1 | /**************************************************************************
|
---|
2 | *
|
---|
3 | * ivf.pass2.cpp -- Memory efficient pass 2 inversion
|
---|
4 | * Copyright (C) 1999 Rodger McNab
|
---|
5 | *
|
---|
6 | * This program is free software; you can redistribute it and/or modify
|
---|
7 | * it under the terms of the GNU General Public License as published by
|
---|
8 | * the Free Software Foundation; either version 2 of the License, or
|
---|
9 | * (at your option) any later version.
|
---|
10 | *
|
---|
11 | * This program is distributed in the hope that it will be useful,
|
---|
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
14 | * GNU General Public License for more details.
|
---|
15 | *
|
---|
16 | * You should have received a copy of the GNU General Public License
|
---|
17 | * along with this program; if not, write to the Free Software
|
---|
18 | * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
|
---|
19 | *
|
---|
20 | **************************************************************************/
|
---|
21 |
|
---|
22 | #define _XOPEN_SOURCE 1
|
---|
23 | // This was added for Solaris, but it makes things worse on Solaris for me...
|
---|
24 | // #define _XOPEN_SOURCE_EXTENDED 1
|
---|
25 |
|
---|
26 | #include <stdio.h>
|
---|
27 |
|
---|
28 | #if defined __WIN32__
|
---|
29 | # include <process.h>
|
---|
30 | # include <io.h>
|
---|
31 | # define getpid _getpid
|
---|
32 | # define unlink _unlink
|
---|
33 | #else
|
---|
34 | # include <unistd.h>
|
---|
35 | # include "non_ansi.h"
|
---|
36 | #endif
|
---|
37 |
|
---|
38 | #include "UCArray.h"
|
---|
39 | #include "sysfuncs.h"
|
---|
40 | #include "mg_files.h"
|
---|
41 | #include "invf.h"
|
---|
42 | #include "mg.h"
|
---|
43 | #include "build.h"
|
---|
44 | #include "locallib.h"
|
---|
45 | #include "bitio_m_random.h"
|
---|
46 | #include "bitio_m_stdio.h"
|
---|
47 | #include "bitio_m_mems.h"
|
---|
48 | #include "bitio_gen.h"
|
---|
49 | #include <stdio.h>
|
---|
50 | #include "words.h"
|
---|
51 | #include "messages.h"
|
---|
52 | #include "netorder.h"
|
---|
53 | #include "FIvfLevelInfo.h"
|
---|
54 | #include "perf_hash.h"
|
---|
55 | #include "string.h"
|
---|
56 |
|
---|
57 | #include "longlong.h"
|
---|
58 |
|
---|
59 | #if defined(GSDL_USE_OBJECTSPACE)
|
---|
60 | # include <ospace\std\map>
|
---|
61 | #elif defined(GSDL_USE_STL_H)
|
---|
62 | # include <map.h>
|
---|
63 | #else
|
---|
64 | # include <map>
|
---|
65 | #endif
|
---|
66 |
|
---|
67 |
|
---|
68 | #ifdef USE_LONG_LONG
|
---|
69 | #define SEEK_X seek_LL
|
---|
70 | #define TELL_X tell_LL
|
---|
71 | #else
|
---|
72 | #define SEEK_X seek
|
---|
73 | #define TELL_X tell
|
---|
74 | #endif
|
---|
75 |
|
---|
76 | #ifndef RND_BUF_SIZE
|
---|
77 | #define RND_BUF_SIZE 8*1024
|
---|
78 | #endif
|
---|
79 |
|
---|
80 |
|
---|
81 | static mg_u_long numDocs = 0;
|
---|
82 | static mg_u_long numChunkDocs = 0;
|
---|
83 | static mg_u_long numDocsInChunk = 0;
|
---|
84 |
|
---|
85 | static mg_u_long numFrags = 0;
|
---|
86 | static mg_u_long numFragsInChunk = 0;
|
---|
87 | static mg_u_long chunkStartFragNum = 0;
|
---|
88 |
|
---|
89 |
|
---|
90 |
|
---|
91 |
|
---|
92 | struct BitPtr {
|
---|
93 | mg_u_long start;
|
---|
94 | mg_u_long here;
|
---|
95 | mg_u_long lastFragNum;
|
---|
96 | mg_u_long lgB;
|
---|
97 |
|
---|
98 | void Clear () { start = here = lastFragNum = lgB = 0; }
|
---|
99 | BitPtr () { Clear(); }
|
---|
100 | };
|
---|
101 |
|
---|
102 | class WordBitPtrs {
|
---|
103 | protected:
|
---|
104 | mg_u_long numWords;
|
---|
105 | mg_u_long numTags;
|
---|
106 | mg_u_long size;
|
---|
107 | BitPtr *wordBitPtrs;
|
---|
108 |
|
---|
109 | void CheckBufOverrun (mg_u_long num) {
|
---|
110 | if (wordBitPtrs[num].here > wordBitPtrs[num+1].start) {
|
---|
111 | cerr << "numDocs: " << numDocs << "\n";
|
---|
112 | cerr << "numChunkDocs: " << numChunkDocs << "\n";
|
---|
113 | cerr << "numDocsInChunk: " << numDocsInChunk << "\n";
|
---|
114 | cerr << "numFrags: " << numFrags << "\n";
|
---|
115 | cerr << "numFragsInChunk: " << numFragsInChunk << "\n";
|
---|
116 | cerr << "chunkStartFragNum: " << chunkStartFragNum << "\n";
|
---|
117 | cerr << "num: " << num << "\n";
|
---|
118 | cerr << "[num].start: " << wordBitPtrs[num].start << "\n";
|
---|
119 | cerr << "[num].here: " << wordBitPtrs[num].here << "\n";
|
---|
120 | cerr << "[num+1].start: " << wordBitPtrs[num+1].start << "\n";
|
---|
121 | FatalError (1, "Bit buffer overrun");
|
---|
122 | }
|
---|
123 | }
|
---|
124 |
|
---|
125 | public:
|
---|
126 | void Clear ();
|
---|
127 | WordBitPtrs () { wordBitPtrs = NULL; Clear(); }
|
---|
128 | ~WordBitPtrs ();
|
---|
129 | void SetSize (mg_u_long _numWords,
|
---|
130 | mg_u_long _numTags);
|
---|
131 |
|
---|
132 | void ResetPtrs () {
|
---|
133 | if (wordBitPtrs == NULL) return;
|
---|
134 | mg_u_long i;
|
---|
135 | for (i=0; i<size; ++i) wordBitPtrs[i].Clear();
|
---|
136 | }
|
---|
137 |
|
---|
138 | BitPtr &GetWordBitPtr (mg_u_long wordNum)
|
---|
139 | { return wordBitPtrs[wordNum]; }
|
---|
140 | mg_u_long &GetWordStart (mg_u_long wordNum)
|
---|
141 | { return wordBitPtrs[wordNum].start; }
|
---|
142 | mg_u_long &GetWordHere (mg_u_long wordNum)
|
---|
143 | { return wordBitPtrs[wordNum].here; }
|
---|
144 | void CheckWordBufOverrun (mg_u_long wordNum)
|
---|
145 | { CheckBufOverrun (wordNum); }
|
---|
146 |
|
---|
147 | BitPtr &GetTagBitPtr (mg_u_long tagNum)
|
---|
148 | { return wordBitPtrs[tagNum + numWords]; }
|
---|
149 | mg_u_long &GetTagStart (mg_u_long tagNum)
|
---|
150 | { return wordBitPtrs[tagNum + numWords].start; }
|
---|
151 | mg_u_long &GetTagHere (mg_u_long tagNum)
