1 | /**************************************************************************
|
---|
2 | *
|
---|
3 | * mg_invf_merge.c -- description
|
---|
4 | * Copyright (C) 1995 Shane Hudson ([email protected])
|
---|
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 | * Last edited: 31 March, 1995
|
---|
21 | * $Id: mg_invf_merge.c 3745 2003-02-20 21:20:24Z mdewsnip $
|
---|
22 | *
|
---|
23 | **************************************************************************/
|
---|
24 |
|
---|
25 | #include "sysfuncs.h"
|
---|
26 |
|
---|
27 | #include "memlib.h"
|
---|
28 | #include "locallib.h"
|
---|
29 | #include "local_strings.h"
|
---|
30 | #include "messages.h"
|
---|
31 | #include "timing.h"
|
---|
32 | #include "netorder.h" /* [RPAP - Jan 97: Endian Ordering] */
|
---|
33 |
|
---|
34 | #include "bitio_m.h"
|
---|
35 | #include "bitio_gen.h"
|
---|
36 | #include "bitio_stdio.h"
|
---|
37 |
|
---|
38 | #include "mg.h"
|
---|
39 | #include "mg_merge.h"
|
---|
40 | #include "mg_files.h"
|
---|
41 | #include "invf.h"
|
---|
42 | #include "words.h"
|
---|
43 |
|
---|
44 | typedef char FileName[256];
|
---|
45 |
|
---|
46 |
|
---|
47 | /*
|
---|
48 | * The merge_info struct contains all the data for an inverted file and
|
---|
49 | * lexicon. We declare 3 merge_info variables, m[OLD] m[NEW] and m[MERGE].
|
---|
50 | */
|
---|
51 | typedef struct merge_info_type
|
---|
52 | {
|
---|
53 | FileName fname;
|
---|
54 | FILE *invf;
|
---|
55 | FILE *dict;
|
---|
56 | FILE *idx;
|
---|
57 | FILE *para; /* [RPAP - Feb 97: Level 3 Merge] */
|
---|
58 | struct invf_dict_header idh;
|
---|
59 | struct invf_file_header ifh;
|
---|
60 | u_long nDocs;
|
---|
61 |
|
---|
62 | /* Lexicon processing variables */
|
---|
63 | u_long fcnt, wcnt;
|
---|
64 | int suff, pref;
|
---|
65 | unsigned char term[MAXSTEMLEN + 1];
|
---|
66 | u_long term_count;
|
---|
67 | int done;
|
---|
68 |
|
---|
69 | /* Variables used for merging */
|
---|
70 | struct stdio_bitio_state sbs; /* for bit-based I/O on inverted files */
|
---|
71 | u_long nTerms; /* = number of invf entries */
|
---|
72 | int *fcntlist; /* store fcnt values read in from the lexicons */
|
---|
73 |
|
---|
74 | }
|
---|
75 | merge_info;
|
---|
76 |
|
---|
77 |
|
---|
78 | /***************************************************
|
---|
79 | **** GLOBALS ****
|
---|
80 | ***************************************************/
|
---|
81 |
|
---|
82 | merge_info m[3]; /* m[OLD], m[NEW] and m[MERGE] */
|
---|
83 | u_long magicsize; /* how big is magic number? currently always 4 */
|
---|
84 | u_long Nstatic; /* static # of documents in file */
|
---|
85 |
|
---|
86 | float *DocWeightBuffer;
|
---|
87 | int weightOption = 0; /* true if weights file is to be appended */
|
---|
88 |
|
---|
89 | /*** Globals for processing the lexicons -- see readTerm() and writeTerm()
|
---|
90 | ***/
|
---|
91 | ProgTime start; /* for printing elapsed time */
|
---|
92 | unsigned char prevTerm[MAXSTEMLEN + 1];
|
---|
93 |
|
---|
94 | /*** Globals for merging -- see processEntry()
|
---|
95 | ***/
|
---|
96 | char *mergedata; /* stores value OLD/NEW/MERGE for each term */
|
---|
97 | int fastMerge = 1; /* True if fast merge selected. Default = true */
|
---|
98 |
|
---|
99 | u_long oldIdxValue; /* stores last read value from m[OLD].idx
|
---|
100 | Used to calculate entry lengths */
|
---|
101 | u_long bytes_output; /* for writing .invf.idx file */
|
---|
102 | u_long oldOnlyBits = 0; /* entries in old only */
|
---|
103 | u_long oldMergeBits = 0; /* entries in old that are merged with
|
---|
104 | a new entry */
|
---|
105 |
|
---|
106 | /**************************************************************************/
|
---|
107 |
|
---|
108 |
|
---|
109 | /*=======================================================================
|
---|
110 | * init_merge_invf(): open files, initialise globals, etc
|
---|
