14 |
* |
* |
15 |
* Based on: |
* Based on: |
16 |
* - C example from PostgreSQL documentation (BSD licence) |
* - C example from PostgreSQL documentation (BSD licence) |
17 |
* - example002.c from Hyper Estraier (GPL) |
* - coreexample002.c and nodeexample002.c from Hyper Estraier (GPL) |
18 |
* - _textin/_textout from pgcurl.c (LGPL) |
* - _textin/_textout from pgcurl.c (LGPL) |
19 |
* |
* |
20 |
* This code is licenced under GPL |
* This code is licenced under GPL |
28 |
#include "miscadmin.h" |
#include "miscadmin.h" |
29 |
#include <estraier.h> |
#include <estraier.h> |
30 |
#include <cabin.h> |
#include <cabin.h> |
31 |
|
#include <estnode.h> |
32 |
|
|
33 |
#define _textin(str) DirectFunctionCall1(textin, CStringGetDatum(str)) |
#define _textin(str) DirectFunctionCall1(textin, CStringGetDatum(str)) |
34 |
#define _textout(str) DatumGetPointer(DirectFunctionCall1(textout, PointerGetDatum(str))) |
#define _textout(str) DatumGetPointer(DirectFunctionCall1(textout, PointerGetDatum(str))) |
40 |
#define SortMem 16 * 1024 |
#define SortMem 16 * 1024 |
41 |
#endif |
#endif |
42 |
|
|
43 |
|
#define ATTR_DELIMITER "{{!}}" |
44 |
|
|
45 |
/* prototype */ |
/* prototype */ |
46 |
char *attr2text(ESTDOC *doc, char *attr); |
char *attr2text(ESTDOC *doc, char *attr); |
47 |
|
char *node_attr2text(ESTRESDOC *rdoc, char *attr); |
48 |
|
|
49 |
|
|
50 |
/* work in progress */ |
/* work in progress */ |
207 |
|
|
208 |
/* minimum valid attribute length is 10: @a STREQ a */ |
/* minimum valid attribute length is 10: @a STREQ a */ |
209 |
if (! PG_ARGISNULL(2) && strlen(attr) >= 10) { |
if (! PG_ARGISNULL(2) && strlen(attr) >= 10) { |
210 |
elog(DEBUG1,"est_cond_add_attr(%s)", attr); |
elog(DEBUG1,"attributes: %s", attr); |
211 |
est_cond_add_attr(cond, attr); |
char *curr_attr; |
212 |
|
curr_attr = strtok(attr, ATTR_DELIMITER); |
213 |
|
while (curr_attr) { |
214 |
|
elog(DEBUG1,"est_cond_add_attr(%s)", curr_attr); |
215 |
|
est_cond_add_attr(cond, curr_attr); |
216 |
|
curr_attr = strtok(NULL, ATTR_DELIMITER); |
217 |
|
} |
218 |
} |
} |
219 |
|
|
220 |
/* set the search phrase to the search condition object */ |
/* set the search phrase to the search condition object */ |
336 |
|
|
337 |
len++; |
len++; |
338 |
len *= sizeof(char); |
len *= sizeof(char); |
339 |
|
|
340 |
|
elog(DEBUG2, "palloc(%d)", len); |
341 |
|
|
342 |
|
val = palloc(len); |
343 |
|
|
344 |
|
memset(val, 0, len); |
345 |
|
strncpy(val, attrval, len); |
346 |
|
|
347 |
|
elog(DEBUG2, "val=%s", val); |
348 |
|
|
349 |
|
return val; |
350 |
|
} |
351 |
|
|
352 |
|
/* |
353 |
|
* variation on theme: use node API which doesn't open index on |
354 |
|
* every query which is much faster for large indexes |
355 |
|
* |
356 |
|
*/ |
357 |
|
|
358 |
|
PG_FUNCTION_INFO_V1(pgest_node); |
359 |
|