|
---|
152 | { return wordBitPtrs[tagNum + numWords].here; }
|
---|
153 | void CheckTagBufOverrun (mg_u_long tagNum)
|
---|
154 | { CheckBufOverrun (tagNum + numWords); }
|
---|
155 |
|
---|
156 | BitPtr &GetEndBitPtr ()
|
---|
157 | { return wordBitPtrs[size-1]; }
|
---|
158 | mg_u_long &GetEndStart ()
|
---|
159 | { return wordBitPtrs[size-1].start; }
|
---|
160 | mg_u_long &GetEndHere ()
|
---|
161 | { return wordBitPtrs[size-1].here; }
|
---|
162 | };
|
---|
163 |
|
---|
164 |
|
---|
165 | struct IP2TagInfo {
|
---|
166 | bool inTag;
|
---|
167 | mg_u_long startFrag;
|
---|
168 | mg_u_long tagNum;
|
---|
169 |
|
---|
170 | IP2TagInfo () {
|
---|
171 | inTag = false;
|
---|
172 | startFrag = 0;
|
---|
173 | tagNum = 0;
|
---|
174 | }
|
---|
175 | };
|
---|
176 |
|
---|
177 | // maps tags to tag information
|
---|
178 | typedef map<UCArray, IP2TagInfo, DictLTUCArray> TagMapDict;
|
---|
179 |
|
---|
180 |
|
---|
181 | // class to handle the translation of occurrence order
|
---|
182 | // to dictionary order for words and tags
|
---|
183 | class OccurToDictConverter {
|
---|
184 | protected:
|
---|
185 | mg_u_long pos;
|
---|
186 | mg_u_long val;
|
---|
187 | FILE *transFile;
|
---|
188 | random_bitio_buffer rbs;
|
---|
189 |
|
---|
190 | mg_u_long wordDictSize;
|
---|
191 | mg_u_long tagDictSize;
|
---|
192 |
|
---|
193 | void SeekStart ();
|
---|
194 | mg_u_long TranslateNum (mg_u_long num);
|
---|
195 |
|
---|
196 | public:
|
---|
197 | OccurToDictConverter ();
|
---|
198 | ~OccurToDictConverter ();
|
---|
199 |
|
---|
200 | void Open (char *filename, mg_u_long _wordDictSize,
|
---|
201 | mg_u_long _tagDictSize);
|
---|
202 |
|
---|
203 | // Close frees all allocated memory
|
---|
204 | void Close ();
|
---|
205 |
|
---|
206 | mg_u_long TranslateWord (mg_u_long occurNum)
|
---|
207 | { return TranslateNum (occurNum); }
|
---|
208 | mg_u_long TranslateTag (mg_u_long occurNum)
|
---|
209 | { return TranslateNum (occurNum+wordDictSize); }
|
---|
210 | };
|
---|
211 |
|
---|
212 |
|
---|
213 | struct InvfStateRec {
|
---|
214 | mg_ullong start;
|
---|
215 | mg_ullong here;
|
---|
216 | mg_u_long lastFragNum;
|
---|
217 | mg_u_long B;
|
---|
218 |
|
---|
219 | void Clear () {
|
---|
220 | start = here = 0;
|
---|
221 | lastFragNum = B = 0;
|
---|
222 | }
|
---|
223 | InvfStateRec () { Clear (); }
|
---|
224 | };
|
---|
225 |
|
---|
226 |
|
---|
227 | #define ISR_SIZE 1024
|
---|
228 |
|
---|
229 | class InvfStateCache {
|
---|
230 | protected:
|
---|
231 | InvfStateRec recCache [ISR_SIZE];
|
---|
232 | mg_u_long startNum;
|
---|
233 |
|
---|
234 | FILE *stateFile;
|
---|
235 |
|
---|
236 | void ClearCache () {
|
---|
237 | unsigned int i = 0;
|
---|
238 | for (i=0; i<ISR_SIZE; ++i) recCache[i].Clear();
|
---|
239 | }
|
---|
240 |
|
---|
241 | public:
|
---|
242 | InvfStateCache ();
|
---|
243 | ~InvfStateCache ();
|
---|
244 |
|
---|
245 | void Open (char *filename);
|
---|
246 | void Close ();
|
---|
247 |
|
---|
248 | // previous references to state records may be
|
---|
249 | // invalidated calling GetRec
|
---|
250 | InvfStateRec &GetRec (mg_u_long num);
|
---|
251 | };
|
---|
252 |
|
---|
253 |
|
---|
254 | static invf_dict_header idh;
|
---|
255 | static WordBitPtrs bitPtrs;
|
---|
256 |
|
---|
257 | static FILE *chunkFile = NULL;
|
---|
258 | static stdio_bitio_buffer chunkBuf;
|
---|
259 |
|
---|
260 | static mg_u_long ivfMemBufSize = 0;
|
---|
261 | static char *ivfMemBuf = NULL;
|
---|
262 |
|
---|
263 | // word and tag dictionaries. a map is used for the tag dictionary
|
---|
264 | // as it should never be very big (and the perfect hash function
|
---|
265 | // sometimes has trouble with small values).
|
---|
266 | static perf_hash_data *wordHashDict = NULL;
|
---|
267 | static TagMapDict tagMapDict;
|
---|
268 |
|
---|
269 | // information about all the different levels
|
---|
270 | static FIvfLevel ivfLevel;
|
---|
271 |
|
---|
272 | static OccurToDictConverter occurConvert;
|
---|
273 |
|
---|
274 | // information about the state of the inverted file
|
---|
275 | static InvfStateCache invfState;
|
---|
276 |
|
---|
277 | static char collectFilename[512];
|
---|
278 |
|
---|
279 |
|
---|
280 | void WordBitPtrs::Clear () {
|
---|
281 | numWords = 0;
|
---|
282 | numTags = 0;
|
---|
283 | size=0;
|
---|
284 | if (wordBitPtrs != NULL) delete [] wordBitPtrs;
|
---|
285 | wordBitPtrs = NULL;
|
---|
286 | }
|
---|
287 |
|
---|
288 | WordBitPtrs::~WordBitPtrs () {
|
---|
289 | if (wordBitPtrs != NULL) delete [] wordBitPtrs;
|
---|
290 | }
|
---|
291 |
|
---|
292 | void WordBitPtrs::SetSize (mg_u_long _numWords,
|
---|
293 | mg_u_long _numTags){
|
---|
294 | Clear();
|
---|
295 | numWords = _numWords;
|
---|
296 | numTags = _numTags;
|
---|
297 | size = numWords + numTags + 1;
|
---|
298 | wordBitPtrs = new BitPtr [size];
|
---|
299 | }
|
---|
300 |
|
---|
301 |
|
---|
302 | void OccurToDictConverter::SeekStart () {
|
---|
303 | if (transFile == NULL) return;
|
---|
304 | rbs.SEEK_X (sizeof (mg_u_long) * 8);
|
---|
305 | pos = 0;
|
---|
306 | }
|
---|
307 |
|
---|
308 | mg_u_long OccurToDictConverter::TranslateNum (mg_u_long num) {
|
---|
309 | if (num < pos) SeekStart ();
|
---|
310 | while (pos <= num) {
|
---|
311 | if (pos < wordDictSize)
|
---|
312 | val = rbs.binary_decode (wordDictSize + 1, NULL) - 1;
|
---|
313 | else
|
---|
314 | val = rbs.binary_decode (tagDictSize + 1, NULL) - 1;
|
---|
315 | ++pos;
|
---|
316 | }
|
---|
317 | return val;
|
---|
318 | }
|
---|
319 |
|
---|
320 | OccurToDictConverter::OccurToDictConverter () {
|
---|
321 | pos = 0;
|
---|
322 | val = 0;
|
---|
323 | transFile = NULL;
|
---|