111 | *=======================================================================*/
|
---|
112 | int
|
---|
113 | init_merge_invf ()
|
---|
114 | {
|
---|
115 | prevTerm[0] = prevTerm[1] = 0;
|
---|
116 |
|
---|
117 | /***
|
---|
118 | * open .dict files
|
---|
119 | ***/
|
---|
120 | m[OLD].dict = open_file (m[OLD].fname, INVF_DICT_SUFFIX, "rb",
|
---|
121 | MAGIC_STEM_BUILD, MG_ABORT); /* [RPAP - Feb 97: WIN32 Port] */
|
---|
122 | magicsize = ftell (m[OLD].dict);
|
---|
123 | fread (&(m[OLD].idh), sizeof (m[OLD].idh), 1, m[OLD].dict);
|
---|
124 |
|
---|
125 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
126 | NTOHUL(m[OLD].idh.lookback);
|
---|
127 | NTOHUL(m[OLD].idh.dict_size);
|
---|
128 | NTOHUL(m[OLD].idh.total_bytes);
|
---|
129 | NTOHUL(m[OLD].idh.index_string_bytes);
|
---|
130 | NTOHD(m[OLD].idh.input_bytes); /* [RJM 07/97: 4G limit] */
|
---|
131 | NTOHUL(m[OLD].idh.num_of_docs);
|
---|
132 | NTOHUL(m[OLD].idh.static_num_of_docs);
|
---|
133 | NTOHUL(m[OLD].idh.num_of_words);
|
---|
134 | NTOHUL(m[OLD].idh.stemmer_num);
|
---|
135 | NTOHUL(m[OLD].idh.stem_method);
|
---|
136 |
|
---|
137 | m[NEW].dict = open_file (m[NEW].fname, INVF_DICT_SUFFIX, "rb",
|
---|
138 | MAGIC_STEM_BUILD, MG_ABORT); /* [RPAP - Feb 97: WIN32 Port] */
|
---|
139 | fread (&(m[NEW].idh), sizeof (m[NEW].idh), 1, m[NEW].dict);
|
---|
140 |
|
---|
141 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
142 | NTOHUL(m[NEW].idh.lookback);
|
---|
143 | NTOHUL(m[NEW].idh.dict_size);
|
---|
144 | NTOHUL(m[NEW].idh.total_bytes);
|
---|
145 | NTOHUL(m[NEW].idh.index_string_bytes);
|
---|
146 | NTOHD(m[NEW].idh.input_bytes); /* [RJM 07/97: 4G limit] */
|
---|
147 | NTOHUL(m[NEW].idh.num_of_docs);
|
---|
148 | NTOHUL(m[NEW].idh.static_num_of_docs);
|
---|
149 | NTOHUL(m[NEW].idh.num_of_words);
|
---|
150 | NTOHUL(m[NEW].idh.stemmer_num);
|
---|
151 | NTOHUL(m[NEW].idh.stem_method);
|
---|
152 |
|
---|
153 | m[MERGE].dict = create_file (m[MERGE].fname, INVF_DICT_SUFFIX, "wb",
|
---|
154 | MAGIC_STEM_BUILD, MG_ABORT); /* [RPAP - Feb 97: WIN32 Port] */
|
---|
155 |
|
---|
156 | /* write space for header */
|
---|
157 | fwrite ((char *) &(m[OLD].idh), sizeof (m[OLD].idh), 1, m[MERGE].dict);
|
---|
158 |
|
---|
159 | m[OLD].nDocs = m[OLD].idh.num_of_docs;
|
---|
160 | m[NEW].nDocs = m[NEW].idh.num_of_docs;
|
---|
161 | Nstatic = m[OLD].idh.static_num_of_docs;
|
---|
162 | m[OLD].nTerms = m[OLD].idh.dict_size;
|
---|
163 | m[NEW].nTerms = m[NEW].idh.dict_size;
|
---|
164 | m[MERGE].nDocs = m[OLD].nDocs + m[NEW].nDocs;
|
---|
165 | m[MERGE].nTerms = 0;
|
---|
166 |
|
---|
167 | /* Set up weight buffer for weights of new documents */
|
---|
168 | if (!(DocWeightBuffer = Xmalloc (m[NEW].nDocs * sizeof (*DocWeightBuffer))))
|
---|
169 | {
|
---|
170 | Message ("Insufficient memory\n");
|
---|
171 | return COMPERROR;
|
---|
172 | }
|
---|
173 | bzero ((char *) DocWeightBuffer, m[NEW].nDocs * sizeof (*DocWeightBuffer));
|
---|
174 |
|
---|
175 | /***
|
---|
176 | * open .invf files
|
---|
177 | * Try ".ORG" extension first for m[OLD].invf since an inverted file
|
---|
178 | * with skips may have been created.
|
---|
179 | ****/
|
---|
180 | if (!(m[OLD].invf = open_file (m[OLD].fname, INVF_SUFFIX ".ORG", "rb",
|
---|
181 | MAGIC_INVF, MG_CONTINUE))) /* [RPAP - Feb 97: WIN32 Port] */
|
---|
182 | m[OLD].invf = open_file (m[OLD].fname, INVF_SUFFIX, "rb",
|
---|
183 | MAGIC_INVF, MG_ABORT); /* [RPAP - Feb 97: WIN32 Port] */
|
---|
184 | fread (&(m[OLD].ifh), sizeof (m[OLD].ifh), 1, m[OLD].invf);
|
---|
185 |
|
---|
186 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
187 | NTOHUL(m[OLD].ifh.no_of_words);
|
---|
188 | NTOHUL(m[OLD].ifh.no_of_ptrs);
|
---|
189 | NTOHUL(m[OLD].ifh.skip_mode);
|
---|
190 | {
|
---|
191 | int i;
|
---|
192 | for (i = 0; i < 16; i++)
|
---|
193 | NTOHUL(m[OLD].ifh.params[i]);
|
---|
194 | }
|
---|
