/[pgestraier]/trunk/pgest.c
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Annotation of /trunk/pgest.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 14 - (hide annotations)
Thu May 26 00:06:10 2005 UTC (18 years, 11 months ago) by dpavlin
File MIME type: text/plain
File size: 9831 byte(s)
tests and fix for limit and offset bug introduced with null handling

1 dpavlin 1 /*
2     * integrate Hyper Estraier into PostgreSQL
3     *
4     * Dobrica Pavlinusic <dpavlin@rot13.org> 2005-05-19
5     *
6     * TODO:
7     * - all
8     *
9     * NOTES:
10     * - clear structures with memset to support hash indexes (who whould like
11     * to create hash index on table returned from function?)
12     * - number of returned rows is set by PostgreSQL evaluator, see:
13     * http://archives.postgresql.org/pgsql-hackers/2005-02/msg00546.php
14     *
15     * Based on:
16     * - C example from PostgreSQL documentation (BSD licence)
17     * - example002.c from Hyper Estraier (GPL)
18     * - _textin/_textout from pgcurl.c (LGPL)
19     *
20     * This code is licenced under GPL
21     */
22    
23     #include "postgres.h"
24     #include "fmgr.h"
25     #include "funcapi.h"
26     #include "utils/builtins.h"
27     #include "utils/array.h"
28     #include "miscadmin.h"
29     #include <estraier.h>
30     #include <cabin.h>
31    
32     #define _textin(str) DirectFunctionCall1(textin, CStringGetDatum(str))
33     #define _textout(str) DatumGetPointer(DirectFunctionCall1(textout, PointerGetDatum(str)))
34     #define GET_STR(textp) DatumGetCString(DirectFunctionCall1(textout, PointerGetDatum(textp)))
35     #define GET_TEXT(cstrp) DatumGetTextP(DirectFunctionCall1(textin, CStringGetDatum(cstrp)))
36    
37 dpavlin 5 /* prototype */
38     char *attr2text(ESTDOC *doc, char *attr);
39 dpavlin 1
40     ESTDB *db;
41     ESTCOND *cond;
42     ESTDOC *doc;
43     const CBLIST *texts;
44     int ecode, *est_result, resnum, i, j;
45 dpavlin 3 int limit = 0;
46     int offset = 0;
47 dpavlin 1
48     /* define PostgreSQL v1 function */
49     PG_FUNCTION_INFO_V1(pgest);
50     Datum pgest(PG_FUNCTION_ARGS) {
51    
52     FuncCallContext *funcctx;
53     int call_cntr;
54     int max_calls;
55     TupleDesc tupdesc;
56     TupleTableSlot *slot;
57     AttInMetadata *attinmeta;
58     char *index_path;
59     char *query;
60 dpavlin 5 char *attr;
61 dpavlin 1
62     /* stuff done only on the first call of the function */
63     if (SRF_IS_FIRSTCALL()) {
64     MemoryContext oldcontext;
65    
66     /* create a function context for cross-call persistence */
67     funcctx = SRF_FIRSTCALL_INIT();
68    
69     /* switch to memory context appropriate for multiple function calls */
70     oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
71 dpavlin 12 /* take arguments from function */
72 dpavlin 1
73 dpavlin 12 /* index path */
74     if (PG_ARGISNULL(0)) {
75     elog(ERROR, "index path can't be null");
76     SRF_RETURN_DONE(funcctx);
77     }
78     index_path = _textout(PG_GETARG_TEXT_P(0));
79    
80     /* query string */
81     if (PG_ARGISNULL(0)) {
82     query = "";
83     } else {
84     query = _textout(PG_GETARG_TEXT_P(1));
85     }
86    
87     /* atribute filter */
88     if (PG_ARGISNULL(2)) {
89     attr = "";
90     } else {
91     attr = _textout(PG_GETARG_TEXT_P(2));
92     }
93    
94     /* limit */
95 dpavlin 14 if (PG_ARGISNULL(3)) {
96     limit = 0;
97     } else {
98     limit = PG_GETARG_INT32(3);
99     }
100 dpavlin 12
101     /* offset */
102 dpavlin 14 if (PG_ARGISNULL(4)) {
103     offset = 0;
104     } else {
105     offset = PG_GETARG_INT32(4);
106     }
107 dpavlin 12
108    
109 dpavlin 1 /* open the database */
110     elog(DEBUG1, "pgest: est_db_open(%s)", index_path);
111    
112     if(!(db = est_db_open(index_path, ESTDBREADER, &ecode))){
113     elog(ERROR, "est_db_open: can't open %s [%d]: %s", index_path, ecode, est_err_msg(ecode));
114     SRF_RETURN_DONE(funcctx);
115     }
116    
117 dpavlin 12 elog(INFO, "pgest: query[%s] attr[%s] limit %d offset %d", query, (PG_ARGISNULL(2) ? "NULL" : attr), limit, offset);
118 dpavlin 1
119     /* create a search condition object */
120     if (!(cond = est_cond_new())) {
121     elog(INFO, "pgest: est_cond_new failed");
122     SRF_RETURN_DONE(funcctx);
123     }
124    
125     /* set the search phrase to the search condition object */
126 dpavlin 12 if (! PG_ARGISNULL(1) && strlen(query) > 0)
127 dpavlin 9 est_cond_set_phrase(cond, query);
128 dpavlin 1
129 dpavlin 5 /* minimum valid attribute length is 10: @a STREQ a */
130 dpavlin 12 if (! PG_ARGISNULL(2) && strlen(attr) >= 10) {
131 dpavlin 5 elog(INFO,"est_cond_add_attr(%s)", attr);
132     est_cond_add_attr(cond, attr);
133     }
134    
135 dpavlin 1 /* get the result of search */
136     est_result = est_db_search(db, cond, &resnum, NULL);
137    
138     /* total number of tuples to be returned */
139 dpavlin 5 if (limit && limit < resnum) {