324 | wordDictSize = 0;
|
---|
325 | tagDictSize = 0;
|
---|
326 | }
|
---|
327 |
|
---|
328 | OccurToDictConverter::~OccurToDictConverter () {
|
---|
329 | if (transFile != NULL) Close ();
|
---|
330 | }
|
---|
331 |
|
---|
332 | void OccurToDictConverter::Open (char *filename, mg_u_long _wordDictSize,
|
---|
333 | mg_u_long _tagDictSize) {
|
---|
334 | if (transFile != NULL) Close ();
|
---|
335 |
|
---|
336 | wordDictSize = _wordDictSize;
|
---|
337 | tagDictSize = _tagDictSize;
|
---|
338 |
|
---|
339 | transFile = open_file (filename, INVF_CHUNK_TRANS_SUFFIX, "rb",
|
---|
340 | MAGIC_CHUNK_TRANS, MG_ABORT);
|
---|
341 | rbs.attachFile (transFile, RND_BUF_SIZE);
|
---|
342 | SeekStart ();
|
---|
343 | val = 0;
|
---|
344 | }
|
---|
345 |
|
---|
346 | void OccurToDictConverter::Close () {
|
---|
347 | if (transFile == NULL) return;
|
---|
348 |
|
---|
349 | rbs.done ();
|
---|
350 | fclose (transFile);
|
---|
351 | transFile = NULL;
|
---|
352 | pos = 0;
|
---|
353 | val = 0;
|
---|
354 |
|
---|
355 | wordDictSize = 0;
|
---|
356 | tagDictSize = 0;
|
---|
357 | }
|
---|
358 |
|
---|
359 |
|
---|
360 |
|
---|
361 |
|
---|
362 | InvfStateCache::InvfStateCache () {
|
---|
363 | startNum = 0;
|
---|
364 | stateFile = NULL;
|
---|
365 | }
|
---|
366 |
|
---|
367 | InvfStateCache::~InvfStateCache () {
|
---|
368 | if (stateFile != NULL) Close ();
|
---|
369 | }
|
---|
370 |
|
---|
371 | void InvfStateCache::Open (char *filename) {
|
---|
372 | if (stateFile != NULL) Close();
|
---|
373 |
|
---|
374 | // open the state file
|
---|
375 | char path[512];
|
---|
376 | sprintf (path, FILE_NAME_FORMAT ".%d", get_basepath (), filename,
|
---|
377 | ".invf.state", (mg_s_long) getpid ());
|
---|
378 | if (!(stateFile = fopen (path, "wb+"))) {
|
---|
379 | Message ("Unable to create \"%s\"", path);
|
---|
380 | exit (1);
|
---|
381 | }
|
---|
382 | unlink (path); // file will be deleted after it is closed
|
---|
383 | // reset the buffer
|
---|
384 | startNum = 0;
|
---|
385 | ClearCache();
|
---|
386 | }
|
---|
387 |
|
---|
388 | void InvfStateCache::Close () {
|
---|
389 | if (stateFile == NULL) return;
|
---|
390 | fclose (stateFile);
|
---|
391 | stateFile = NULL;
|
---|
392 | startNum = 0;
|
---|
393 | }
|
---|
394 |
|
---|
395 | InvfStateRec &InvfStateCache::GetRec (mg_u_long num) {
|
---|
396 | // see if cached
|
---|
397 | if ((num >= startNum) && (num < startNum + ISR_SIZE))
|
---|
398 | return recCache[num-startNum];
|
---|
399 |
|
---|
400 | // not cached, write out this lot of records and read in
|
---|
401 | fseek (stateFile, startNum*sizeof (InvfStateRec), SEEK_SET);
|
---|
402 | fwrite ((char *) recCache, sizeof (InvfStateRec), ISR_SIZE, stateFile);
|
---|
403 |
|
---|
404 | // read in the new set of records
|
---|
405 | ClearCache ();
|
---|
406 | startNum = num - (num % ISR_SIZE);
|
---|
407 | fseek (stateFile, startNum*sizeof (InvfStateRec), SEEK_SET);
|
---|
408 | size_t numbytes = fread ((char *) recCache, sizeof (InvfStateRec), ISR_SIZE, stateFile);
|
---|
409 | if (numbytes != ISR_SIZE){
|
---|
410 | #ifdef DEBUG
|
---|
411 | fprintf(stderr, "[mgpp/text/ivf.pass2.cpp L410] number of bytes read by fread does not match the requested amount\n");
|
---|
412 | #endif
|
---|
413 | }
|
---|
414 |
|
---|
415 | return recCache[num-startNum];
|
---|
416 | }
|
---|
417 |
|
---|
418 |
|
---|
419 |
|
---|
420 | static void ClearCharBuf (char *buf, mg_u_long size) {
|
---|
421 | char *end = buf + size;
|
---|
422 | while (buf != end) *buf++ = 0;
|
---|
423 | }
|
---|
424 |
|
---|
425 | static void ReadWordDict (char *filename) {
|
---|
426 | // read in the perfect hash function for the word dictionary
|
---|
427 | FILE *wordHashFile = open_file (filename, INVF_DICT_HASH_SUFFIX, "rb",
|
---|
428 | MAGIC_HASH, MG_ABORT);
|
---|
429 | if (!(wordHashDict = read_perf_hash_data (wordHashFile))) {
|
---|
430 | FatalError (1, "Unable to read in hash data for word dictionary");
|
---|
431 | }
|
---|
432 | fclose (wordHashFile);
|
---|
433 | }
|
---|
434 |
|
---|
435 | static void ReadTagDict (char *filename, invf_dict_header &_idh) {
|
---|
436 | // open the file
|
---|
437 | FILE *dictFile = open_file (filename, INVF_DICT_SUFFIX, "rb",
|
---|
438 | MAGIC_STEM_BUILD, MG_ABORT);
|
---|
439 |
|
---|
440 | // seek to the start of the tag dictionary
|
---|
441 | fseek (dictFile, _idh.tag_dict_start, SEEK_SET);
|
---|
442 |
|
---|
443 | mg_u_long tagNum;
|
---|
444 | dict_el thisEl;
|
---|
445 | for (tagNum = 0; tagNum < _idh.tag_dict_size; ++tagNum) {
|
---|
446 | thisEl.Read (dictFile);
|
---|
447 | tagMapDict[thisEl.el].tagNum = tagNum;
|
---|
448 | }
|
---|
449 |
|
---|
450 | fclose (dictFile);
|
---|
451 | }
|
---|
452 |
|
---|
453 | static void ReadLevelFile (char *filename) {
|
---|
454 | FILE *f;
|
---|
455 | f = open_file (filename, INVF_LEVEL_SUFFIX, "rb",
|
---|
456 | MAGIC_INVF_LEVELS, MG_ABORT);
|
---|
457 | ivfLevel.Read (f);
|
---|
458 | fclose (f);
|
---|
459 | }
|
---|
460 |
|
---|
461 | void CheckIntOverflow (mg_ullong totalIBits, mg_ullong lastTotalIBits) {
|
---|
462 | if (totalIBits < lastTotalIBits) {
|
---|
463 | fprintf(stderr, "ERROR: The totalIBits counter (%lu byte unsigned integer) has overflowed.\n", sizeof (mg_ullong));
|
---|
464 | if (sizeof (mg_ullong) < 8) {
|
---|
465 | fprintf(stderr, " Try compiling with GCC to enable use of 8 bytes for this counter.\n");
|
---|
466 | }
|
---|
467 | fprintf(stderr, " Build aborted.\n");
|
---|
468 | exit(1);
|
---|
469 | }
|
---|
470 | }
|
---|
471 |
|
---|
472 | // assumes the inverted file state file has been opened
|
---|
473 | static void InitInvfState (char *filename,
|
---|