195 | NTOHUL(m[OLD].ifh.InvfLevel);
|
---|
196 |
|
---|
197 | m[NEW].invf = open_file (m[NEW].fname, INVF_SUFFIX, "rb",
|
---|
198 | MAGIC_INVF, MG_ABORT);
|
---|
199 | fread (&(m[NEW].ifh), sizeof (m[NEW].ifh), 1, m[NEW].invf);
|
---|
200 |
|
---|
201 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
202 | NTOHUL(m[NEW].ifh.no_of_words);
|
---|
203 | NTOHUL(m[NEW].ifh.no_of_ptrs);
|
---|
204 | NTOHUL(m[NEW].ifh.skip_mode);
|
---|
205 | {
|
---|
206 | int i;
|
---|
207 | for (i = 0; i < 16; i++)
|
---|
208 | NTOHUL(m[NEW].ifh.params[i]);
|
---|
209 | }
|
---|
210 | NTOHUL(m[NEW].ifh.InvfLevel);
|
---|
211 |
|
---|
212 | if (m[OLD].ifh.skip_mode != 0)
|
---|
213 | FatalError (1, "The old invf file contains skips. Unable to merge.");
|
---|
214 |
|
---|
215 | if (m[OLD].ifh.InvfLevel != m[NEW].ifh.InvfLevel)
|
---|
216 | FatalError (1, "The two invf files have different inversion levels!");
|
---|
217 |
|
---|
218 | /********** [RPAP - Feb 97: Level 3 Merge]
|
---|
219 |
|
---|
220 | if ((m[OLD].ifh.InvfLevel > 2) || (m[NEW].ifh.InvfLevel > 2))
|
---|
221 | FatalError (1, "mgmerge cannot handle level 3 inverted files!");
|
---|
222 |
|
---|
223 | *************/
|
---|
224 |
|
---|
225 | m[MERGE].invf = create_file (m[MERGE].fname, INVF_SUFFIX, "wb",
|
---|
226 | MAGIC_INVF, MG_ABORT); /* [RPAP - Feb 97: WIN32 Port] */
|
---|
227 |
|
---|
228 | /* write space for header in inverted file */
|
---|
229 | m[MERGE].ifh.no_of_words = m[MERGE].ifh.no_of_ptrs = 0;
|
---|
230 | m[MERGE].ifh.skip_mode = 0;
|
---|
231 | m[MERGE].ifh.InvfLevel = 2;
|
---|
232 | {
|
---|
233 | int i;
|
---|
234 | for (i = 0; i < 16; i++)
|
---|
235 | m[MERGE].ifh.params[i] = 0;
|
---|
236 | }
|
---|
237 | fwrite ((char *) &(m[MERGE].ifh), sizeof (m[MERGE].ifh), 1, m[MERGE].invf);
|
---|
238 |
|
---|
239 | /***
|
---|
240 | * open .invf.idx files
|
---|
241 | ***/
|
---|
242 | m[OLD].idx = open_file (m[OLD].fname, INVF_IDX_SUFFIX, "rb",
|
---|
243 | MAGIC_INVI, MG_ABORT);
|
---|
244 | fread (&oldIdxValue, sizeof (u_long), 1, m[OLD].idx);
|
---|
245 | NTOHUL(oldIdxValue); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
246 |
|
---|
247 | m[NEW].idx = open_file (m[NEW].fname, INVF_IDX_SUFFIX, "rb",
|
---|
248 | MAGIC_INVI, MG_ABORT);
|
---|
249 |
|
---|
250 | m[MERGE].idx = create_file (m[MERGE].fname, INVF_IDX_SUFFIX, "wb",
|
---|
251 | MAGIC_INVI, MG_ABORT);
|
---|
252 |
|
---|
253 |
|
---|
254 | /* [RPAP - Feb 97: Level 3 Merge] */
|
---|
255 |
|
---|
256 | /***
|
---|
257 | * open .invf.parargraph files
|
---|
258 | ***/
|
---|
259 | if (m[OLD].ifh.InvfLevel == 3)
|
---|
260 | {
|
---|
261 | m[OLD].para = NULL;
|
---|
262 | m[NEW].para = open_file (m[NEW].fname, INVF_PARAGRAPH_SUFFIX, "rb",
|
---|
263 | MAGIC_PARAGRAPH, MG_ABORT);
|
---|
264 | m[MERGE].para = open_file (m[MERGE].fname, INVF_PARAGRAPH_SUFFIX, "rb+",
|
---|
265 | MAGIC_PARAGRAPH, MG_ABORT);
|
---|
266 | }
|
---|
267 | else
|
---|
268 | {
|
---|
269 | m[OLD].para = NULL;
|
---|
270 | m[NEW].para = NULL;
|
---|
271 | m[MERGE].para = NULL;
|
---|
272 | }
|
---|
273 |
|
---|
274 | return OK;
|
---|
275 | }
|
---|
276 |
|
---|
277 |
|
---|
278 | /*=======================================================================
|
---|
279 | * procEntry_slow(): merge an entry, slow method
|
---|
280 | *=======================================================================*/
|
---|
281 | void
|
---|
282 | procEntry_slow (int tnum, int type, u_long oldN, u_long newN, u_long mergeN)
|
---|
283 | /*
|
---|
284 | * Process an invf entry.
|
---|
285 | * Slow Merge. (Always decode/re-encode).
|
---|
286 | *
|
---|
287 | * tnum is the entry number, type = OLD or NEW or MERGE.
|
---|
288 | * oldN, newN and mergeN are the N parameters for Bblock coding.
|
---|
289 | * They are passed so other procEntry_fast can call this one
|
---|
290 | * when a decode/recode is needed to save repeating code.
