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 |
/* prototype */ |
38 |
char *attr2text(ESTDOC *doc, char *attr); |
39 |
|
40 |
ESTDB *db; |
41 |
ESTCOND *cond; |
42 |
ESTDOC *doc; |
43 |
const CBLIST *texts; |
44 |
int ecode, *est_result, resnum, i, j; |
45 |
int limit = 0; |
46 |
int offset = 0; |
47 |
|
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 |
char *attr; |
61 |
|
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 |
/* take arguments from function */ |
72 |
|
73 |
/* 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 |
if (PG_ARGISNULL(3)) { |
96 |
limit = 0; |
97 |
} else { |
98 |
limit = PG_GETARG_INT32(3); |
99 |
} |
100 |
|
101 |
/* offset */ |
102 |
if (PG_ARGISNULL(4)) { |
103 |
offset = 0; |
104 |
} else { |
105 |
offset = PG_GETARG_INT32(4); |
106 |
} |
107 |
|
108 |
|
109 |
/* 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 |
elog(INFO, "pgest: query[%s] attr[%s] limit %d offset %d", query, (PG_ARGISNULL(2) ? "NULL" : attr), limit, offset); |
118 |
|
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 |
if (! PG_ARGISNULL(1) && strlen(query) > 0) |
127 |
est_cond_set_phrase(cond, query); |
128 |
|
129 |
/* minimum valid attribute length is 10: @a STREQ a */ |
130 |
if (! PG_ARGISNULL(2) && strlen(attr) >= 10) { |
131 |
elog(INFO,"est_cond_add_attr(%s)", attr); |
132 |
est_cond_add_attr(cond, attr); |
133 |
} |
134 |
|
135 |
/* 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 |
if (limit && limit < resnum) { |
140 |
funcctx->max_calls = limit - offset; |
141 |
} else { |
142 |
funcctx->max_calls = resnum - offset; |
143 |
} |
144 |
|
145 |
/* check if results exists */ |
146 |
if ( 0 == funcctx->max_calls ) |
147 |
elog(INFO, "pgest: no results for: %s", query ); |
148 |
|
149 |
elog(DEBUG1, "pgest: found %d hits for %s", resnum, query); |
150 |
|
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 |
elog(DEBUG1, "SRF_IS_FIRSTCALL done"); |
170 |
} |
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 |
|
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 |
if (call_cntr < max_calls) { |
186 |
char **values; |
187 |
HeapTuple tuple; |
188 |
Datum result; |
189 |
|
190 |
elog(DEBUG1, "pgest: loop count %d", call_cntr); |
191 |
|
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 |
if (doc = est_db_get_doc(db, est_result[call_cntr + offset], 0)) { |
204 |
|
205 |
elog(DEBUG1, "URI: %s\n Title: %s\n", |
206 |
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 |
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,"@size"); |
218 |
|
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 |
values[0] = DatumGetCString( "" ); |
225 |
values[1] = DatumGetCString( "" ); |
226 |
values[2] = DatumGetCString( "" ); |
227 |
values[3] = DatumGetCString( "" ); |
228 |
} |
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 |
elog(DEBUG1, "loop over"); |
254 |
|
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(pgest_attr); |
266 |
Datum pgest_attr(PG_FUNCTION_ARGS) |
267 |
{ |
268 |
ArrayType *attr_arr = PG_GETARG_ARRAYTYPE_P(5); |
269 |
Oid element_type = ARR_ELEMTYPE(attr_arr); |
270 |
int ndims = ARR_NDIM(attr_arr); |
271 |
int *dim_counts = ARR_DIMS(attr_arr); |
272 |
int *dim_lower_bounds = ARR_LBOUND(attr_arr); |
273 |
int ncols = 0; |
274 |
int nrows = 0; |
275 |
int indx[MAXDIM]; |
276 |
int16 typlen; |
277 |
bool typbyval; |
278 |
char typalign; |
279 |
ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo; |
280 |
AttInMetadata *attinmeta; |
281 |
TupleDesc tupdesc; |
282 |
Tuplestorestate *tupstore = NULL; |
283 |
HeapTuple tuple; |
284 |
MemoryContext per_query_ctx; |
285 |
MemoryContext oldcontext; |
286 |
Datum dvalue; |
287 |
char **values; |
288 |
