1 | /*
|
---|
2 | * MGWrapperImpl.c
|
---|
3 | * Copyright (C) 2002 New Zealand Digital Library, http://www.nzdl.org
|
---|
4 | *
|
---|
5 | * This program is free software; you can redistribute it and/or modify
|
---|
6 | * it under the terms of the GNU General Public License as published by
|
---|
7 | * the Free Software Foundation; either version 2 of the License, or
|
---|
8 | * (at your option) any later version.
|
---|
9 | *
|
---|
10 | * This program is distributed in the hope that it will be useful,
|
---|
11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
13 | * GNU General Public License for more details.
|
---|
14 | *
|
---|
15 | * You should have received a copy of the GNU General Public License
|
---|
16 | * along with this program; if not, write to the Free Software
|
---|
17 | * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
|
---|
18 | */
|
---|
19 |
|
---|
20 | #include <jni.h>
|
---|
21 |
|
---|
22 | #ifdef __MINGW32__
|
---|
23 |
|
---|
24 | // Cross compiling for Windows
|
---|
25 | // Want the type definitions in *win32* version of jni_md.h but
|
---|
26 | // this then leads to C-mangled style functions which we *don't*
|
---|
27 | // want. The following achieves this
|
---|
28 |
|
---|
29 | #undef JNIEXPORT
|
---|
30 | #undef JNIIMPORT
|
---|
31 | #undef JNICALL
|
---|
32 |
|
---|
33 | #define JNIEXPORT
|
---|
34 | #define JNIIMPORT
|
---|
35 | #define JNICALL
|
---|
36 | #endif
|
---|
37 |
|
---|
38 | #include "MGWrapperImpl.h"
|
---|
39 | #include "org_greenstone_mg_MGWrapper.h"
|
---|
40 |
|
---|
41 |
|
---|
42 | #include "backend.h"
|
---|
43 | #include "environment.h"
|
---|
44 | #include "text_get.h"
|
---|
45 | #include "stemmer.h"
|
---|
46 |
|
---|
47 | /*************************************************************************
|
---|
48 | NOTES
|
---|
49 |
|
---|
50 | - Features supported by MGPP but not by MG (AFAIK)
|
---|
51 |
|
---|
52 | - Sorting results by rank.
|
---|
53 | Done here as a post-processing operation. Could be more efficient in
|
---|
54 | some cases: the current solution is not very good when the number of
|
---|
55 | matching documents is large and the number of desired matches is
|
---|
56 | small. In this case it would be better to iterate through the array
|
---|
57 | picking out the best documents rather than sorting them all.
|
---|
58 |
|
---|
59 | - Asking for query term frequencies to be returned.
|
---|
60 | This cannot be turned off in MG. If the term frequencies are not
|
---|
61 | required, they are simply not passed back to the Java side.
|
---|
62 |
|
---|
63 | - Choosing the index for queries.
|
---|
64 | It is possible for MG to build Section and Document indexes (for
|
---|
65 | example), but these are two separate indexes, and must be loaded
|
---|
66 | separately. This module can load more than one index at a time, thus
|
---|
67 | processing queries to different indexes more quickly.
|
---|
68 | NOTE: This replaces TWO options in the MGPP version: returnLevel and
|
---|
69 | queryLevel.
|
---|
70 |
|
---|
71 | *************************************************************************/
|
---|
72 |
|
---|
73 |
|
---|
74 | #define MAX_INDEXES_CACHED 3
|
---|
75 |
|
---|
76 | /* copied from mgquery, needed to convert paragraph numbers to document numbers
|
---|
77 | for greenstone */
|
---|
78 | #if defined(PARADOCNUM) || defined(NZDL)
|
---|
79 | static int GetDocNumFromParaNum(query_data *qd, int paranum) {
|
---|
80 | int Documents = qd->td->cth.num_of_docs;
|
---|
81 | int *Paragraph = qd->paragraph;
|
---|
82 | int low = 1, high = Documents;
|
---|
83 | int mid = (low+high)/2;
|
---|
84 |
|
---|
85 | while ((mid = (low+high)/2) >=1 && mid <= Documents)
|
---|
86 | {
|
---|
87 | if (paranum > Paragraph[mid])
|
---|
88 | low = mid+1;
|
---|
89 | else if (paranum <= Paragraph[mid-1])
|
---|
90 | high = mid-1;
|
---|
91 | else
|
---|
92 | return mid;
|
---|
93 | }
|
---|
94 | FatalError(1, "Bad paragraph number.\n");
|
---|
95 | return 0;
|
---|
96 | }
|
---|
97 | #endif
|
---|
98 |
|
---|
99 |
|
---|
100 | /*********************************************
|
---|
101 | initialisation stuff
|
---|
102 | *********************************************/
|
---|
103 |
|
---|
104 | /* cached ids for java stuff */
|
---|
105 | jfieldID FID_mg_data = NULL; /* MGWrapperData */
|
---|
106 | jfieldID FID_query_result = NULL; /* MGQueryResult */
|
---|
107 | jmethodID MID_addDoc = NULL; /* MGQueryResult.addDoc() */
|
---|
108 | jmethodID MID_addTerm = NULL; /* MGQueryResult.addTerm() */
|
---|
109 | jmethodID MID_addEquivTerm = NULL; /* MGQueryResult.addEquivTerm() */
|
---|
110 | jmethodID MID_setTotalDocs = NULL; /* MGQueryResult.setTotalDocs() */
|
---|
111 | jmethodID MID_clearResult = NULL; /* MGQueryResult.clear() */
|
---|
112 |
|
---|
113 |
|
---|
114 | /* to access objects and methods on java side, need their field/method ids -
|
---|
115 | this initialises them at the start to avoid recalculating them each time they
|
---|
116 | are needed
|
---|
117 | Note: the descriptors need to be exactly right, otherwise you get an error
|
---|
118 | saying "no such field" but no reference to the fact that it has the right
|
---|
119 | name but the wrong type.