|
---|
291 | */
|
---|
292 | {
|
---|
293 | int doc = 0;
|
---|
294 | int fcntOLD = m[OLD].fcntlist[tnum];
|
---|
295 | int fcntNEW = m[NEW].fcntlist[tnum];
|
---|
296 | int inblk, outblk;
|
---|
297 | u_long inbits = 0, outbits = 0;
|
---|
298 |
|
---|
299 | /* write .invf.idx pointer */
|
---|
300 | HTONUL(bytes_output); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
301 | fwrite (&bytes_output, sizeof (u_long), 1, m[MERGE].idx);
|
---|
302 | NTOHUL(bytes_output); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
303 |
|
---|
304 | outblk = BIO_Bblock_Init (mergeN, (fcntOLD + fcntNEW));
|
---|
305 |
|
---|
306 | /*** OLD entry -> merge entry ***/
|
---|
307 | if (type != NEW)
|
---|
308 | {
|
---|
309 | int i;
|
---|
310 | inblk = BIO_Bblock_Init (oldN, fcntOLD);
|
---|
311 | for (i = 0; i < fcntOLD; i++)
|
---|
312 | {
|
---|
313 | int num;
|
---|
314 | num = BIO_Stdio_Bblock_Decode (inblk, &(m[OLD].sbs), &inbits);
|
---|
315 | doc += num;
|
---|
316 | BIO_Stdio_Bblock_Encode (num, outblk, &(m[MERGE].sbs), &outbits);
|
---|
317 | if (m[OLD].ifh.InvfLevel >= 2) /* frequencies */
|
---|
318 | {
|
---|
319 | num = BIO_Stdio_Gamma_Decode (&(m[OLD].sbs), &inbits);
|
---|
320 | BIO_Stdio_Gamma_Encode (num, &(m[MERGE].sbs), &outbits);
|
---|
321 | }
|
---|
322 | }
|
---|
323 | if (type == OLD)
|
---|
324 | oldOnlyBits += (inbits + m[OLD].sbs.Btg);
|
---|
325 | else
|
---|
326 | oldMergeBits += (inbits + m[OLD].sbs.Btg);
|
---|
327 | }
|
---|
328 |
|
---|
329 | /*** NEW entry -> merge entry ***/
|
---|
330 | if (type != OLD)
|
---|
331 | {
|
---|
332 | int i;
|
---|
333 | int offset;
|
---|
334 | double logN = log ((double) m[MERGE].nDocs);
|
---|
335 | double idf = logN - log ((double) (fcntOLD + fcntNEW));
|
---|
336 |
|
---|
337 | inblk = BIO_Bblock_Init (newN, fcntNEW);
|
---|
338 | offset = (m[OLD].nDocs - doc); /* amount to be added to 1ST pointer */
|
---|
339 | for (i = 0; i < fcntNEW; i++)
|
---|
340 | {
|
---|
341 | register int num;
|
---|
342 | register double weight;
|
---|
343 | num = BIO_Stdio_Bblock_Decode (inblk, &(m[NEW].sbs), &inbits);
|
---|
344 | if (i == 0)
|
---|
345 | num += offset;
|
---|
346 | doc += num;
|
---|
347 | BIO_Stdio_Bblock_Encode (num, outblk, &(m[MERGE].sbs), &outbits);
|
---|
348 | if (m[OLD].ifh.InvfLevel >= 2) /* frequencies */
|
---|
349 | {
|
---|
350 | num = BIO_Stdio_Gamma_Decode (&(m[NEW].sbs), &inbits);
|
---|
351 | BIO_Stdio_Gamma_Encode (num, &(m[MERGE].sbs), &outbits);
|
---|
352 | weight = num * idf;
|
---|
353 | DocWeightBuffer[doc - m[OLD].nDocs - 1] += weight * weight;
|
---|
354 | }
|
---|
355 | }
|
---|
356 | }
|
---|
357 | /*** Now the padding bits ***/
|
---|
358 | while (m[OLD].sbs.Btg)
|
---|
359 | BIO_Stdio_Decode_Bit (&(m[OLD].sbs));
|
---|
360 | while (m[NEW].sbs.Btg)
|
---|
361 | BIO_Stdio_Decode_Bit (&(m[NEW].sbs));
|
---|
362 | while (m[MERGE].sbs.Btg != 8)
|
---|
363 | {
|
---|
364 | BIO_Stdio_Encode_Bit (0, &(m[MERGE].sbs));
|
---|
365 | outbits++;
|
---|
366 | }
|
---|
367 | bytes_output += (outbits >> 3);
|
---|
368 | return;
|
---|
369 | }
|
---|
370 |
|
---|
371 |
|
---|
372 | /*=======================================================================
|
---|
373 | * procEntry_fast(): merge an entry, WITHOUT decoding if it's in the
|
---|
374 | * old inverted file only
|
---|
375 | *=======================================================================*/
|
---|
376 | void
|
---|
377 | procEntry_fast (int tnum, int type)
|
---|
378 | /*
|
---|
379 | * Merge an invf entry.
|
---|
380 | * Faster Method : Copy (don't re-code) entries for terms only in IFold.
|
---|
381 | * The N parameter is Nstatic, from m[OLD].idh.static_num_of_docs.
|
---|
382 | * If entry type isnt OLD, just call procEntry_slow()
|
---|
383 | */
|
---|
384 | {
|
---|
385 | u_long newIdxValue, len = 0;
|
---|
386 |
|
---|
387 | /* Calculate IFold entry length in bytes if not NEW type */
|
---|
388 | if (type != NEW) /* read in an index number from m[OLD].idx */
|
---|
389 | {
|
---|
390 | fread (&newIdxValue, sizeof (u_long), 1, m[OLD].idx);
|
---|
391 | NTOHUL(newIdxValue); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
392 | len = newIdxValue - oldIdxValue;
|
---|
393 | oldIdxValue = newIdxValue;
|
---|
394 | }
|
---|
395 |
|
---|
396 | if (type != OLD) /* must decode/recode */
|
---|
397 | procEntry_slow (tnum, type, Nstatic, m[NEW].nDocs, Nstatic);
|
---|
398 | else
|
---|
399 | {
|
---|
400 | /* write .invf.idx pointer */
|
---|
401 | HTONUL(bytes_output); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
402 | fwrite (&bytes_output, sizeof (u_long), 1, m[MERGE].idx);
|
---|
403 | NTOHUL(bytes_output); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
404 |
|
---|
405 | /* copy entry */
|
---|
406 | {
|
---|
407 | int i;
|
---|
408 | char c;
|
---|
409 | for (i = 0; i < len; i++)
|
---|
410 | {
|
---|
411 | c = fgetc (m[OLD].invf);
|
---|
412 | fputc (c, m[MERGE].invf);
|
---|
413 | }
|
---|
414 | }
|
---|
415 | bytes_output += len;
|
---|
416 | oldOnlyBits += (len * 8);
|
---|
417 | }
|
---|
418 | return;
|
---|
419 | }
|
---|
420 |
|
---|
421 |
|
---|
422 | /*=======================================================================
|
---|
423 | * processEntry(): merge an entry. Call either procEntry_slow() or
|
---|
424 | * procEntry_fast.