474 | invf_dict_header &_idh,
|
---|
475 | InvfStateCache &_invfState,
|
---|
476 | mg_ullong &totalIBits,
|
---|
477 | bool wordLevelIndex) {
|
---|
478 | // read in the dictionary, setting inverted state information
|
---|
479 | FILE *dictFile = open_file (filename, INVF_DICT_SUFFIX, "rb",
|
---|
480 | MAGIC_STEM_BUILD, MG_ABORT);
|
---|
481 |
|
---|
482 | // seek to the start of the word dictionary
|
---|
483 | fseek (dictFile, _idh.word_dict_start, SEEK_SET);
|
---|
484 |
|
---|
485 | // add the word entries
|
---|
486 | word_dict_el wordEl;
|
---|
487 | wordEl.SetNumLevels (_idh.num_levels);
|
---|
488 | mg_u_long dictWordNum, p;
|
---|
489 | mg_ullong lastTotalIBits;
|
---|
490 | mg_u_long N = _idh.num_frags;
|
---|
491 | for (dictWordNum=0; dictWordNum<_idh.word_dict_size; ++dictWordNum) {
|
---|
492 | // lastTotalIBits is used to detect integer overflow
|
---|
493 | lastTotalIBits = totalIBits;
|
---|
494 |
|
---|
495 | // read the next word and associated information
|
---|
496 | wordEl.Read (dictFile, _idh.num_levels);
|
---|
497 |
|
---|
498 | // update the state record
|
---|
499 | p = wordEl.frag_occur;
|
---|
500 | InvfStateRec &wisr = _invfState.GetRec (dictWordNum);
|
---|
501 | wisr.start = totalIBits;
|
---|
502 | wisr.here = totalIBits;
|
---|
503 | wisr.B = BIO_Bblock_Init (N, p);
|
---|
504 |
|
---|
505 | // add the length of the fragment numbers
|
---|
506 | totalIBits += BIO_Bblock_Bound_b (N, p, wisr.B);
|
---|
507 |
|
---|
508 | // if needed, add the length of the fragment frequency information
|
---|
509 | if (!wordLevelIndex)
|
---|
510 | totalIBits += BIO_Gamma_Bound (wordEl.freq, wordEl.frag_occur);
|
---|
511 |
|
---|
512 | // align next byte
|
---|
513 | #ifdef USE_LONG_LONG
|
---|
514 | totalIBits = (totalIBits + 7ull) & 0xfffffffffffffff8ull;
|
---|
515 | #else
|
---|
516 | totalIBits = (totalIBits + 7ul) & 0xfffffff8ul;
|
---|
517 | #endif
|
---|
518 |
|
---|
519 | CheckIntOverflow (totalIBits, lastTotalIBits);
|
---|
520 | }
|
---|
521 |
|
---|
522 |
|
---|
523 | // seek to the start of the tag dictionary
|
---|
524 | fseek (dictFile, _idh.tag_dict_start, SEEK_SET);
|
---|
525 |
|
---|
526 | // add the tag entries
|
---|
527 | dict_el tagEl;
|
---|
528 | mg_u_long dictTagNum;
|
---|
529 | N = _idh.num_frags;
|
---|
530 | for (dictTagNum=0; dictTagNum<_idh.tag_dict_size; ++dictTagNum) {
|
---|
531 | // lastTotalIBits is used to detect integer overflow
|
---|
532 | lastTotalIBits = totalIBits;
|
---|
533 |
|
---|
534 | // read the next tag and associated information
|
---|
535 | tagEl.Read (dictFile);
|
---|
536 |
|
---|
537 | // update the state record
|
---|
538 | p = tagEl.frag_occur*2;
|
---|
539 | InvfStateRec &tisr = _invfState.GetRec (dictTagNum + _idh.word_dict_size);
|
---|
540 | tisr.start = totalIBits;
|
---|
541 | tisr.here = totalIBits;
|
---|
542 | tisr.B = BIO_Bblock_Init (N+p, p);
|
---|
543 |
|
---|
544 | // add the length of the fragment numbers (two numbers for each
|
---|
545 | // tag, one for start and one for end)
|
---|
546 | totalIBits += BIO_Bblock_Bound_b (N+p, p, tisr.B);
|
---|
547 |
|
---|
548 | // align next byte
|
---|
549 | #ifdef USE_LONG_LONG
|
---|
550 | totalIBits = (totalIBits + 7ull) & 0xfffffffffffffff8ull;
|
---|
551 | #else
|
---|
552 | totalIBits = (totalIBits + 7ul) & 0xfffffff8ul;
|
---|
553 | #endif
|
---|
554 |
|
---|
555 | CheckIntOverflow (totalIBits, lastTotalIBits);
|
---|
556 | }
|
---|
557 |
|
---|
558 | fclose (dictFile);
|
---|
559 | }
|
---|
560 |
|
---|
561 | /*
|
---|
562 | // assumes the chunk tag information has been placed in .first
|
---|
563 | static void PrintChunkInfo (mg_u_long chunkMem,
|
---|
564 | mg_u_long numChunkWords,
|
---|
565 | mg_u_long numChunkTags) {
|
---|
566 | static mg_u_long chunksRead = 0;
|
---|
567 | ++chunksRead;
|
---|
568 | cout << "Chunk Number: " << chunksRead << "\n";
|
---|
569 | cout << "numChunkDocs " << numDocsInChunk << "\n";
|
---|
570 | cout << "numChunkFrags " << numFragsInChunk << "\n";
|
---|
571 | cout << "mem " << chunkMem << "\n";
|
---|
572 | cout << "numWords " << numChunkWords << "\n";
|
---|
573 | cout << "numTags " << numChunkTags << "\n\n";
|
---|
574 |
|
---|
575 | TagMapDict::iterator tagMapHere = tagMapDict.begin();
|
---|
576 | TagMapDict::iterator tagMapEnd = tagMapDict.end();
|
---|
577 | while (tagMapHere != tagMapEnd) {
|
---|
578 | mg_u_long tagMapNum = (*tagMapHere).second.tagNum;
|
---|
579 | cout << (*tagMapHere).first << " " << tagMapNum << " "
|
---|
580 | << bitPtrs.GetTagBitPtr(tagMapNum).here << "\n";
|
---|
581 | ++tagMapHere;
|
---|
582 | }
|
---|
583 | }
|
---|
584 | */
|
---|
585 |
|
---|
586 | void ReadChunk (invf_dict_header &_idh, bool wordLevelIndex) {
|
---|
587 | // reset globals
|
---|
588 | numChunkDocs = 0;
|
---|
589 | chunkStartFragNum = numFrags;
|
---|
590 |
|
---|
591 | // read in information about this chunk
|
---|
592 | numDocsInChunk = chunkBuf.gamma_decode (NULL) - 1;
|
---|
593 | if (numDocsInChunk == 0)
|
---|
594 | FatalError (1, "The number of docs in the current chunk is 0");
|
---|
595 |
|
---|
596 | numFragsInChunk = chunkBuf.gamma_decode (NULL) - 1;
|
---|
597 | mg_u_long chunkMem = chunkBuf.gamma_decode (NULL) - 1;
|
---|
598 |
|
---|
599 | if (chunkMem > ivfMemBufSize)
|
---|
600 | FatalError (1, "Chunk memory size is greater than maximum");
|
---|
601 |
|
---|
602 | mg_u_long numChunkWords = chunkBuf.gamma_decode (NULL) - 1;
|
---|
603 | mg_u_long numChunkTags = chunkBuf.gamma_decode (NULL) - 1;
|
---|
604 |
|
---|
605 |
|
---|
606 | // reset stuff
|
---|
607 | ClearCharBuf (ivfMemBuf, ivfMemBufSize);
|
---|
608 | bitPtrs.ResetPtrs();
|
---|
609 |
|
---|