|
---|
120 | Note: apparently the jclass is a local ref and should only work
|
---|
121 | in the method that created it. It seems to work ok, but I'll make it
|
---|
122 | global cos the book said I should, and it may avoid future hassles.
|
---|
123 | */
|
---|
124 | JNIEXPORT void JNICALL
|
---|
125 | Java_org_greenstone_mg_MGWrapper_initIDs(JNIEnv *j_env, jclass j_cls)
|
---|
126 | {
|
---|
127 | jclass JC_MGQueryResult;
|
---|
128 |
|
---|
129 | /* a long-"J" */
|
---|
130 | FID_mg_data = (*j_env)->GetFieldID(j_env, j_cls, "mg_data_ptr_", "J");
|
---|
131 | assert(FID_mg_data != NULL);
|
---|
132 |
|
---|
133 | /* an object -"L<class name>;" */
|
---|
134 | FID_query_result = (*j_env)->GetFieldID(j_env, j_cls, "mg_query_result_",
|
---|
135 | "Lorg/greenstone/mg/MGQueryResult;");
|
---|
136 | assert(FID_query_result != NULL);
|
---|
137 |
|
---|
138 | /* the methods we want to use */
|
---|
139 | JC_MGQueryResult = (*j_env)->FindClass(j_env, "org/greenstone/mg/MGQueryResult");
|
---|
140 |
|
---|
141 | /* addDoc(long doc, float rank) */
|
---|
142 | MID_addDoc = (*j_env)->GetMethodID(j_env, JC_MGQueryResult, "addDoc", "(JF)V");
|
---|
143 | assert(MID_addDoc != NULL);
|
---|
144 |
|
---|
145 | /* addTerm(String term, int stem) */
|
---|
146 | MID_addTerm = (*j_env)->GetMethodID(j_env, JC_MGQueryResult, "addTerm", "(Ljava/lang/String;I)V");
|
---|
147 | assert(MID_addTerm != NULL);
|
---|
148 |
|
---|
149 | /* addEquivTerm(String term, String equivTerm, long match, long freq) */
|
---|
150 | MID_addEquivTerm = (*j_env)->GetMethodID(j_env, JC_MGQueryResult, "addEquivTerm", "(Ljava/lang/String;Ljava/lang/String;JJ)V");
|
---|
151 | assert(MID_addEquivTerm != NULL);
|
---|
152 |
|
---|
153 | /* setTotalDocs(long) */
|
---|
154 | MID_setTotalDocs = (*j_env)->GetMethodID(j_env, JC_MGQueryResult, "setTotalDocs", "(J)V");
|
---|
155 | assert(MID_setTotalDocs != NULL);
|
---|
156 |
|
---|
157 | /* clear(void) */
|
---|
158 | MID_clearResult = (*j_env)->GetMethodID(j_env, JC_MGQueryResult, "clear", "()V");
|
---|
159 | assert(MID_clearResult != NULL);
|
---|
160 | }
|
---|
161 |
|
---|
162 |
|
---|
163 | /* the java side MGWrapper has a pointer to a C object - MGWrapperData
|
---|
164 | initialise this and set the pointer
|
---|
165 | */
|
---|
166 | JNIEXPORT jboolean JNICALL
|
---|
167 | Java_org_greenstone_mg_MGWrapper_initCSide(JNIEnv *j_env, jobject j_obj)
|
---|
168 | {
|
---|
169 | /* Allocate a MGWrapperData object to store query parameters */
|
---|
170 | MGWrapperData* data = (MGWrapperData*) malloc(sizeof(MGWrapperData));
|
---|
171 | assert(data != NULL);
|
---|
172 |
|
---|
173 | /* Set default values - no stemming, no case-folding, boolean OR queries */
|
---|
174 | data->defaultStemMethod = 0;
|
---|
175 | data->defaultBoolCombine = 0;
|
---|
176 |
|
---|
177 | /* Allocate a QueryInfo object to store more query parameters */
|
---|
178 | data->queryInfo = (QueryInfo*) malloc(sizeof(QueryInfo));
|
---|
179 | assert(data->queryInfo != NULL);
|
---|
180 |
|
---|
181 | /* Set default values - 50 documents max, return term freqs, sort by rank */
|
---|
182 | data->queryInfo->index = NULL;
|
---|
183 | data->queryInfo->maxDocs = 50;
|
---|
184 | data->queryInfo->needTermFreqs = 1;
|
---|
185 |
|
---|
186 | /* Save the object on the Java side */
|
---|
187 | (*j_env)->SetLongField(j_env, j_obj, FID_mg_data, (long) data);
|
---|
188 |
|
---|
189 | /* Initialise MG environment variables */
|
---|
190 | InitEnv();
|
---|
191 | SetEnv("expert", "true", NULL);
|
---|
192 | SetEnv("mode", "docnums", NULL);
|
---|
193 |
|
---|
194 | return 1; /* true - no errors */
|
---|
195 | }
|
---|
196 |
|
---|
197 |
|
---|
198 | /*******************************************
|
---|
199 | Index caching
|
---|
200 | *******************************************/
|
---|
201 |
|
---|
202 | query_data* cached_indexes[MAX_INDEXES_CACHED] = { NULL };
|
---|
203 |
|
---|
204 |
|
---|
205 | /* Get the index data necessary to perform a query or document retrieval */