|
---|
425 | *=======================================================================*/
|
---|
426 | void
|
---|
427 | processEntry (int tnum, int type)
|
---|
428 | /*
|
---|
429 | * Read an entry from IFold and/or IFnew, and write to IFmerge.
|
---|
430 | * tnum is the term number, type is OLD or NEW or MERGE
|
---|
431 | */
|
---|
432 | {
|
---|
433 | if (fastMerge)
|
---|
434 | procEntry_fast (tnum, type);
|
---|
435 | else
|
---|
436 | procEntry_slow (tnum, type, Nstatic, m[NEW].nDocs, m[MERGE].nDocs);
|
---|
437 | return;
|
---|
438 | }
|
---|
439 |
|
---|
440 |
|
---|
441 | /*=======================================================================
|
---|
442 | * readTerm()
|
---|
443 | *=======================================================================*/
|
---|
444 | void
|
---|
445 | readTerm (int x)
|
---|
446 | /*
|
---|
447 | * Read the next term from the appropriate ".invf.dict" file.
|
---|
448 | * x should be either OLD or NEW.
|
---|
449 | * Sets m[x].done to true when that lexicon has been completely read.
|
---|
450 | */
|
---|
451 | {
|
---|
452 | int i;
|
---|
453 |
|
---|
454 | if (m[x].term_count >= m[x].nTerms)
|
---|
455 | {
|
---|
456 | m[x].term[0] = 0;
|
---|
457 | m[x].done = 1;
|
---|
458 | return;
|
---|
459 | }
|
---|
460 | m[x].pref = fgetc (m[x].dict);
|
---|
461 | m[x].suff = fgetc (m[x].dict);
|
---|
462 | m[x].term[0] = m[x].pref + m[x].suff;
|
---|
463 |
|
---|
464 | for (i = 0; i < m[x].suff; i++)
|
---|
465 | {
|
---|
466 | m[x].term[m[x].pref + i + 1] = fgetc (m[x].dict);
|
---|
467 | }
|
---|
468 | m[x].term_count = m[x].term_count + 1;
|
---|
469 |
|
---|
470 | fread ((char *) &(m[x].fcnt), sizeof (u_long), 1, m[x].dict);
|
---|
471 | fread ((char *) &(m[x].wcnt), sizeof (u_long), 1, m[x].dict);
|
---|
472 |
|
---|
473 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
474 | NTOHUL(m[x].fcnt);
|
---|
475 | NTOHUL(m[x].wcnt);
|
---|
476 |
|
---|
477 | return;
|
---|
478 | }
|
---|
479 |
|
---|
480 |
|
---|
481 | /*=======================================================================
|
---|
482 | * writeTerm()
|
---|
483 | *=======================================================================*/
|
---|
484 | void
|
---|
485 | writeTerm (int x)
|
---|
486 | /*
|
---|
487 | * Write the current term to m[MERGE].dict file.
|
---|
488 | * x is OLD, NEW or MERGE.
|
---|
489 | */
|
---|
490 | {
|
---|
491 | unsigned char i, prefix, suffix;
|
---|
492 | if (x == MERGE)
|
---|
493 | {
|
---|
494 | m[MERGE].fcnt = m[OLD].fcnt + m[NEW].fcnt;
|
---|
495 | m[MERGE].wcnt = m[OLD].wcnt + m[NEW].wcnt;
|
---|
496 | m[MERGE].pref = m[OLD].pref;
|
---|
497 | m[MERGE].suff = m[OLD].suff;
|
---|
498 | memcpy (m[MERGE].term, m[OLD].term, m[OLD].term[0] + 1);
|
---|
499 | }
|
---|
500 | if (x == OLD)
|
---|
501 | {
|
---|
502 | m[MERGE].fcnt = m[OLD].fcnt;
|
---|
503 | m[MERGE].wcnt = m[OLD].wcnt;
|
---|
504 | }
|
---|
505 | if (x == NEW)
|
---|
506 | {
|
---|
507 | m[MERGE].fcnt = m[NEW].fcnt;
|
---|
508 | m[MERGE].wcnt = m[NEW].wcnt;
|
---|
509 | }
|
---|
510 | prefix = prefixlen (prevTerm, m[x].term);
|
---|
511 | suffix = m[x].term[0] - prefix;
|
---|
512 | fputc (prefix, m[MERGE].dict); /* prefix length */
|
---|
513 | fputc (suffix, m[MERGE].dict); /* suffix length */
|
---|
514 | for (i = 0; i < suffix; i++)
|
---|
515 | fputc (m[x].term[prefix + i + 1], m[MERGE].dict);
|
---|
516 |
|
---|
517 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
518 | HTONUL(m[MERGE].fcnt);
|
---|
519 | HTONUL(m[MERGE].wcnt);
|
---|
520 |
|
---|
521 | fwrite ((char *) &(m[MERGE].fcnt), sizeof (m[MERGE].fcnt), 1, m[MERGE].dict);
|
---|
522 | fwrite ((char *) &(m[MERGE].wcnt), sizeof (m[MERGE].wcnt), 1, m[MERGE].dict);
|
---|
523 |
|
---|
524 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
525 | NTOHUL(m[MERGE].fcnt);
|
---|
526 | NTOHUL(m[MERGE].wcnt);
|
---|
527 |
|
---|
528 | memcpy (prevTerm, m[x].term, prefix + suffix + 1);
|
---|
529 | m[MERGE].idh.total_bytes += m[x].term[0] + 1;
|
---|
530 | m[MERGE].idh.index_string_bytes += m[x].term[0] + 2 - prefix;
|
---|
531 | fflush (m[MERGE].dict);
|
---|
532 | }
|
---|
533 |
|
---|
534 |
|
---|
535 | /*=======================================================================
|
---|
536 | * process_merge_invf(): The main loop
|
---|
537 | *=======================================================================*/
|
---|
538 | int
|
---|
539 | process_merge_invf (void)
|
---|
540 | /*
|
---|
541 | * The main function to merge the lexicons and inverted files.