610 | // read in the entries in occurrence order storing the
|
---|
611 | // "chunkWordCount" in "start" and the "chunkFragCount"
|
---|
612 | // in "here"
|
---|
613 | mg_u_long numOccur;
|
---|
614 | mg_u_long wordNum;
|
---|
615 | for (numOccur=0; numOccur<numChunkWords; ++numOccur) {
|
---|
616 | wordNum = occurConvert.TranslateWord (numOccur);
|
---|
617 | BitPtr &wordPtr = bitPtrs.GetWordBitPtr (wordNum);
|
---|
618 | wordPtr.start = chunkBuf.gamma_decode (NULL) - 1;
|
---|
619 | if (wordPtr.start >= 2)
|
---|
620 | wordPtr.here = chunkBuf.gamma_decode (NULL);
|
---|
621 | else wordPtr.here = wordPtr.start;
|
---|
622 | }
|
---|
623 | mg_u_long tagNum;
|
---|
624 | for (numOccur=0; numOccur<numChunkTags; ++numOccur) {
|
---|
625 | tagNum = occurConvert.TranslateTag (numOccur);
|
---|
626 | BitPtr &tagPtr = bitPtrs.GetTagBitPtr (tagNum);
|
---|
627 | // only chunkFragCount is encoded for tags
|
---|
628 | tagPtr.start = chunkBuf.gamma_decode (NULL) - 1;
|
---|
629 | tagPtr.here = tagPtr.start;
|
---|
630 | }
|
---|
631 |
|
---|
632 | /* PrintChunkInfo (chunkMem, numChunkWords, numChunkTags);*/
|
---|
633 |
|
---|
634 | // create the bit ptrs in dictionary order
|
---|
635 | mg_u_long totalIBits = 0; // only dealing with memory
|
---|
636 | mg_u_long chunkWordCount, chunkFragCount;
|
---|
637 | for (wordNum=0; wordNum<_idh.word_dict_size; ++wordNum) {
|
---|
638 | BitPtr &wordPtr = bitPtrs.GetWordBitPtr (wordNum);
|
---|
639 | chunkWordCount = wordPtr.start;
|
---|
640 | chunkFragCount = wordPtr.here;
|
---|
641 | wordPtr.start = totalIBits;
|
---|
642 | wordPtr.here = totalIBits;
|
---|
643 | wordPtr.lastFragNum = chunkStartFragNum;
|
---|
644 | wordPtr.lgB = 0;
|
---|
645 | if (chunkWordCount > 0) {
|
---|
646 | wordPtr.lgB = floorlog_2 (BIO_Bblock_Init_W (numFragsInChunk,
|
---|
647 | chunkFragCount));
|
---|
648 | totalIBits += BIO_Bblock_Bound (numFragsInChunk, chunkFragCount);
|
---|
649 | // use unary encoding for memory buffer encoding of fragment freq
|
---|
650 | if (!wordLevelIndex) {
|
---|
651 | totalIBits += chunkWordCount;
|
---|
652 | }
|
---|
653 | }
|
---|
654 | }
|
---|
655 | for (tagNum=0; tagNum<_idh.tag_dict_size; ++tagNum) {
|
---|
656 | BitPtr &tagPtr = bitPtrs.GetTagBitPtr (tagNum);
|
---|
657 | chunkFragCount = tagPtr.here;
|
---|
658 | tagPtr.start = totalIBits;
|
---|
659 | tagPtr.here = totalIBits;
|
---|
660 | tagPtr.lastFragNum = chunkStartFragNum;
|
---|
661 | tagPtr.lgB = 0;
|
---|
662 | if (chunkFragCount > 0) {
|
---|
663 | mg_u_long pTag = chunkFragCount*2;
|
---|
664 | tagPtr.lgB = floorlog_2 (BIO_Bblock_Init_W (numFragsInChunk+pTag,
|
---|
665 | pTag));
|
---|
666 | mg_u_long bLen = BIO_Bblock_Bound (numFragsInChunk+pTag,
|
---|
667 | pTag);
|
---|
668 | // cout << tagNum + _idh.word_dict_size << " ";
|
---|
669 | // cout << "numFrags: " << numFragsInChunk
|
---|
670 | // << " chunkFragCount: " << chunkFragCount
|
---|
671 | // << " B: " << 1 << tagPtr.lgB
|
---|
672 | // << " blen: " << blen << "\n";
|
---|
673 | totalIBits += bLen;
|
---|
674 | }
|
---|
675 | }
|
---|
676 | bitPtrs.GetEndStart() = totalIBits;
|
---|
677 | bitPtrs.GetEndHere() = totalIBits;
|
---|
678 |
|
---|
679 | if ((totalIBits + 7ul) >> 3ul > chunkMem) {
|
---|
680 | cerr << "totalIBits: " << totalIBits << "\n";
|
---|
681 | cerr << "bytes: " << ((totalIBits + 7ul) >> 3ul) << "\n";
|
---|
682 | cerr << "chunkMem: " << chunkMem << "\n";
|
---|
683 | FatalError (1, "Pointers exceed buffer size");
|
---|
684 | }
|
---|
685 | }
|
---|
686 |
|
---|
687 |
|
---|
688 |
|
---|
689 |
|
---|
690 | int init_ivf_2 (const TagInfo &/*tagInfo*/, char *filename) {
|
---|
691 | // read in compressed dictionary header
|
---|
692 | FILE *dictFile = open_file (filename, INVF_DICT_SUFFIX, "rb",
|
---|
693 | MAGIC_STEM_BUILD, MG_ABORT);
|
---|
694 | idh.Read (dictFile);
|
---|
695 | fclose (dictFile);
|
---|
696 |
|
---|
697 | // set the size of the bit ptrs
|
---|
698 | bitPtrs.SetSize (idh.word_dict_size, idh.tag_dict_size);
|
---|
699 |
|
---|
700 | // open the chunk file and read in the maximum memory needed
|
---|
701 | // for the inverted memory buffer
|
---|
702 | chunkFile = open_file (filename, INVF_CHUNK_SUFFIX, "rb",
|
---|
703 | MAGIC_CHUNK, MG_ABORT);
|
---|
704 | ReadUL (chunkFile, ivfMemBufSize);
|
---|
705 | chunkBuf.attachFile (chunkFile);
|
---|
706 |
|
---|
707 | // allocate memory for the inverted buffer
|
---|
708 | ivfMemBuf = new char [ivfMemBufSize];
|
---|
709 | ClearCharBuf (ivfMemBuf, ivfMemBufSize);
|
---|
710 |
|
---|
711 | // read in the word dictionary
|
---|
712 | ReadWordDict (filename);
|
---|
713 |
|
---|
714 | // read in the tag dictionary
|
---|
715 | ReadTagDict (filename, idh);
|
---|
716 |
|
---|
717 | // read in the level information
|
---|
718 | ReadLevelFile (filename);
|
---|
719 | bool wordLevelIndex = ivfLevel.indexLevel.empty();
|
---|
720 |
|
---|
721 | // set up the translation table file
|
---|
722 | occurConvert.Open (filename, idh.word_dict_size, idh.tag_dict_size);
|
---|
723 |
|
---|
724 | // reset some globals
|
---|
725 | numDocs = 0;
|
---|
726 | numChunkDocs = 0;
|
---|
727 | numDocsInChunk = 0;
|
---|
728 | numFrags = 0;
|
---|
729 | numFragsInChunk = 0;
|
---|
730 | chunkStartFragNum = 0;
|
---|
731 |
|
---|
732 | strcpy (collectFilename, filename);
|
---|
733 |
|
---|
734 |
|
---|
735 | // create the inverted file
|
---|
736 | mg_ullong totalIBits = 0;
|
---|
737 | FILE *invfFile = create_file (filename, INVF_SUFFIX, "wb",
|
---|
738 | MAGIC_INVF, MG_ABORT);
|
---|
739 | totalIBits += sizeof (mg_u_long) * 8; // magic number
|
---|
740 | totalIBits += 8 * 200; // 200 byte gap -- why??????