|
---|
206 | query_data*
|
---|
207 | loadIndexData(char* base_dir, char* index_path, char* text_path)
|
---|
208 | {
|
---|
209 | char* index_path_name;
|
---|
210 | char* text_path_name;
|
---|
211 | query_data* qd;
|
---|
212 | int i = 0;
|
---|
213 |
|
---|
214 | /* Form the path name of the desired indexes */
|
---|
215 | index_path_name = (char*) malloc(strlen(base_dir) + strlen(index_path) + 1);
|
---|
216 | assert(index_path_name != NULL);
|
---|
217 | strcpy(index_path_name, base_dir);
|
---|
218 | strcat(index_path_name, index_path);
|
---|
219 | printf("Index pathname: %s\n", index_path_name);
|
---|
220 |
|
---|
221 | text_path_name = (char*) malloc(strlen(base_dir) + strlen(text_path) + 1);
|
---|
222 | assert(text_path_name != NULL);
|
---|
223 | strcpy(text_path_name, base_dir);
|
---|
224 | strcat(text_path_name, text_path);
|
---|
225 | printf("Text pathname: %s\n", text_path_name);
|
---|
226 |
|
---|
227 | /* Search through the cached indexes for the desired one */
|
---|
228 | while (i < MAX_INDEXES_CACHED && cached_indexes[i] != NULL) {
|
---|
229 | printf("(Cached) Pathname: %s\n", cached_indexes[i]->pathname);
|
---|
230 | printf("(Cached) Textpathname: %s\n", cached_indexes[i]->textpathname);
|
---|
231 |
|
---|
232 | /* Check if the index has already been loaded */
|
---|
233 | if ((strcmp(index_path_name, cached_indexes[i]->pathname) == 0) &&
|
---|
234 | (strcmp(text_path_name, cached_indexes[i]->textpathname) == 0)) {
|
---|
235 | /* Index has already been loaded and cached, so return it */
|
---|
236 | printf("Found index!\n");
|
---|
237 | free(index_path_name);
|
---|
238 | free(text_path_name);
|
---|
239 | return cached_indexes[i];
|
---|
240 | }
|
---|
241 |
|
---|
242 | i++;
|
---|
243 | }
|
---|
244 |
|
---|
245 | /* Text strings no longer needed */
|
---|
246 | free(index_path_name);
|
---|
247 | free(text_path_name);
|
---|
248 |
|
---|
249 | /* The index is not cached, so load it now */
|
---|
250 | qd = InitQuerySystem(base_dir, index_path, text_path, NULL);
|
---|
251 | if (!qd) {
|
---|
252 | printf("Error: Could not InitQuerySystem()...\n");
|
---|
253 | return NULL;
|
---|
254 | }
|
---|
255 |
|
---|
256 | /* The index loaded OK, so cache it */
|
---|
257 | /* This could be a little more sophisticated, eg. replace least frequently used index */
|
---|
258 | if (i >= MAX_INDEXES_CACHED)
|
---|
259 | i = MAX_INDEXES_CACHED - 1;
|
---|
260 |
|
---|
261 | /* Free the index being replaced */
|
---|
262 | if (cached_indexes[i] != NULL)
|
---|
263 | FinishQuerySystem(cached_indexes[i]);
|
---|
264 |
|
---|
265 | /* Cache the loaded index, and return it */
|
---|
266 | cached_indexes[i] = qd;
|
---|
267 | return cached_indexes[i];
|
---|
268 | }
|
---|
269 |
|
---|
270 |
|
---|
271 | /* Clean up by unloading all cached indexes */
|
---|
272 | JNIEXPORT jboolean JNICALL
|
---|
273 | Java_org_greenstone_mg_MGWrapper_unloadIndexData(JNIEnv* j_env, jobject j_obj)
|
---|
274 | {
|
---|
275 | /* Free all the loaded indexes */
|
---|
276 | int i = 0;
|
---|
277 | while (i < MAX_INDEXES_CACHED && cached_indexes[i] != NULL) {
|
---|
278 | FinishQuerySystem(cached_indexes[i]);
|
---|
279 | cached_indexes[i] = NULL;
|
---|
280 | i++;
|
---|
281 | }
|
---|
282 |
|
---|
283 | return 1; /* true - no errors */
|
---|
284 | }
|
---|
285 |
|
---|
286 |
|
---|
287 | /****************************************************
|
---|
288 | retrieve a document
|
---|
289 | ****************************************************/
|
---|
290 |
|
---|
291 | /* Returns a document from mg as a string */
|
---|
292 | JNIEXPORT jstring JNICALL
|
---|
293 | Java_org_greenstone_mg_MGWrapper_getDocument(JNIEnv *j_env, jobject j_obj,
|
---|
294 | jstring j_base_dir, jstring j_text_path,
|
---|
295 | jlong j_docnum)
|
---|
296 | {
|
---|
297 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
298 |
|
---|
299 | char* index_path;
|
---|
300 | const char* base_dir;
|
---|
