sis_write(ir->sis, ir->store);
|
2590
|
-
|
2591
|
-
commit_lock->release(commit_lock);
|
2592
|
-
ir->store->close_lock(commit_lock);
|
2593
|
-
mutex_unlock(&ir->store->mutex);
|
2594
|
-
|
2595
|
-
if (ir->write_lock != NULL) {
|
2596
|
-
ir->write_lock->release(ir->write_lock); /* release write lock */
|
2597
|
-
ir->store->close_lock(ir->write_lock);
|
2598
|
-
ir->write_lock = NULL;
|
2599
|
-
}
|
2600
|
-
} else {
|
2601
|
-
ir->do_commit(ir);
|
2602
|
-
}
|
2603
|
-
ir->has_changes = false;
|
2604
|
-
}
|
2605
|
-
}
|
2606
|
-
|
2607
|
-
void ir_commit(IndexReader *ir)
|
2608
|
-
{
|
2609
|
-
mutex_lock(&ir->mutex);
|
2610
|
-
ir_commit_internal(ir);
|
2611
|
-
mutex_unlock(&ir->mutex);
|
2612
|
-
}
|
2613
|
-
|
2614
|
-
void ir_close(IndexReader *ir)
|
2615
|
-
{
|
2616
|
-
mutex_lock(&ir->mutex);
|
2617
|
-
ir_commit_internal(ir);
|
2618
|
-
ir->do_close(ir);
|
2619
|
-
store_deref(ir->store);
|
2620
|
-
if (ir->is_owner) {
|
2621
|
-
sis_destroy(ir->sis);
|
2622
|
-
}
|
2623
|
-
if (ir->cache) {
|
2624
|
-
h_destroy(ir->cache);
|
2625
|
-
}
|
2626
|
-
if (ir->sort_cache) {
|
2627
|
-
h_destroy(ir->sort_cache);
|
2628
|
-
}
|
2629
|
-
|
2630
|
-
mutex_destroy(&ir->mutex);
|
2631
|
-
free(ir);
|
2632
|
-
}
|
2633
|
-
|
2634
|
-
/**
|
2635
|
-
* Don't call this method if the cache already exists
|
2636
|
-
**/
|
2637
|
-
void ir_add_cache(IndexReader *ir)
|
2638
|
-
{
|
2639
|
-
ir->cache = co_hsh_create();
|
2640
|
-
}
|
2641
|
-
|
2642
|
-
bool ir_is_latest(IndexReader *ir)
|
2643
|
-
{
|
2644
|
-
bool is_latest = false;
|
2645
|
-
Lock *commit_lock = ir->store->open_lock(ir->store, COMMIT_LOCK_NAME);
|
2646
|
-
if (!commit_lock->obtain(commit_lock)) {
|
2647
|
-
ir->store->close_lock(commit_lock);
|
2648
|
-
RAISE(STATE_ERROR, COMMIT_LOCK_ERROR_MSG);
|
2649
|
-
}
|
2650
|
-
TRY
|
2651
|
-
is_latest = (sis_read_current_version(ir->store) == ir->sis->version);
|
2652
|
-
XFINALLY
|
2653
|
-
commit_lock->release(commit_lock);
|
2654
|
-
ir->store->close_lock(commit_lock);
|
2655
|
-
XENDTRY
|
2656
|
-
return is_latest;
|
2657
|
-
}
|
2658
|
-
|