|
---|
542 | */
|
---|
543 | {
|
---|
544 | /*** Some extra Initialisation stuff first ***/
|
---|
545 |
|
---|
546 | m[OLD].term[0] = m[NEW].term[0] = '\0';
|
---|
547 | m[OLD].term_count = m[NEW].term_count = 0;
|
---|
548 | m[OLD].done = m[NEW].done = 0;
|
---|
549 | m[MERGE].idh.total_bytes = m[MERGE].idh.index_string_bytes = 0;
|
---|
550 |
|
---|
551 | /*** malloc arrays: mergedata, m[OLD].fcntlist, m[NEW].fcntlist ***/
|
---|
552 | mergedata = malloc ((m[OLD].nTerms + m[NEW].nTerms) * sizeof (char));
|
---|
553 | if (mergedata == 0)
|
---|
554 | {
|
---|
555 | fprintf (stderr, "MALLOC error!\n");
|
---|
556 | exit (1);
|
---|
557 | }
|
---|
558 | {
|
---|
559 | int i;
|
---|
560 | for (i = 0; i < 2; i++)
|
---|
561 | {
|
---|
562 | m[i].fcntlist = malloc ((m[OLD].nTerms + m[NEW].nTerms) * sizeof (int));
|
---|
563 | if (m[i].fcntlist == 0)
|
---|
564 | {
|
---|
565 | fprintf (stderr, "MALLOC error!\n");
|
---|
566 | exit (1);
|
---|
567 | }
|
---|
568 | }
|
---|
569 | }
|
---|
570 |
|
---|
571 | /*========================*/
|
---|
572 | /*** LEXICON PASS ***/
|
---|
573 | /*========================*/
|
---|
574 |
|
---|
575 | /*** read first terms ***/
|
---|
576 | readTerm (OLD);
|
---|
577 | readTerm (NEW);
|
---|
578 |
|
---|
579 | m[MERGE].nTerms = 0;
|
---|
580 | while (m[OLD].done == 0 || m[NEW].done == 0)
|
---|
581 | {
|
---|
582 | int i;
|
---|
583 | if (m[OLD].done)
|
---|
584 | i = 1; /* NEW will always be greater */
|
---|
585 | else if (m[NEW].done)
|
---|
586 | i = -1; /* OLD will always be greater */
|
---|
587 | else
|
---|
588 | i = casecompare (m[OLD].term, m[NEW].term);
|
---|
589 |
|
---|
590 | if (i < 0)
|
---|
591 | { /* term in OLD only */
|
---|
592 | mergedata[m[MERGE].nTerms] = (char) OLD;
|
---|
593 | m[OLD].fcntlist[m[MERGE].nTerms] = m[OLD].fcnt;
|
---|
594 | m[NEW].fcntlist[m[MERGE].nTerms] = 0;
|
---|
595 | m[MERGE].nTerms++;
|
---|
596 | writeTerm (OLD);
|
---|
597 | readTerm (OLD);
|
---|
598 | }
|
---|
599 | if (i == 0)
|
---|
600 | { /* term in both lexions */
|
---|
601 | mergedata[m[MERGE].nTerms] = (char) MERGE;
|
---|
602 | m[OLD].fcntlist[m[MERGE].nTerms] = m[OLD].fcnt;
|
---|
603 | m[NEW].fcntlist[m[MERGE].nTerms] = m[NEW].fcnt;
|
---|
604 | m[MERGE].nTerms++;
|
---|
605 | writeTerm (MERGE);
|
---|
606 | readTerm (OLD);
|
---|
607 | readTerm (NEW);
|
---|
608 | }
|
---|
609 | if (i > 0)
|
---|
610 | { /* term in NEW only */
|
---|
611 | mergedata[m[MERGE].nTerms] = (char) NEW;
|
---|
612 | m[NEW].fcntlist[m[MERGE].nTerms] = m[NEW].fcnt;
|
---|
613 | m[OLD].fcntlist[m[MERGE].nTerms] = 0;
|
---|
614 | m[MERGE].nTerms++;
|
---|
615 | writeTerm (NEW);
|
---|
616 | readTerm (NEW);
|
---|
617 | }
|
---|
618 | }/*while*/
|
---|
619 | Message ("%s\n", ElapsedTime (&start, NULL));
|
---|
620 | /* print some results about terms */
|
---|
621 | {
|
---|
622 | fprintf (stderr, " Terms: OLD = %ld, NEW = %ld, MERGE = %ld\n",
|
---|
623 | m[OLD].nTerms, m[NEW].nTerms, m[MERGE].nTerms);
|
---|
624 | fprintf (stderr, " In OLD only: %ld, In NEW only: %ld, BOTH: %ld\n",
|
---|
625 | (m[MERGE].nTerms - m[NEW].nTerms),
|
---|
626 | (m[MERGE].nTerms - m[OLD].nTerms),
|
---|
627 | (m[OLD].nTerms + m[NEW].nTerms - m[MERGE].nTerms));
|
---|
628 | }
|
---|
629 |
|
---|
630 |
|
---|
631 | /*========================*/
|
---|
632 | /*** INVF PASS ***/
|
---|
633 | /*========================*/
|
---|
634 |
|
---|
635 | bytes_output = ftell (m[MERGE].invf);
|
---|
636 | BIO_Stdio_Decode_Start (m[OLD].invf, &m[OLD].sbs);
|
---|
637 | BIO_Stdio_Decode_Start (m[NEW].invf, &m[NEW].sbs);
|
---|
638 | BIO_Stdio_Encode_Start (m[MERGE].invf, &m[MERGE].sbs);
|
---|
639 | {
|
---|
640 | int i;
|
---|
641 | for (i = 0; i < m[MERGE].nTerms; i++)
|
---|
642 | {
|
---|
643 | processEntry (i, mergedata[i]);
|
---|
644 | }
|
---|
645 | /* write final .invf.idx pointer */
|
---|
646 | HTONUL(bytes_output); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
647 | fwrite (&bytes_output, sizeof (u_long), 1, m[MERGE].idx);
|
---|