301 | const char* text_path;
|
---|
302 | query_data* qd;
|
---|
303 |
|
---|
304 | mg_u_long pos, len;
|
---|
305 | u_char* c_buffer = NULL;
|
---|
306 | u_char* uc_buffer = NULL;
|
---|
307 | int ULen;
|
---|
308 |
|
---|
309 | jstring result;
|
---|
310 |
|
---|
311 | /* Make sure an index has been specified */
|
---|
312 | index_path = data->queryInfo->index;
|
---|
313 | assert(index_path != NULL);
|
---|
314 |
|
---|
315 | /* Obtain C versions of the two string parameters */
|
---|
316 | base_dir = (*j_env)->GetStringUTFChars(j_env, j_base_dir, NULL);
|
---|
317 | if (base_dir == NULL) {
|
---|
318 | return NULL;
|
---|
319 | }
|
---|
320 | text_path = (*j_env)->GetStringUTFChars(j_env, j_text_path, NULL);
|
---|
321 | if (text_path == NULL) {
|
---|
322 | (*j_env)->ReleaseStringUTFChars(j_env, j_base_dir, base_dir);
|
---|
323 | return NULL;
|
---|
324 | }
|
---|
325 |
|
---|
326 | /* Load the appropriate index for satisfying this request */
|
---|
327 | printf("Document retrieval, index path: %s\n", index_path);
|
---|
328 | qd = loadIndexData((char*) base_dir, (char*) index_path, (char*) text_path);
|
---|
329 |
|
---|
330 | /* The C text strings are no longer needed */
|
---|
331 | (*j_env)->ReleaseStringUTFChars(j_env, j_base_dir, base_dir);
|
---|
332 | (*j_env)->ReleaseStringUTFChars(j_env, j_text_path, text_path);
|
---|
333 |
|
---|
334 | /* Check that the index was loaded successfully */
|
---|
335 | if (qd==NULL) {
|
---|
336 | return NULL;
|
---|
337 | }
|
---|
338 | /*assert(qd != NULL);*/
|
---|
339 |
|
---|
340 | /* Get the document position and length in the text file */
|
---|
341 | printf("Fetching document number %d...\n", (mg_u_long) j_docnum);
|
---|
342 | FetchDocStart(qd, (mg_u_long) j_docnum, &pos, &len);
|
---|
343 | printf("Fetched document start. Pos: %d, Len: %d\n", pos, len);
|
---|
344 |
|
---|
345 | /* Allocate memory for the document text (from mg/src/text/mgquery.c:RawDocOutput()) */
|
---|
346 | c_buffer = (u_char*) malloc(len);
|
---|
347 | assert(c_buffer != NULL);
|
---|
348 | uc_buffer = (u_char*) malloc((int) (qd->td->cth.ratio * 1.01 * len) + 100);
|
---|
349 | assert(uc_buffer != NULL);
|
---|
350 |
|
---|
351 | /* Seek to the correct position in the file and read the document text */
|
---|
352 | Fseek (qd->td->TextFile, pos, 0);
|
---|
353 | Fread (c_buffer, 1, len, qd->td->TextFile);
|
---|
354 |
|
---|
355 | /* Decompress the document text into another buffer, and terminate it */
|
---|
356 | DecodeText (qd->cd, c_buffer, len, uc_buffer, &ULen);
|
---|
357 | uc_buffer[ULen] = '\0';
|
---|
358 |
|
---|
359 | /* Load the document text into a Java string */
|
---|
360 | result = (*j_env)->NewStringUTF(j_env, uc_buffer);
|
---|
361 | assert(result != NULL);
|
---|
362 |
|
---|
363 | /* Free C buffers */
|
---|
364 | free(c_buffer);
|
---|
365 | free(uc_buffer);
|
---|
366 |
|
---|
367 | /* Return the document text */
|
---|
368 | return result;
|
---|
369 | }
|
---|
370 |
|
---|
371 |
|
---|
372 | /*******************************************
|
---|
373 | do a query
|
---|
374 | *******************************************/
|
---|
375 |
|
---|
376 | /* do the actual query - the results are written to query_result held on the Java side */
|
---|
377 | JNIEXPORT void JNICALL
|
---|
378 | Java_org_greenstone_mg_MGWrapper_runQuery(JNIEnv *j_env, jobject j_obj,
|
---|
379 | jstring j_base_dir, jstring j_text_path,
|
---|
380 | jstring j_query)
|
---|
381 | {
|
---|
382 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
383 |
|
---|
384 | char* index_path;
|
---|
385 | const char* base_dir;
|
---|
386 | const char* text_path;
|
---|
387 | query_data* qd;
|
---|
388 |
|
---|
389 | jobject result_ptr;
|
---|
390 | char* query;
|
---|
391 | int i, j;
|
---|
392 |
|
---|
393 | jthrowable exc;
|
---|
394 | /* First of all, clear the previous result */
|
---|
395 | /* The result to write to */
|
---|
396 | result_ptr = (*j_env)->GetObjectField(j_env, j_obj, FID_query_result);