int rsinfo_ncols; |
289 |
int i, j; |
290 |
/* estvars */ |
291 |
char *index_path; |
292 |
char *query; |
293 |
char *attr; |
294 |
|
295 |
|
296 |
/* only allow 1D input array */ |
297 |
if (ndims == 1) |
298 |
{ |
299 |
ncols = dim_counts[0]; |
300 |
} |
301 |
else |
302 |
ereport(ERROR, |
303 |
(errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
304 |
errmsg("invalid input array"), |
305 |
errdetail("Input array must have 1 dimension"))); |
306 |
|
307 |
/* check to see if caller supports us returning a tuplestore */ |
308 |
if (!rsinfo || !(rsinfo->allowedModes & SFRM_Materialize)) |
309 |
ereport(ERROR, |
310 |
(errcode(ERRCODE_SYNTAX_ERROR), |
311 |
errmsg("materialize mode required, but it is not " \ |
312 |
"allowed in this context"))); |
313 |
|
314 |
/* get info about element type needed to construct the array */ |
315 |
get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign); |
316 |
|
317 |
/* get the requested return tuple description */ |
318 |
tupdesc = rsinfo->expectedDesc; |
319 |
rsinfo_ncols = tupdesc->natts; |
320 |
|
321 |
/* |
322 |
* The requested tuple description better match up with the array |
323 |
* we were given. |
324 |
*/ |
325 |
if (rsinfo_ncols != ncols) |
326 |
ereport(ERROR, |
327 |
(errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
328 |
errmsg("invalid input array"), |
329 |
errdetail("Number of elements in array must match number of query specified columns."))); |
330 |
|
331 |
/* OK, use it */ |
332 |
attinmeta = TupleDescGetAttInMetadata(tupdesc); |
333 |
|
334 |
/* Now go to work */ |
335 |
rsinfo->returnMode = SFRM_Materialize; |
336 |
|
337 |
per_query_ctx = fcinfo->flinfo->fn_mcxt; |
338 |
oldcontext = MemoryContextSwitchTo(per_query_ctx); |
339 |
|
340 |
/* initialize our tuplestore */ |
341 |
tupstore = tuplestore_begin_heap(true, false, SortMem); |
342 |
|
343 |
|
344 |
/* take rest of arguments from function */ |
345 |
|
346 |
/* index path */ |
347 |
if (PG_ARGISNULL(0)) { |
348 |
ereport(ERROR, |
349 |
(errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
350 |
errmsg("index path can't be null"), |
351 |
errdetail("Index path must be valid full path to HyperEstraier index"))); |
352 |
} |
353 |
index_path = _textout(PG_GETARG_TEXT_P(0)); |
354 |
|
355 |
/* query string */ |
356 |
if (PG_ARGISNULL(0)) { |
357 |
query = ""; |
358 |
} else { |
359 |
query = _textout(PG_GETARG_TEXT_P(1)); |
360 |
} |
361 |
|
362 |
/* atribute filter */ |
363 |
if (PG_ARGISNULL(2)) { |
364 |
attr = ""; |
365 |
} else { |
366 |
attr = _textout(PG_GETARG_TEXT_P(2)); |
367 |
} |
368 |
|
369 |
/* limit */ |
370 |
if (PG_ARGISNULL(3)) { |
371 |
limit = 0; |
372 |
} else { |
373 |
limit = PG_GETARG_INT32(3); |
374 |
} |
375 |
|
376 |
/* offset */ |
377 |
if (PG_ARGISNULL(4)) { |
378 |
offset = 0; |
379 |
} else { |
380 |
offset = PG_GETARG_INT32(4); |
381 |
} |
382 |
|
383 |
|
384 |
/* open the database */ |
385 |
elog(DEBUG1, "pgest_attr: est_db_open(%s)", index_path); |
386 |
|
387 |
if(!(db = est_db_open(index_path, ESTDBREADER, &ecode))){ |
388 |
ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
389 |
errmsg("est_db_open: can't open %s: %d", index_path, ecode), |
390 |
errdetail(est_err_msg(ecode)))); |
391 |
} |
392 |
|
393 |
elog(INFO, "pgest_attr: query[%s] attr[%s] limit %d offset %d", query, (PG_ARGISNULL(2) ? "NULL" : attr), limit, offset); |
394 |
|
395 |
/* create a search condition object */ |
396 |
if (!(cond = est_cond_new())) { |
397 |
ereport(ERROR, (errcode(ERRCODE_QUERY_CANCELED), |
398 |
errmsg("pgest_attr: est_cond_new failed"))); |
399 |
} |
400 |
|
401 |
/* set the search phrase to the search condition object */ |
402 |
if (! PG_ARGISNULL(1) && strlen(query) > 0) |