648 | NTOHUL(bytes_output); /* [RPAP - Jan 97: Endian Ordering] */
|
---|
649 |
|
---|
650 | fprintf (stderr, "Old invf: old only bits: %ld, merged bits: %ld\n",
|
---|
651 | oldOnlyBits, oldMergeBits);
|
---|
652 | }
|
---|
653 |
|
---|
654 | return OK;
|
---|
655 | }
|
---|
656 |
|
---|
657 |
|
---|
658 | /*=======================================================================
|
---|
659 | * done_merge_invf(): write headers, close files, append weights to
|
---|
660 | * weights file
|
---|
661 | *=======================================================================*/
|
---|
662 | int
|
---|
663 | done_merge_invf (void)
|
---|
664 | {
|
---|
665 | /*** write new lexicon header ***/
|
---|
666 | m[MERGE].idh.dict_size = m[MERGE].nTerms;
|
---|
667 | m[MERGE].idh.lookback = m[OLD].idh.lookback;
|
---|
668 | m[MERGE].idh.input_bytes = m[OLD].idh.input_bytes + m[NEW].idh.input_bytes;
|
---|
669 | m[MERGE].idh.total_bytes = m[OLD].idh.total_bytes + m[NEW].idh.total_bytes;
|
---|
670 | m[MERGE].idh.num_of_docs = m[MERGE].nDocs;
|
---|
671 | m[MERGE].idh.num_of_words = m[OLD].idh.num_of_words + m[NEW].idh.num_of_words;
|
---|
672 | m[MERGE].idh.stemmer_num = m[OLD].idh.stemmer_num;
|
---|
673 | m[MERGE].idh.stem_method = m[OLD].idh.stem_method;
|
---|
674 |
|
---|
675 | /* if fastMerge, static num of docs stays the same! */
|
---|
676 | if (fastMerge)
|
---|
677 | m[MERGE].idh.static_num_of_docs = m[OLD].idh.static_num_of_docs;
|
---|
678 | else
|
---|
679 | m[MERGE].idh.static_num_of_docs = m[MERGE].nDocs;
|
---|
680 |
|
---|
681 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
682 | HTONUL(m[MERGE].idh.lookback);
|
---|
683 | HTONUL(m[MERGE].idh.dict_size);
|
---|
684 | HTONUL(m[MERGE].idh.total_bytes);
|
---|
685 | HTONUL(m[MERGE].idh.index_string_bytes);
|
---|
686 | HTOND(m[MERGE].idh.input_bytes); /* [RJM 07/97: 4G limit] */
|
---|
687 | HTONUL(m[MERGE].idh.num_of_docs);
|
---|
688 | HTONUL(m[MERGE].idh.static_num_of_docs);
|
---|
689 | HTONUL(m[MERGE].idh.num_of_words);
|
---|
690 | HTONUL(m[MERGE].idh.stemmer_num);
|
---|
691 | HTONUL(m[MERGE].idh.stem_method);
|
---|
692 |
|
---|
693 | fseek (m[MERGE].dict, magicsize, 0);
|
---|
694 | fwrite (&(m[MERGE].idh), sizeof (struct invf_dict_header), 1, m[MERGE].dict);
|
---|
695 |
|
---|
696 | /*** write new inverted file header ***/
|
---|
697 | m[MERGE].ifh.no_of_words = m[MERGE].nTerms;
|
---|
698 | m[MERGE].ifh.no_of_ptrs = m[OLD].ifh.no_of_ptrs + m[NEW].ifh.no_of_ptrs;
|
---|
699 | m[MERGE].ifh.skip_mode = m[OLD].ifh.skip_mode;
|
---|
700 | m[MERGE].ifh.InvfLevel = m[OLD].ifh.InvfLevel;
|
---|
701 |
|
---|
702 | /* ifh.params[16] -- I have NO IDEA what these do!!!!! */
|
---|
703 | /* only set if a inverted file is in skipped format, I think */
|
---|
704 | {
|
---|
705 | int i;
|
---|
706 | for (i = 0; i < 16; i++)
|
---|
707 | m[MERGE].ifh.params[i] = m[OLD].ifh.params[i];
|
---|
708 | }
|
---|
709 |
|
---|
710 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
711 | HTONUL(m[MERGE].ifh.no_of_words);
|
---|
712 | HTONUL(m[MERGE].ifh.no_of_ptrs);
|
---|
713 | HTONUL(m[MERGE].ifh.skip_mode);
|
---|
714 | {
|
---|
715 | int i;
|
---|
716 | for (i = 0; i < 16; i++)
|
---|
717 | HTONUL(m[MERGE].ifh.params[i]);
|
---|
718 | }
|
---|
719 | HTONUL(m[MERGE].ifh.InvfLevel);
|
---|
720 |
|
---|
721 | fseek (m[MERGE].invf, magicsize, 0); /* go to start of header */
|
---|
722 | fwrite (&(m[MERGE].ifh), sizeof (struct invf_file_header), 1, m[MERGE].invf);
|
---|
723 |
|
---|
724 | /*** Write new document weights --- leave old weights alone ***/
|
---|
725 | if ((weightOption) && (m[OLD].ifh.InvfLevel >= 2))
|
---|
726 | {
|
---|
727 | FILE *weightfile = open_file (m[MERGE].fname, WEIGHTS_SUFFIX, "r+b",
|
---|
728 | MAGIC_WGHT, MG_CONTINUE); /* [RPAP - Feb 97: WIN32 Port] */
|
---|
729 | if (weightfile)
|
---|
730 | {
|
---|
731 | u_long i; /* [RPAP - Jan 97: Endian Ordering] */
|
---|
732 |
|
---|
733 | fprintf (stderr, "writing new weights: Nnew = %ld\n", m[NEW].nDocs);
|
---|
734 | fseek (weightfile, 0, 2); /* end of file */
|