|
---|
397 | assert(result_ptr != NULL);
|
---|
398 |
|
---|
399 | /* Clear any previous query results */
|
---|
400 | (*j_env)->CallVoidMethod(j_env, result_ptr, MID_clearResult);
|
---|
401 | exc = (*j_env)->ExceptionOccurred(j_env);
|
---|
402 | if (exc) {
|
---|
403 | (*j_env)->ExceptionDescribe(j_env);
|
---|
404 | return;
|
---|
405 | }
|
---|
406 |
|
---|
407 | /* Make sure an index has been specified */
|
---|
408 | index_path = data->queryInfo->index;
|
---|
409 | if (index_path == NULL) {
|
---|
410 | return;
|
---|
411 | }
|
---|
412 |
|
---|
413 | /* Obtain C versions of the two string parameters */
|
---|
414 | base_dir = (*j_env)->GetStringUTFChars(j_env, j_base_dir, NULL);
|
---|
415 | if (base_dir == NULL) {
|
---|
416 | return;
|
---|
417 | }
|
---|
418 | text_path = (*j_env)->GetStringUTFChars(j_env, j_text_path, NULL);
|
---|
419 | if (text_path == NULL) {
|
---|
420 | (*j_env)->ReleaseStringUTFChars(j_env, j_base_dir, base_dir);
|
---|
421 | return;
|
---|
422 | }
|
---|
423 |
|
---|
424 | /* Load the appropriate index for satisfying this request */
|
---|
425 | qd = loadIndexData((char*) base_dir, (char*) index_path, (char*) text_path);
|
---|
426 |
|
---|
427 | /* The C text strings are no longer needed */
|
---|
428 | (*j_env)->ReleaseStringUTFChars(j_env, j_base_dir, base_dir);
|
---|
429 | (*j_env)->ReleaseStringUTFChars(j_env, j_text_path, text_path);
|
---|
430 |
|
---|
431 | /* Check that the index was loaded successfully */
|
---|
432 | if (qd == NULL) {
|
---|
433 | return;
|
---|
434 | }
|
---|
435 |
|
---|
436 | /* Remove anything hanging around from last time */
|
---|
437 | FreeQueryDocs(qd);
|
---|
438 |
|
---|
439 | /* Obtain a C version of the query string */
|
---|
440 | query = (char*) (*j_env)->GetStringUTFChars(j_env, j_query, NULL);
|
---|
441 | if (query == NULL) {
|
---|
442 | return;
|
---|
443 | }
|
---|
444 | printf("Searching for query \"%s\"...\n", query);
|
---|
445 |
|
---|
446 | /* Make sure the query isn't empty */
|
---|
447 | if (strlen(query) == 0) {
|
---|
448 | printf("Warning: Empty query.\n");
|
---|
449 | return;
|
---|
450 | }
|
---|
451 |
|
---|
452 | /* "Some" queries are done as ranked queries */
|
---|
453 | if (data->defaultBoolCombine == 0) {
|
---|
454 | RankedQueryInfo rqi;
|
---|
455 | rqi.QueryFreqs = 1; /* Use the frequency of each query term in the query - OK? */
|
---|
456 | rqi.Exact = 1; /* Perform exact ranking */
|
---|
457 | rqi.MaxDocsToRetrieve = data->queryInfo->maxDocs; /* Get only the desired number */
|
---|
458 | rqi.MaxParasToRetrieve = rqi.MaxDocsToRetrieve; /* OK? */
|
---|
459 | /* we may need to get more paragraphs to get enough documents. I copied the following from mgquery. it seems to work, not sure why - kjdon */
|
---|
460 | if (qd->id->ifh.InvfLevel == 3 && GetEnv ("maxparas")) {
|
---|
461 | rqi.MaxParasToRetrieve = atoi (GetEnv ("maxparas"));
|
---|
462 | }
|
---|
463 |
|
---|
464 | rqi.Sort = 1; /* Sort the query terms by frequency before ranking */
|
---|
465 | rqi.AccumMethod = 'L'; /* Use a list when accumulating (has bugs though...) */
|
---|
466 | /* rqi.MaxAccums = -1; */ /* Use as many accumulators as necessary - CRASHES with list */
|
---|
467 | rqi.MaxAccums = 100000;
|
---|
468 | rqi.MaxTerms = -1; /* Use all the query terms */
|
---|
469 | /* rqi.StopAtMaxAccum = 0;*/ /* Don't care (using as many accumulators as necessary) */
|
---|
470 | rqi.StopAtMaxAccum = 1;
|
---|
471 | rqi.HashTblSize = 1000; /* Don't care (not using a hash table) */
|
---|
472 | rqi.skip_dump = NULL; /* Don't dump skip information */
|
---|
473 |
|
---|
474 | /* RankedQuery() reads 'casefold' and 'stem' parameters from the environment */
|
---|
475 | SetEnv("casefold", ((data->defaultStemMethod & 1) ? "on" : "off"), NULL);
|
---|
476 | SetEnv("stem", ((data->defaultStemMethod & 2) ? "on" : "off"), NULL);
|
---|
477 |
|
---|
478 | RankedQuery(qd, query, &rqi);
|
---|
479 | }
|
---|
480 | /* "All" queries are done as boolean queries */