|
---|
741 | fclose (invfFile);
|
---|
742 |
|
---|
743 | // init the inverted file state cache
|
---|
744 | invfState.Open (filename);
|
---|
745 | InitInvfState (filename, idh, invfState, totalIBits, wordLevelIndex);
|
---|
746 |
|
---|
747 | return COMPALLOK;
|
---|
748 | }
|
---|
749 |
|
---|
750 | static void CloseTextTag (IP2TagInfo &tInfo, const UCArray &/*tagName*/) {
|
---|
751 | if (!tInfo.inTag) return;
|
---|
752 |
|
---|
753 | // add this tag to the inverted list
|
---|
754 | BitPtr &tagBitPtr = bitPtrs.GetTagBitPtr (tInfo.tagNum);
|
---|
755 | mg_u_long endFrag = numFrags;
|
---|
756 | int b = 1 << tagBitPtr.lgB;
|
---|
757 |
|
---|
758 | /*
|
---|
759 | cout << (tInfo.tagNum+idh.word_dict_size) << " \"<" << tagName << ">\" "
|
---|
760 | << tInfo.startFrag << " " << endFrag << "\n";
|
---|
761 | */
|
---|
762 |
|
---|
763 | mems_bitio_buffer buffer ((u_char *) ivfMemBuf, tagBitPtr.here);
|
---|
764 | buffer.bblock_encode (tInfo.startFrag - tagBitPtr.lastFragNum + 1,
|
---|
765 | b, NULL);
|
---|
766 | buffer.bblock_encode (endFrag - tInfo.startFrag + 1, b, NULL);
|
---|
767 | tagBitPtr.lastFragNum = endFrag;
|
---|
768 | tagBitPtr.here = buffer.position();
|
---|
769 | buffer.encodeDone();
|
---|
770 |
|
---|
771 | // check for buffer overrun
|
---|
772 | bitPtrs.CheckTagBufOverrun (tInfo.tagNum);
|
---|
773 |
|
---|
774 | // reset information about this tag
|
---|
775 | tInfo.inTag = false;
|
---|
776 | tInfo.startFrag = 0;
|
---|
777 | }
|
---|
778 |
|
---|
779 | static void ProcessOpenTag (const TextEl &el, bool &inFrag) {
|
---|
780 | // close tag if open
|
---|
781 | IP2TagInfo &tInfo = tagMapDict[el.tagName];
|
---|
782 | if (tInfo.inTag) CloseTextTag (tInfo, el.tagName);
|
---|
783 |
|
---|
784 | // open this tag
|
---|
785 | tInfo.inTag = true;
|
---|
786 | tInfo.startFrag = numFrags;
|
---|
787 |
|
---|
788 | // check for start of next fragment
|
---|
789 | bool wordLevelIndex = ivfLevel.indexLevel.empty();
|
---|
790 | if (!wordLevelIndex && el.tagName == ivfLevel.indexLevel) {
|
---|
791 | ++numFrags;
|
---|
792 | inFrag = true;
|
---|
793 | }
|
---|
794 | }
|
---|
795 |
|
---|
796 | static void ProcessCloseTag (const TextEl &el, bool &inFrag) {
|
---|
797 | // check for end of fragment
|
---|
798 | bool wordLevelIndex = ivfLevel.indexLevel.empty();
|
---|
799 | if (!wordLevelIndex && el.tagName == ivfLevel.indexLevel) {
|
---|
800 | inFrag = false;
|
---|
801 | }
|
---|
802 |
|
---|
803 | IP2TagInfo &tInfo = tagMapDict[el.tagName];
|
---|
804 | CloseTextTag (tInfo, el.tagName);
|
---|
805 | }
|
---|
806 |
|
---|
807 | static void ProcessText (const TextEl &el, bool &inFrag) {
|
---|
808 | // make sure this text is to be indexed
|
---|
809 | bool wordLevelIndex = ivfLevel.indexLevel.empty();
|
---|
810 | if (!wordLevelIndex && !inFrag) return;
|
---|
811 |
|
---|
812 | const unsigned char *textHere = &(el.text[0]);
|
---|
813 | const unsigned char *textEnd = &(el.text[el.text.size() - 1]);
|
---|
814 | unsigned char mgWord[MAXSTEMLEN + 1];
|
---|
815 |
|
---|
816 | if (!inaword_mgpp (textHere, textEnd))
|
---|
817 | ParseNonindexWord (textHere, textEnd);
|
---|
818 |
|
---|
819 |
|
---|
820 | // Alternately parse off words and non-words from the input
|
---|
821 |
|
---|
822 | while (textHere <= textEnd) {
|
---|
823 | textHere = ParseIndexMGWord (textHere, textEnd, mgWord);
|
---|
824 | textHere = ParseNonindexWord (textHere, textEnd);
|
---|
825 |
|
---|
826 | if (mgWord[0] > 0) {
|
---|
827 | if (wordLevelIndex) ++numFrags;
|
---|
828 |
|
---|
829 | mg_u_long wordNum = perf_hash (wordHashDict, mgWord);
|
---|
830 |
|
---|
831 | /*
|
---|
832 | cout << wordNum << " \"";
|
---|
833 | cout.write (mgWord+1, *mgWord);
|
---|
834 | cout << "\" " << numFrags << "\n";
|
---|
835 | */
|
---|
836 |
|
---|
837 | // add this word to the inverted list
|
---|
838 | BitPtr &wordBitPtr = bitPtrs.GetWordBitPtr (wordNum);
|
---|
839 | mg_u_long fragNum = numFrags;
|
---|
840 | int b = 1 << wordBitPtr.lgB;
|
---|
841 |
|
---|
842 | mems_bitio_buffer buffer ((u_char *) ivfMemBuf, wordBitPtr.here);
|
---|
843 |
|
---|
844 | // note: this assumes that fragments don't carry over between
|
---|
845 | // chunks (which they don't because all tags are closed at the
|
---|
846 | // end of each document and chunks are based on document
|
---|
847 | // boundaries), i.e. the first fragment number must be greater
|
---|
848 | // than the starting fragment number of the chunk.