403 |
est_cond_set_phrase(cond, query); |
404 |
|
405 |
/* minimum valid attribute length is 10: @a STREQ a */ |
406 |
if (! PG_ARGISNULL(2) && strlen(attr) >= 10) { |
407 |
elog(INFO,"est_cond_add_attr(%s)", attr); |
408 |
est_cond_add_attr(cond, attr); |
409 |
} |
410 |
|
411 |
/* get the result of search */ |
412 |
est_result = est_db_search(db, cond, &resnum, NULL); |
413 |
|
414 |
/* check if results exists */ |
415 |
if ( 0 == resnum ) { |
416 |
elog(INFO, "pgest_attr: no results for: %s", query ); |
417 |
} |
418 |
|
419 |
/* total number of tuples to be returned */ |
420 |
if (limit && limit < resnum) { |
421 |
nrows = limit - offset; |
422 |
} else { |
423 |
nrows = resnum - offset; |
424 |
} |
425 |
|
426 |
|
427 |
elog(DEBUG1, "pgest_attr: found %d hits for %s", resnum, query); |
428 |
|
429 |
|
430 |
values = (char **) palloc(ncols * sizeof(char *)); |
431 |
|
432 |
for (i = 0; i < nrows; i++) |
433 |
{ |
434 |
|
435 |
/* get result from estraier */ |
436 |
if (! ( doc = est_db_get_doc(db, est_result[i + offset], 0)) ) { |
437 |
elog(INFO, "can't find result %d", i + offset); |
438 |
} else { |
439 |
elog(DEBUG1, "URI: %s\n Title: %s\n", |
440 |
est_doc_attr(doc, "@uri"), |
441 |
est_doc_attr(doc, "@title") |
442 |
); |
443 |
} |
444 |
|
445 |
/* iterate over results */ |
446 |
for (j = 0; j < ncols; j++) |
447 |
{ |
448 |
bool isnull; |
449 |
|
450 |
/* array value of this position */ |
451 |
indx[0] = j + dim_lower_bounds[0]; |
452 |
|
453 |
dvalue = array_ref(attr_arr, ndims, indx, -1, typlen, typbyval, typalign, &isnull); |
454 |
|
455 |
if (!isnull && doc) |
456 |
values[j] = DatumGetCString( |
457 |
attr2text(doc, |
458 |
DirectFunctionCall1(textout, dvalue) |
459 |
)); |
460 |
else |
461 |
values[j] = NULL; |
462 |
} |
463 |
/* construct the tuple */ |
464 |
tuple = BuildTupleFromCStrings(attinmeta, values); |
465 |
|
466 |
/* now store it */ |
467 |
tuplestore_puttuple(tupstore, tuple); |
468 |
|
469 |
|
470 |
/* delete estraier document object */ |
471 |
est_doc_delete(doc); |
472 |
} |
473 |
|
474 |
tuplestore_donestoring(tupstore); |
475 |
rsinfo->setResult = tupstore; |
476 |
|
477 |
/* |
478 |
* SFRM_Materialize mode expects us to return a NULL Datum. The actual |
479 |
* tuples are in our tuplestore and passed back through |
480 |
* rsinfo->setResult. rsinfo->setDesc is set to the tuple description |
481 |
* that we actually used to build our tuples with, so the caller can |
482 |
* verify we did what it was expecting. |
483 |
*/ |
484 |
rsinfo->setDesc = tupdesc; |
485 |
MemoryContextSwitchTo(oldcontext); |
486 |
|
487 |
if(!est_db_close(db, &ecode)){ |
488 |
ereport(ERROR, (errcode(ERRCODE_IO_ERROR), |
489 |
errmsg("est_db_close: %d", ecode), |
490 |
errdetail(est_err_msg(ecode)))); |
491 |
} |
492 |
|
493 |
return (Datum) 0; |
494 |
} |
495 |
|
496 |
|
497 |
/* make text var from attr */ |
498 |
char *attr2text(ESTDOC *doc, char *attr) { |
499 |
char *val; |
500 |
const char *attrval; |
501 |
int len; |
502 |
int attrlen; |
503 |
|
504 |
elog(DEBUG1, "doc: %08x, attr: %s", doc, attr); |
505 |
|
506 |
if ( (attrval = est_doc_attr(doc, attr)) && (attrlen = strlen(attrval)) ) { |
507 |
val = (char *) palloc(attrlen * sizeof(char)); |
508 |
} else { |
509 |
return (Datum) NULL; |
510 |
} |
511 |
|
512 |
len = strlen(attrval); |
513 |
elog(DEBUG1, "attr2text(%s) = '%s' %d bytes", attr, attrval, len); |
514 |
|
515 |
len++; |
516 |
len *= sizeof(char); |
517 |
|
518 |
elog(DEBUG2, "palloc(%d)", len); |
519 |
|
520 |
val = palloc(len); |
521 |
|
522 |
memset(val, 0, len); |
523 |
strncpy(val, attrval, len); |
524 |
|
525 |
elog(DEBUG2, "val=%s", val); |
526 |
|
527 |
return val; |
528 |
} |
529 |
|