|
---|
481 | else {
|
---|
482 | BooleanQueryInfo bqi;
|
---|
483 | bqi.MaxDocsToRetrieve = data->queryInfo->maxDocs;
|
---|
484 |
|
---|
485 | /* Had to add "words$o" to LIB_OBJS in mg/src/text/Makefile and recompile mg for this */
|
---|
486 | BooleanQuery(qd, query, &bqi, data->defaultStemMethod);
|
---|
487 | }
|
---|
488 |
|
---|
489 | /* Finished with the C query string */
|
---|
490 | (*j_env)->ReleaseStringUTFChars(j_env, j_query, query);
|
---|
491 |
|
---|
492 | /* Check the query was processed successfully */
|
---|
493 | if (qd->DL == NULL || qd->QTL == NULL || qd->TL == NULL) {
|
---|
494 | return;
|
---|
495 | }
|
---|
496 |
|
---|
497 | /* Record the total number of matching documents */
|
---|
498 | (*j_env)->CallVoidMethod(j_env, result_ptr, MID_setTotalDocs, (jlong) qd->DL->num);
|
---|
499 | exc = (*j_env)->ExceptionOccurred(j_env);
|
---|
500 | if (exc) {
|
---|
501 | (*j_env)->ExceptionDescribe(j_env);
|
---|
502 | return;
|
---|
503 | }
|
---|
504 |
|
---|
505 | /* Record the matching documents, but only the number requested */
|
---|
506 | printf("Number of matching documents: %d\n", qd->DL->num);
|
---|
507 |
|
---|
508 | for (i = 0; (i < qd->DL->num && i < data->queryInfo->maxDocs); i++) {
|
---|
509 | int doc_num = qd->DL->DE[i].DocNum;
|
---|
510 | float doc_weight = qd->DL->DE[i].Weight;
|
---|
511 |
|
---|
512 | #if defined(PARADOCNUM) || defined(NZDL)
|
---|
513 | if (qd->id->ifh.InvfLevel == 3) {
|
---|
514 | /* pararaph level, need to convert to doc level*/
|
---|
515 | doc_num = GetDocNumFromParaNum(qd, doc_num);
|
---|
516 | }
|
---|
517 | #endif
|
---|
518 |
|
---|
519 |
|
---|
520 | /* Call the addDoc function (Java side) to record a matching document */
|
---|
521 | (*j_env)->CallVoidMethod(j_env, result_ptr, MID_addDoc,
|
---|
522 | (jlong) doc_num, (jfloat) doc_weight);
|
---|
523 | exc = (*j_env)->ExceptionOccurred(j_env);
|
---|
524 | if (exc) {
|
---|
525 | (*j_env)->ExceptionDescribe(j_env);
|
---|
526 | return;
|
---|
527 | }
|
---|
528 | }
|
---|
529 |
|
---|
530 | /* Record the term information, if desired */
|
---|
531 | if (data->queryInfo->needTermFreqs) {
|
---|
532 | /* The following code is a lot more complicated than it could be, but it is necessary
|
---|
533 | to compensate for an oddity in MG. */
|
---|
534 | unsigned char** stemmed_terms = malloc(sizeof(unsigned char*) * qd->TL->num);
|
---|
535 |
|
---|
536 | printf("Number of terms: %d\n", qd->TL->num);
|
---|
537 | printf("Number of query terms: %d\n", qd->QTL->num);
|
---|
538 |
|
---|
539 | /* Generate the stemmed form of each of the relevant terms */
|
---|
540 | for (i = 0; i < qd->TL->num; i++) {
|
---|
541 | u_char* raw_term = qd->TL->TE[i].Word;
|
---|
542 | unsigned int term_length = raw_term[0];
|
---|
543 |
|
---|
544 | u_char* raw_stemmed_term = malloc(term_length + 1);
|
---|
545 | unsigned int stemmed_term_length;
|
---|
546 |
|
---|
547 | /* Copy the term, and stem it */
|
---|
548 | for (j = 0; j <= term_length; j++)
|
---|
549 | raw_stemmed_term[j] = raw_term[j];
|
---|
550 | stemmer(data->defaultStemMethod, qd->sd->sdh.stemmer_num, raw_stemmed_term);
|
---|
551 |
|
---|
552 | /* Allocate memory to store the stemmed term, and fill it */
|
---|
553 | stemmed_term_length = raw_stemmed_term[0];
|
---|
554 | stemmed_terms[i] = malloc(stemmed_term_length + 1);
|
---|
555 | assert(stemmed_terms[i] != NULL);
|
---|
556 | strncpy(stemmed_terms[i], &(raw_stemmed_term[1]), stemmed_term_length);
|
---|
557 | stemmed_terms[i][stemmed_term_length] = '\0';
|
---|
558 | }
|
---|
559 |
|
---|
560 | /* Record every query term, along with their equivalent terms */
|
---|
561 | for (i = 0; i < qd->QTL->num; i++) {
|
---|
562 | u_char* raw_query_term = qd->QTL->QTE[i].Term;
|
---|
563 | unsigned int query_term_length = raw_query_term[0];
|
---|
564 | unsigned char* query_term;
|
---|
565 | jstring j_query_term;
|
---|
566 |
|
---|
567 | u_char* raw_stemmed_query_term = malloc(query_term_length + 1);