|
---|
849 | if (fragNum > wordBitPtr.lastFragNum) {
|
---|
850 | buffer.bblock_encode ((fragNum - wordBitPtr.lastFragNum - 1) + 1,
|
---|
851 | b, NULL);
|
---|
852 | if (!wordLevelIndex) buffer.encodeBit (1); // freq = 1
|
---|
853 |
|
---|
854 | } else if (!wordLevelIndex) {
|
---|
855 | // add one to the frequency count for this word
|
---|
856 | buffer.seek (buffer.position()-1);
|
---|
857 | buffer.encodeBit (0); // unary encoding -- last = 1
|
---|
858 | buffer.encodeBit (1);
|
---|
859 | }
|
---|
860 |
|
---|
861 | wordBitPtr.lastFragNum = fragNum;
|
---|
862 | wordBitPtr.here = buffer.position();
|
---|
863 | buffer.encodeDone();
|
---|
864 |
|
---|
865 | // check for buffer overrun
|
---|
866 | bitPtrs.CheckWordBufOverrun (wordNum);
|
---|
867 | }
|
---|
868 | }
|
---|
869 | }
|
---|
870 |
|
---|
871 | // combine the in memory inverted buffer with the disk
|
---|
872 | // based inverted file
|
---|
873 | static void DiskMerge (char *filename) {
|
---|
874 | bool wordLevelIndex = ivfLevel.indexLevel.empty();
|
---|
875 |
|
---|
876 | // make sure we have something to process
|
---|
877 | if (numChunkDocs <= 0) return;
|
---|
878 |
|
---|
879 | // open the inverted file
|
---|
880 | FILE *invfFile = open_file (filename, INVF_SUFFIX, "rb+",
|
---|
881 | MAGIC_INVF, MG_ABORT);
|
---|
882 | random_bitio_buffer invfOutBuf (invfFile);
|
---|
883 |
|
---|
884 | // set up to decode the entries in memory
|
---|
885 | mems_bitio_buffer memInBuf ((u_char *) ivfMemBuf, 0);
|
---|
886 |
|
---|
887 | // write out the word information
|
---|
888 | mg_u_long wordNum;
|
---|
889 | int b;
|
---|
890 | mg_u_long currFragNum;
|
---|
891 | mg_u_long delta;
|
---|
892 | mg_u_long currFreq;
|
---|
893 | for (wordNum=0; wordNum<idh.word_dict_size; ++wordNum) {
|
---|
894 | // go to the end of the last inverted file entry
|
---|
895 | InvfStateRec &wordDiskState = invfState.GetRec (wordNum);
|
---|
896 | invfOutBuf.SEEK_X (wordDiskState.here);
|
---|
897 |
|
---|
898 | // go to the start of the inverted chunk info in memory
|
---|
899 | BitPtr &wordBitPtr = bitPtrs.GetWordBitPtr (wordNum);
|
---|
900 | memInBuf.seek(wordBitPtr.start);
|
---|
901 |
|
---|
902 | // decode each entry and re-write to disk
|
---|
903 | currFragNum = chunkStartFragNum;
|
---|
904 | while (memInBuf.position() < wordBitPtr.here) {
|
---|
905 | // decode word entry
|
---|
906 | b = 1 << wordBitPtr.lgB;
|
---|
907 | delta = memInBuf.bblock_decode (b, NULL);
|
---|
908 | currFragNum += delta;
|
---|
909 | if (!wordLevelIndex) currFreq = memInBuf.unary_decode (NULL);
|
---|
910 | else currFreq = 1;
|
---|
911 |
|
---|
912 | // recode on disk
|
---|
913 | invfOutBuf.bblock_encode (currFragNum-wordDiskState.lastFragNum,
|
---|
914 | wordDiskState.B, NULL);
|
---|
915 | if (!wordLevelIndex) invfOutBuf.gamma_encode (currFreq, NULL);
|
---|
916 | wordDiskState.lastFragNum = currFragNum;
|
---|
917 | }
|
---|
918 |
|
---|
919 | wordDiskState.here = invfOutBuf.TELL_X();
|
---|
920 | }
|
---|
921 |
|
---|
922 | // write out the tag information
|
---|
923 | mg_u_long tagNum;
|
---|
924 | mg_u_long currTagStart;
|
---|
925 | mg_u_long currTagEnd;
|
---|
926 | for (tagNum=0; tagNum<idh.tag_dict_size; ++tagNum) {
|
---|
927 | // go to the end of the last inverted file entry
|
---|
928 | InvfStateRec &tagDiskState = invfState.GetRec (tagNum+idh.word_dict_size);
|
---|
929 | invfOutBuf.SEEK_X (tagDiskState.here);
|
---|
930 |
|
---|
931 | // go to the start of the inverted chunk info in memory
|
---|
932 | BitPtr &tagBitPtr = bitPtrs.GetTagBitPtr (tagNum);
|
---|
933 | memInBuf.seek(tagBitPtr.start);
|
---|
934 |
|
---|
935 | // decode each entry and re-write to disk
|
---|
936 | currTagEnd = chunkStartFragNum;
|
---|
937 | while (memInBuf.position() < tagBitPtr.here) {
|
---|
938 | // decode tag entry
|
---|
939 | b = 1 << tagBitPtr.lgB;
|
---|
940 | delta = memInBuf.bblock_decode (b, NULL) - 1;
|
---|
941 | currTagStart = currTagEnd + delta;
|
---|
942 | delta = memInBuf.bblock_decode (b, NULL) - 1;
|
---|
943 | currTagEnd = currTagStart + delta;
|
---|
944 |
|
---|
945 | // recode on disk
|
---|
946 | invfOutBuf.bblock_encode (currTagStart-tagDiskState.lastFragNum+1,
|
---|
947 | tagDiskState.B, NULL);
|
---|
948 | invfOutBuf.bblock_encode (currTagEnd-currTagStart+1,
|
---|
949 | tagDiskState.B, NULL);
|
---|
950 |
|
---|
951 | tagDiskState.lastFragNum = currTagEnd;
|
---|
952 | }
|
---|
953 |
|
---|
954 | tagDiskState.here = invfOutBuf.TELL_X();
|
---|
955 | }
|
---|
956 |
|
---|
957 | memInBuf.done();
|
---|
958 |
|
---|
959 | invfOutBuf.encodeDone();
|
---|
960 | fclose (invfFile);
|
---|
961 | }
|
---|
962 |
|
---|
963 |
|
---|
964 | int process_ivf_2 (const TagInfo &/*tagInfo*/, const TextElArray &doc) {
|
---|
965 | bool wordLevelIndex = ivfLevel.indexLevel.empty();
|
---|
966 | bool inFrag = false;
|
---|
967 | if (wordLevelIndex) inFrag = true; // unconditional
|
---|
968 |
|
---|
969 | // get next chunk information if need to. the chunk information
|
---|
970 | // is needed before the first document is processed
|
---|
971 | if (numChunkDocs >= numDocsInChunk) ReadChunk (idh, wordLevelIndex);
|
---|
972 |
|
---|
973 | // process each text element
|
---|
974 | TextElArray::const_iterator here = doc.begin();
|
---|
975 | TextElArray::const_iterator end = doc.end();
|
---|
976 | while (here != end) {
|
---|
977 | // process this element
|
---|
978 | if ((*here).elType == OpenTagE) ProcessOpenTag (*here, inFrag);
|
---|
979 | else if ((*here).elType == CloseTagE) ProcessCloseTag (*here, inFrag);
|
---|
980 | else ProcessText (*here, inFrag);
|
---|
981 |
|