140     funcctx->max_calls = limit - offset;
141     } else {
142     funcctx->max_calls = resnum - offset;
143     }
144 dpavlin 1
145     /* check if results exists */
146     if ( 0 == funcctx->max_calls )
147     elog(INFO, "pgest: no results for: %s", query );
148    
149 dpavlin 5 elog(DEBUG1, "pgest: found %d hits for %s", resnum, query);
150 dpavlin 1
151     /* Build a tuple description for a __pgest tuple */
152     tupdesc = RelationNameGetTupleDesc("__pgest");
153    
154     /* allocate a slot for a tuple with this tupdesc */
155     slot = TupleDescGetSlot(tupdesc);
156    
157     /* assign slot to function context */
158     funcctx->slot = slot;
159    
160     /*
161     * generate attribute metadata needed later to produce tuples from raw
162     * C strings
163     */
164     attinmeta = TupleDescGetAttInMetadata(tupdesc);
165     funcctx->attinmeta = attinmeta;
166    
167     MemoryContextSwitchTo(oldcontext);
168    
169 dpavlin 2 elog(DEBUG1, "SRF_IS_FIRSTCALL done");
170 dpavlin 1 }
171    
172     /* stuff done on every call of the function */
173     funcctx = SRF_PERCALL_SETUP();
174    
175     call_cntr = funcctx->call_cntr;
176     max_calls = funcctx->max_calls;
177     slot = funcctx->slot;
178     attinmeta = funcctx->attinmeta;
179 dpavlin 3
180     if (limit && call_cntr > limit - 1) {
181     elog(INFO, "call_cntr: %d limit: %d", call_cntr, limit);
182     SRF_RETURN_DONE(funcctx);
183     }
184    
185 dpavlin 1 if (call_cntr < max_calls) {
186     char **values;
187     HeapTuple tuple;
188     Datum result;
189    
190 dpavlin 2 elog(DEBUG1, "pgest: loop count %d", call_cntr);
191 dpavlin 1
192     if (! est_result) {
193     elog(ERROR, "pgest: no estraier results");
194     SRF_RETURN_DONE(funcctx);
195     }
196    
197     /*
198     * Prepare a values array for storage in our slot.
199     * This should be an array of C strings which will
200     * be processed later by the type input functions.
201     */
202    
203 dpavlin 3 if (doc = est_db_get_doc(db, est_result[call_cntr + offset], 0)) {
204 dpavlin 1
205 dpavlin 2 elog(DEBUG1, "URI: %s\n Title: %s\n",
206 dpavlin 1 est_doc_attr(doc, "@uri"),
207     est_doc_attr(doc, "@title")
208     );
209    
210     values = (char **) palloc(4 * sizeof(char *));
211    
212     // values[0] = (char *) palloc(strlen(_estval) * sizeof(char));
213    
214 dpavlin 5 values[0] = (char *) attr2text(doc,"@id");
215     values[1] = (char *) attr2text(doc,"@uri");
216     values[2] = (char *) attr2text(doc,"@title");
217     values[3] = (char *) attr2text(doc,"@type");
218 dpavlin 1
219     /* destloy the document object */
220     elog(DEBUG2, "est_doc_delete");
221     est_doc_delete(doc);
222     } else {
223     elog(INFO, "no result from estraier");
224 dpavlin 7 values[0] = DatumGetCString( "" );
225     values[1] = DatumGetCString( "" );
226     values[2] = DatumGetCString( "" );
227     values[3] = DatumGetCString( "" );
228 dpavlin 1 }
229    
230    
231     elog(DEBUG2, "build tuple");
232     /* build a tuple */
233     tuple = BuildTupleFromCStrings(attinmeta, values);
234    
235     elog(DEBUG2, "make tuple into datum");
236     /* make the tuple into a datum */
237     result = TupleGetDatum(slot, tuple);
238    
239     elog(DEBUG2, "cleanup");
240     /* clean up ? */
241     /*
242     pfree(values[0]);
243     pfree(values[1]);
244     pfree(values[2]);
245     pfree(values[3]);
246     pfree(values);
247     */
248    
249     elog(DEBUG2, "cleanup over");
250    
251     SRF_RETURN_NEXT(funcctx, result);
252     } else {
253 dpavlin 2 elog(DEBUG1, "loop over");
254 dpavlin 1
255     if(!est_db_close(db, &ecode)){
256     elog(INFO, "est_db_close error: %s", est_err_msg(ecode));
257     }
258    
259     /* do when there is no more left */
260     SRF_RETURN_DONE(funcctx);
261     }
262     }
263    
264     /* work in progress */
265     PG_FUNCTION_INFO_V1(pgest2);
266     Datum pgest2(PG_FUNCTION_ARGS)
267     {
268     int nrows = 3;
269     int16 typlen;
270     bool typbyval;
271     char typalign;
272     ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
273     AttInMetadata *attinmeta;
274     TupleDesc tupdesc;
275     Tuplestorestate *tupstore = NULL;
276     HeapTuple tuple;
277     MemoryContext per_query_ctx;
278     MemoryContext oldcontext;
279     Datum dvalue;
280     char **values;
281     int ncols;
282     int i, j;
283    
284     /* check to see if caller supports us returning a tuplestore */
285     if (!rsinfo || !(rsinfo->allowedModes & SFRM_Materialize))
286     ereport(ERROR,
287     (errcode(ERRCODE_SYNTAX_ERROR),
288     errmsg("materialize mode required, but it is not " \
289     "allowed in this context")));
290    
291     /* get the requested return tuple description */
292     tupdesc = rsinfo->expectedDesc;
293     ncols = tupdesc->natts;
294    
295     /*
296     * The requested tuple description better match up with the array
297     * we were given.