---|
735 |
|
---|
736 | /* [RPAP - Jan 97: Endian Ordering] */
|
---|
737 | for (i = 0; i < m[NEW].nDocs; i++)
|
---|
738 | HTONF(DocWeightBuffer[i]);
|
---|
739 |
|
---|
740 | fwrite ((char *) DocWeightBuffer, sizeof (float), m[NEW].nDocs,
|
---|
741 | weightfile);
|
---|
742 | }
|
---|
743 | fclose (weightfile);
|
---|
744 | }
|
---|
745 |
|
---|
746 |
|
---|
747 | /* [RPAP - Feb 97: Level 3 Merge] */
|
---|
748 | if (m[OLD].ifh.InvfLevel == 3)
|
---|
749 | {
|
---|
750 | /* Update paragraph file */
|
---|
751 | int buffer_size = 1000;
|
---|
752 | u_char *buffer;
|
---|
753 | int num;
|
---|
754 |
|
---|
755 | if (!(buffer = (u_char *) Xmalloc (buffer_size)))
|
---|
756 | FatalError (1, "Could not allocate memory for paragraph buffer");
|
---|
757 |
|
---|
758 | fseek (m[MERGE].para, 0, SEEK_END);
|
---|
759 | while ((num = fread ((char *) buffer, sizeof (*buffer), buffer_size, m[NEW].para)) != 0)
|
---|
760 | fwrite ((char *) buffer, sizeof (*buffer), num, m[MERGE].para);
|
---|
761 |
|
---|
762 | Xfree (buffer);
|
---|
763 | }
|
---|
764 |
|
---|
765 | /*** close files ***/
|
---|
766 | fclose (m[MERGE].invf);
|
---|
767 | fclose (m[OLD].invf);
|
---|
768 | fclose (m[NEW].invf);
|
---|
769 | fclose (m[MERGE].dict);
|
---|
770 | fclose (m[OLD].dict);
|
---|
771 | fclose (m[NEW].dict);
|
---|
772 | fclose (m[MERGE].idx);
|
---|
773 | fclose (m[OLD].idx);
|
---|
774 | fclose (m[NEW].idx);
|
---|
775 |
|
---|
776 | /* [RPAP - Feb 97: Level 3 Merge] */
|
---|
777 | if (m[OLD].ifh.InvfLevel == 3)
|
---|
778 | {
|
---|
779 | fclose (m[NEW].para);
|
---|
780 | fclose (m[MERGE].para);
|
---|
781 | }
|
---|
782 |
|
---|
783 | return OK;
|
---|
784 | }
|
---|
785 |
|
---|
786 |
|
---|
787 | /*=======================================================================
|
---|
788 | * usage()
|
---|
789 | *=======================================================================*/
|
---|
790 | void
|
---|
791 | usage (char *progname)
|
---|
792 | {
|
---|
793 | fprintf (stderr, "usage: %s [-s] [-w] [-d directory] -f mg_collection\n",
|
---|
794 | progname);
|
---|
795 | fprintf (stderr, " example: %s -s -w -f MERGE/mailfiles\n", progname);
|
---|
796 | exit (1);
|
---|
797 | }
|
---|
798 |
|
---|
799 |
|
---|
800 | /*=======================================================================
|
---|
801 | * main()
|
---|
802 | *=======================================================================*/
|
---|
803 | int
|
---|
804 | main (int argc, char *argv[])
|
---|
805 | {
|
---|
806 | int ch;
|
---|
807 | char *progname;
|
---|
808 | progname = argv[0];
|
---|
809 | m[MERGE].fname[0] = '\0';
|
---|
810 |
|
---|
811 | /* message and timing information */
|
---|
812 | msg_prefix = argv[0];
|
---|
813 |
|
---|
814 | while ((ch = getopt (argc, argv, "f:d:swh")) != -1)
|
---|
815 | switch (ch)
|
---|
816 | {
|
---|
817 | case 'f':
|
---|
818 | strcpy (m[MERGE].fname, optarg);
|
---|
819 | break;
|
---|
820 | case 'd':
|
---|
821 | set_basepath (optarg);
|
---|
822 | break;
|
---|
823 | case 's':
|
---|
824 | fastMerge = 0;
|
---|
825 | break;
|
---|
826 | case 'w':
|
---|
827 | weightOption = 1;
|
---|
828 | break;
|
---|
829 | case 'h':
|
---|
830 | case '?':
|
---|
831 | default:
|
---|
832 | usage (progname);
|
---|
833 | }
|
---|
834 |
|
---|
835 | if (m[MERGE].fname[0] == '\0')
|
---|
836 | usage (progname);
|
---|
837 | strcpy (m[OLD].fname, m[MERGE].fname);
|
---|
838 | strcat (m[OLD].fname, ".old");
|
---|
839 | strcpy (m[NEW].fname, m[MERGE].fname);
|
---|
840 | strcat (m[NEW].fname, ".new");
|
---|
841 |
|
---|
842 | GetTime (&start);
|
---|
843 | init_merge_invf ();
|
---|
844 | process_merge_invf ();
|
---|
845 | done_merge_invf ();
|
---|
846 | Message ("%s\n", ElapsedTime (&start, NULL));
|
---|
847 | exit (0);
|
---|
848 | }
|
---|
849 |
|
---|
850 | /*************************************************************************/
|
---|
851 | /* EOF: mg_invf_merge_new.c */
|
---|
852 | /*************************************************************************/
|
---|