---|
982 | ++here;
|
---|
983 | }
|
---|
984 |
|
---|
985 | // close off any unclosed tags
|
---|
986 | TagMapDict::iterator tdHere = tagMapDict.begin();
|
---|
987 | TagMapDict::iterator tdEnd = tagMapDict.end();
|
---|
988 | while (tdHere != tdEnd) {
|
---|
989 | CloseTextTag ((*tdHere).second, (*tdHere).first);
|
---|
990 | ++tdHere;
|
---|
991 | }
|
---|
992 |
|
---|
993 | // we've processed one more document
|
---|
994 | ++numDocs;
|
---|
995 | ++numChunkDocs;
|
---|
996 |
|
---|
997 | // merge the memory based inverted file with the one on
|
---|
998 | // disk if this is the end of this chunk
|
---|
999 | if (numChunkDocs >= numDocsInChunk) DiskMerge (collectFilename);
|
---|
1000 |
|
---|
1001 | return COMPALLOK;
|
---|
1002 | }
|
---|
1003 |
|
---|
1004 |
|
---|
1005 | static void CondenseInvfFile (char *filename, mg_u_long &bytesOutput) {
|
---|
1006 | FILE *inInvfFile = open_file (filename, INVF_SUFFIX, "rb",
|
---|
1007 | MAGIC_INVF, MG_ABORT);
|
---|
1008 | FILE *outInvfFile = open_file (filename, INVF_SUFFIX, "rb+",
|
---|
1009 | MAGIC_INVF, MG_ABORT);
|
---|
1010 |
|
---|
1011 | // skip the magic number
|
---|
1012 | fseek (outInvfFile, sizeof (mg_u_long), SEEK_SET);
|
---|
1013 |
|
---|
1014 | // write the inverted file header -- use defaults for most things
|
---|
1015 | invf_file_header ifh;
|
---|
1016 | ifh.no_of_words = idh.word_dict_size;
|
---|
1017 | ifh.no_of_tags = idh.tag_dict_size;
|
---|
1018 | ifh.word_level_index = (ivfLevel.indexLevel.empty()) ? 1 : 0;
|
---|
1019 | ifh.Write (outInvfFile);
|
---|
1020 |
|
---|
1021 | bytesOutput = ftell (outInvfFile);
|
---|
1022 |
|
---|
1023 | // process each meaningful byte in the file
|
---|
1024 | mg_u_long numEntries = ifh.no_of_words + ifh.no_of_tags;
|
---|
1025 | mg_u_long entryNum;
|
---|
1026 | mg_ullong lastStart = 0;
|
---|
1027 | for (entryNum = 0; entryNum < numEntries; ++entryNum) {
|
---|
1028 | InvfStateRec &stateRec = invfState.GetRec (entryNum);
|
---|
1029 |
|
---|
1030 | // overrun check
|
---|
1031 | if (stateRec.start < lastStart)
|
---|
1032 | FatalError (1, "Inverted file Buffer overrun");
|
---|
1033 | lastStart = stateRec.start;
|
---|
1034 |
|
---|
1035 | mg_u_long oldEntryStart = stateRec.start >> 3;
|
---|
1036 | mg_u_long oldEntryStartOver = stateRec.start & 7; // should be 0
|
---|
1037 | mg_u_long oldEntryEnd = (stateRec.here + 7) >> 3; // byte after end
|
---|
1038 | mg_u_long oldEntryEndOver = stateRec.here & 7;
|
---|
1039 |
|
---|
1040 | fseek (inInvfFile, oldEntryStart, SEEK_SET);
|
---|
1041 |
|
---|
1042 | stateRec.here -= stateRec.start;
|
---|
1043 | stateRec.start = (mg_ullong)bytesOutput * 8 + oldEntryStartOver;
|
---|
1044 | stateRec.here += stateRec.start;
|
---|
1045 | while (oldEntryStart < oldEntryEnd) {
|
---|
1046 | unsigned char c = getc (inInvfFile);
|
---|
1047 | if (oldEntryStart == oldEntryEnd - 1) {
|
---|
1048 | u_char ands[8] =
|
---|
1049 | {0xff, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe};
|
---|
1050 | c &= ands[oldEntryEndOver];
|
---|
1051 | }
|
---|
1052 | putc (c, outInvfFile);
|
---|
1053 | ++bytesOutput;
|
---|
1054 | ++oldEntryStart;
|
---|
1055 | }
|
---|
1056 | }
|
---|
1057 |
|
---|
1058 | fclose (inInvfFile);
|
---|
1059 |
|
---|
1060 | #ifdef __WIN32__
|
---|
1061 | if (_chsize (_fileno (outInvfFile), bytesOutput) != 0)
|
---|
1062 | Message ("Could not truncate invf.");
|
---|
1063 | #else
|
---|
1064 | if(ftruncate (fileno (outInvfFile), bytesOutput) != 0)
|
---|
1065 | {
|
---|
1066 | fprintf(stderr, "Could not truncate invf");
|
---|
1067 | }
|
---|
1068 | #endif
|
---|
1069 |
|
---|
1070 | fclose (outInvfFile);
|
---|
1071 | }
|
---|
1072 |
|
---|
1073 | static void OutputInvfIdx (char *filename, mg_u_long invfNumBytes) {
|
---|
1074 | FILE *invfIdxFile = create_file (filename, INVF_IDX_SUFFIX, "wb",
|
---|
1075 | MAGIC_INVI, MG_ABORT);
|
---|
1076 |
|
---|
1077 | // process each meaningful byte in the file
|
---|
1078 | mg_u_long numEntries = idh.word_dict_size + idh.tag_dict_size;
|
---|
1079 | mg_u_long entryNum;
|
---|
1080 | for (entryNum = 0; entryNum < numEntries; ++entryNum) {
|
---|
1081 | InvfStateRec &stateRec = invfState.GetRec (entryNum);
|
---|
1082 |
|
---|
1083 | // assumes that inverted entries start at beginning of each byte
|
---|
1084 | if (!WriteUL (invfIdxFile, (stateRec.start >> 3))) break;
|
---|
1085 | }
|
---|
1086 |
|
---|
1087 | WriteUL (invfIdxFile, invfNumBytes);
|
---|
1088 |
|
---|
1089 | fclose (invfIdxFile);
|
---|
1090 | }
|
---|
1091 |
|
---|
1092 |
|
---|
1093 | int done_ivf_2 (const TagInfo &/*tagInfo*/, char *filename) {
|
---|
1094 | // close most open files
|
---|
1095 | if (chunkFile != NULL) {
|
---|
1096 | chunkBuf.done();
|
---|
1097 | fclose (chunkFile);
|
---|
1098 | chunkFile = NULL;
|
---|
1099 | }
|
---|
1100 | occurConvert.Close();
|
---|
1101 |
|
---|
1102 | // free allocated memory
|
---|
1103 | bitPtrs.Clear();
|
---|
1104 | if (ivfMemBuf != NULL) { delete [] ivfMemBuf; ivfMemBuf = NULL; }
|
---|
1105 | free_perf_hash (wordHashDict);
|
---|
1106 | wordHashDict = NULL;
|
---|
1107 | tagMapDict.erase (tagMapDict.begin(), tagMapDict.end());
|
---|
1108 |
|
---|
1109 | // condense the inverted file and truncate it
|
---|
1110 | // this function also writes out the inverted header
|
---|
1111 | mg_u_long invfNumBytes = 0;
|
---|
1112 | CondenseInvfFile (filename, invfNumBytes);
|
---|
1113 |
|
---|
1114 | OutputInvfIdx (filename, invfNumBytes);
|
---|
1115 |
|
---|
1116 | // close the rest of the open files
|
---|
1117 | invfState.Close ();
|
---|
1118 |
|
---|
1119 | return COMPALLOK;
|
---|
1120 | }
|
---|