|
---|
568 | unsigned int stemmed_query_term_length;
|
---|
569 | unsigned char* stemmed_query_term;
|
---|
570 |
|
---|
571 | /* Allocate memory to store the query term, and fill it */
|
---|
572 | query_term = malloc(query_term_length + 1);
|
---|
573 | assert(query_term != NULL);
|
---|
574 | strncpy(query_term, &(raw_query_term[1]), query_term_length);
|
---|
575 | query_term[query_term_length] = '\0';
|
---|
576 |
|
---|
577 | /* Allocate a new jstring for the query term */
|
---|
578 | j_query_term = (*j_env)->NewStringUTF(j_env, query_term);
|
---|
579 | assert(j_query_term != NULL);
|
---|
580 |
|
---|
581 | /* Call the addTerm function (Java side) to record the query term */
|
---|
582 | (*j_env)->CallVoidMethod(j_env, result_ptr, MID_addTerm,
|
---|
583 | j_query_term, (jint) data->defaultStemMethod);
|
---|
584 | exc = (*j_env)->ExceptionOccurred(j_env);
|
---|
585 | if (exc) {
|
---|
586 | (*j_env)->ExceptionDescribe(j_env);
|
---|
587 | return;
|
---|
588 | }
|
---|
589 |
|
---|
590 | /* Copy the query term, and stem it */
|
---|
591 | for (j = 0; j <= query_term_length; j++)
|
---|
592 | raw_stemmed_query_term[j] = raw_query_term[j];
|
---|
593 | stemmer(data->defaultStemMethod, qd->sd->sdh.stemmer_num, raw_stemmed_query_term);
|
---|
594 |
|
---|
595 | /* Allocate memory to store the stemmed query term, and fill it */
|
---|
596 | stemmed_query_term_length = raw_stemmed_query_term[0];
|
---|
597 | stemmed_query_term = malloc(stemmed_query_term_length + 1);
|
---|
598 | assert(stemmed_query_term != NULL);
|
---|
599 | strncpy(stemmed_query_term, &(raw_stemmed_query_term[1]), stemmed_query_term_length);
|
---|
600 | stemmed_query_term[stemmed_query_term_length] = '\0';
|
---|
601 |
|
---|
602 | /* Find all the terms equivalent to the query term */
|
---|
603 | for (j = 0; j < qd->TL->num; j++) {
|
---|
604 | /* Check if the stemmed query term matches the stemmed term */
|
---|
605 | if (strcmp(stemmed_query_term, stemmed_terms[j]) == 0) {
|
---|
606 | u_char* raw_term = qd->TL->TE[j].Word;
|
---|
607 | unsigned int term_length = raw_term[0];
|
---|
608 | unsigned char* term;
|
---|
609 | jstring j_term;
|
---|
610 |
|
---|
611 | /* Allocate memory to store the query term, and fill it */
|
---|
612 | term = malloc(term_length + 1);
|
---|
613 | assert(term != NULL);
|
---|
614 | strncpy(term, &(raw_term[1]), term_length);
|
---|
615 | term[term_length] = '\0';
|
---|
616 |
|
---|
617 | /* Allocate a new jstring for the query term */
|
---|
618 | j_term = (*j_env)->NewStringUTF(j_env, term);
|
---|
619 | assert(j_term != NULL);
|
---|
620 |
|
---|
621 | /* Call the addEquivTerm function (Java side) to record the equivalent term */
|
---|
622 | (*j_env)->CallVoidMethod(j_env, result_ptr, MID_addEquivTerm,
|
---|
623 | j_query_term, j_term,
|
---|
624 | (jlong) qd->TL->TE[j].WE.doc_count,
|
---|
625 | (jlong) qd->TL->TE[j].WE.count);
|
---|
626 | exc = (*j_env)->ExceptionOccurred(j_env);
|
---|
627 | if (exc) {
|
---|
628 | (*j_env)->ExceptionDescribe(j_env);
|
---|
629 | return;
|
---|
630 | }
|
---|
631 | }
|
---|
632 | }
|
---|
633 | }
|
---|
634 | }
|
---|
635 | }
|
---|
636 |
|
---|
637 |
|
---|
638 | /*******************************************
|
---|
639 | set query options
|
---|
640 | *******************************************/
|
---|
641 |
|
---|
642 | /* Turn casefolding on or off */
|
---|
643 | JNIEXPORT void JNICALL
|
---|
644 | Java_org_greenstone_mg_MGWrapper_setCase(JNIEnv *j_env, jobject j_obj,
|
---|
645 | jboolean j_on)
|
---|
646 | {
|
---|
647 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
648 |
|
---|
649 | if (j_on) {
|
---|
650 | data->defaultStemMethod |= 1;
|
---|
651 | } else {
|
---|
652 | data->defaultStemMethod &= 0xe;
|
---|
653 | }
|
---|
654 | }
|
---|
655 |
|
---|
656 |
|
---|
657 | /* Turn stemming on or off */
|
---|
658 | JNIEXPORT void JNICALL
|
---|
659 | Java_org_greenstone_mg_MGWrapper_setStem(JNIEnv *j_env, jobject j_obj,