298     */
299     /* OK, use it */
300     attinmeta = TupleDescGetAttInMetadata(tupdesc);
301    
302     /* Now go to work */
303     rsinfo->returnMode = SFRM_Materialize;
304    
305     per_query_ctx = fcinfo->flinfo->fn_mcxt;
306     oldcontext = MemoryContextSwitchTo(per_query_ctx);
307    
308     /* initialize our tuplestore */
309     tupstore = tuplestore_begin_heap(true, false, SortMem);
310    
311     values = (char **) palloc(ncols * sizeof(char *));
312    
313     for (i = 0; i < nrows; i++)
314     {
315     for (j = 0; j < ncols; j++)
316     {
317     values[j] = DatumGetCString( "foo" );
318     }
319     /* construct the tuple */
320     tuple = BuildTupleFromCStrings(attinmeta, values);
321    
322     /* now store it */
323     tuplestore_puttuple(tupstore, tuple);
324     }
325    
326     tuplestore_donestoring(tupstore);
327     rsinfo->setResult = tupstore;
328    
329     /*
330     * SFRM_Materialize mode expects us to return a NULL Datum. The actual
331     * tuples are in our tuplestore and passed back through
332     * rsinfo->setResult. rsinfo->setDesc is set to the tuple description
333     * that we actually used to build our tuples with, so the caller can
334     * verify we did what it was expecting.
335     */
336     rsinfo->setDesc = tupdesc;
337     MemoryContextSwitchTo(oldcontext);
338    
339     return (Datum) 0;
340     }
341    
342    
343     /* make text var from attr */
344     char *attr2text(ESTDOC *doc, char *attr) {
345     char *val;
346     const char *attrval;
347     int len;
348 dpavlin 4 int attrlen;
349 dpavlin 1
350 dpavlin 2 elog(DEBUG1, "doc: %08x, attr: %s", doc, attr);
351 dpavlin 1
352 dpavlin 4 if ( (attrval = est_doc_attr(doc, attr)) && (attrlen = strlen(attrval)) ) {
353     val = (char *) palloc(attrlen * sizeof(char));
354 dpavlin 1 } else {
355     return (Datum) NULL;
356     }
357    
358     len = strlen(attrval);
359 dpavlin 2 elog(DEBUG1, "attr2text(%s) = '%s' %d bytes", attr, attrval, len);
360 dpavlin 1
361     len++;
362     len *= sizeof(char);
363    
364     elog(DEBUG2, "palloc(%d)", len);
365    
366     val = palloc(len);
367    
368     memset(val, 0, len);
369     strncpy(val, attrval, len);
370    
371     elog(DEBUG2, "val=%s", val);
372    
373     return val;
374     }
375    
376     /* make integer variable from property */
377     /*
378     char *prop2int(SW_RESULT sw_res, char *propname) {
379     char *val;
380     unsigned long prop;
381     int len;
382    
383     elog(DEBUG2, "prop2int(%s)", propname);
384    
385     prop = estResultPropertyULong( sw_res, propname );
386     if (error_or_abort( est_handle )) return NULL;
387    
388     elog(DEBUG1, "prop2int(%s) = %lu", propname, prop);
389    
390     len = 128 * sizeof(char);
391     elog(DEBUG2, "palloc(%d)", len);
392    
393     val = palloc(len);
394     memset(val, 0, len);
395    
396     snprintf(val, len, "%lu", prop);
397    
398     elog(DEBUG2, "val=%s", val);
399    
400     return val;
401     }
402     */

  ViewVC Help
Powered by ViewVC 1.1.26