|
---|
660 | jboolean j_on)
|
---|
661 | {
|
---|
662 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
663 |
|
---|
664 | if (j_on) {
|
---|
665 | data->defaultStemMethod |= 2;
|
---|
666 | } else {
|
---|
667 | data->defaultStemMethod &= 0xd;
|
---|
668 | }
|
---|
669 | }
|
---|
670 |
|
---|
671 |
|
---|
672 | /* Set the maximum number of documents to return from a query */
|
---|
673 | JNIEXPORT void JNICALL
|
---|
674 | Java_org_greenstone_mg_MGWrapper_setMaxDocs(JNIEnv *j_env, jobject j_obj,
|
---|
675 | jint j_max)
|
---|
676 | {
|
---|
677 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
678 | data->queryInfo->maxDocs = j_max;
|
---|
679 | }
|
---|
680 |
|
---|
681 | /* set the maximum number of numeric to split*/
|
---|
682 | JNIEXPORT void JNICALL
|
---|
683 | Java_org_greenstone_mg_MGWrapper_setMaxNumeric (JNIEnv *j_env,
|
---|
684 | jobject j_obj,
|
---|
685 | jint j_max) {
|
---|
686 |
|
---|
687 | char text[20];
|
---|
688 | char* maxnumeric;
|
---|
689 | sprintf(text,"%d",j_max);
|
---|
690 | maxnumeric = text;
|
---|
691 | SetEnv("maxnumeric",maxnumeric, NULL);
|
---|
692 | }
|
---|
693 |
|
---|
694 |
|
---|
695 | /* Turn term frequency recording on or off */
|
---|
696 | JNIEXPORT void JNICALL
|
---|
697 | Java_org_greenstone_mg_MGWrapper_setReturnTerms(JNIEnv *j_env, jobject j_obj,
|
---|
698 | jboolean j_on)
|
---|
699 | {
|
---|
700 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
701 | data->queryInfo->needTermFreqs = j_on;
|
---|
702 | }
|
---|
703 |
|
---|
704 |
|
---|
705 | /* Choose MG index to search */
|
---|
706 | JNIEXPORT void JNICALL
|
---|
707 | Java_org_greenstone_mg_MGWrapper_setIndex(JNIEnv *j_env, jobject j_obj,
|
---|
708 | jstring j_index)
|
---|
709 | {
|
---|
710 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
711 |
|
---|
712 | /* Get the index name as a C string */
|
---|
713 | const char* index = (*j_env)->GetStringUTFChars(j_env, j_index, NULL);
|
---|
714 | assert(index != NULL);
|
---|
715 | printf("Choosing index %s...\n", index);
|
---|
716 |
|
---|
717 | /* Free the previous index name */
|
---|
718 | if (data->queryInfo->index)
|
---|
719 | free(data->queryInfo->index);
|
---|
720 |
|
---|
721 | /* Allocate memory for the index name, and fill it */
|
---|
722 | data->queryInfo->index = (char*) malloc(strlen(index) + 1);
|
---|
723 | assert(data->queryInfo->index != NULL);
|
---|
724 | strcpy(data->queryInfo->index, index);
|
---|
725 |
|
---|
726 | /* Release the index string */
|
---|
727 | (*j_env)->ReleaseStringUTFChars(j_env, j_index, index);
|
---|
728 | }
|
---|
729 |
|
---|
730 |
|
---|
731 | /* Choose boolean AND or boolean OR queries */
|
---|
732 | JNIEXPORT void JNICALL
|
---|
733 | Java_org_greenstone_mg_MGWrapper_setMatchMode(JNIEnv *j_env, jobject j_obj,
|
---|
734 | jint j_mode)
|
---|
735 | {
|
---|
736 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
737 | data->defaultBoolCombine = j_mode;
|
---|
738 | }
|
---|
739 |
|
---|
740 |
|
---|
741 | /* Get a text representation of the current parameter values */
|
---|
742 | JNIEXPORT jstring JNICALL
|
---|
743 | Java_org_greenstone_mg_MGWrapper_getQueryParams(JNIEnv *j_env, jobject j_obj)
|
---|
744 | {
|
---|
745 | MGWrapperData* data = (MGWrapperData*) (*j_env)->GetLongField(j_env, j_obj, FID_mg_data);
|
---|
746 | char result[512]; /* Assume this is big enough */
|
---|
747 |
|
---|
748 | /* Print the data to a character array */
|
---|
749 | sprintf(result, "Query params:\nindex\t\t%s\ncasefold\t%d\nstem\t\t%d\nquery type\t%s\nmax docs\t%d\n",
|
---|
750 | (data->queryInfo->index == NULL ? "<none loaded>" : data->queryInfo->index),
|
---|
751 | (data->defaultStemMethod & 1),
|
---|
752 | (data->defaultStemMethod & 2),
|
---|
753 | (data->defaultBoolCombine == 1 ? "all" : "some"),
|
---|
754 | (data->queryInfo->maxDocs));
|
---|
755 |
|
---|
756 | /* Convert to a jstring, and return */
|
---|
757 | return (*j_env)->NewStringUTF(j_env, result);
|
---|
758 | }
|
---|