Datum pgest_node(PG_FUNCTION_ARGS) |
360 |
|
{ |
361 |
|
ArrayType *attr_arr = PG_GETARG_ARRAYTYPE_P(8); |
362 |
|
Oid attr_element_type = ARR_ELEMTYPE(attr_arr); |
363 |
|
int attr_ndims = ARR_NDIM(attr_arr); |
364 |
|
int *attr_dim_counts = ARR_DIMS(attr_arr); |
365 |
|
int *attr_dim_lower_bounds = ARR_LBOUND(attr_arr); |
366 |
|
int ncols = 0; |
367 |
|
int nrows = 0; |
368 |
|
int indx[MAXDIM]; |
369 |
|
int16 attr_len; |
370 |
|
bool attr_byval; |
371 |
|
char attr_align; |
372 |
|
ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo; |
373 |
|
AttInMetadata *attinmeta; |
374 |
|
TupleDesc tupdesc; |
375 |
|
Tuplestorestate *tupstore = NULL; |
376 |
|
HeapTuple tuple; |
377 |
|
MemoryContext per_query_ctx; |
378 |
|
MemoryContext oldcontext; |
379 |
|
Datum dvalue; |
380 |
|
char **values; |
381 |
|
int rsinfo_ncols; |
382 |
|
int i, j; |
383 |
|
/* estvars */ |
384 |
|
ESTNODE *node; |
385 |
|
ESTCOND *cond; |
386 |
|
ESTNODERES *nres; |
387 |
|
ESTRESDOC *rdoc; |
388 |
|
const CBLIST *texts; |
389 |
|
int resnum = 0; |
390 |
|
int limit = 0; |
391 |
|
int offset = 0; |
392 |
|
|
393 |
|
char *node_url; |
394 |
|
char *user, *passwd; |
395 |
|
char *query; |
396 |
|
char *attr; |
397 |
|
char *order; |
398 |
|
|
399 |
|
|
400 |
|
/* only allow 1D input array */ |
401 |
|
if (attr_ndims == 1) |
402 |
|
{ |
403 |
|
ncols = attr_dim_counts[0]; |
404 |
|
} |
405 |
|
else |
406 |
|
ereport(ERROR, |
407 |
|
(errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
408 |
|
errmsg("invalid input array"), |
409 |
|
errdetail("Input array must have 1 dimension"))); |
410 |
|
|
411 |
|
/* check to see if caller supports us returning a tuplestore */ |
412 |
|
if (!rsinfo || !(rsinfo->allowedModes & SFRM_Materialize)) |
413 |
|
ereport(ERROR, |
414 |
|
(errcode(ERRCODE_SYNTAX_ERROR), |
415 |
|
errmsg("materialize mode required, but it is not " \ |
416 |
|
"allowed in this context"))); |
417 |
|
|
418 |
|
/* get info about element type needed to construct the array */ |
419 |
|
get_typlenbyvalalign(attr_element_type, &attr_len, &attr_byval, &attr_align); |
420 |
|
|
421 |
|
/* get the requested return tuple description */ |
422 |
|
tupdesc = rsinfo->expectedDesc; |
423 |
|
rsinfo_ncols = tupdesc->natts; |
424 |
|
|
425 |
|
/* |
426 |
|
* The requested tuple description better match up with the array |
427 |
|
* we were given. |
428 |
|
*/ |
429 |
|
if (rsinfo_ncols != ncols) |
430 |
|
ereport(ERROR, |
431 |
|
(errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
432 |
|
errmsg("invalid input array"), |
433 |
|
errdetail("Number of elements in array must match number of query specified columns."))); |
434 |
|
|
435 |
|
/* OK, use it */ |
436 |
|
attinmeta = TupleDescGetAttInMetadata(tupdesc); |
437 |
|
|
438 |
|
/* Now go to work */ |
439 |
|
rsinfo->returnMode = SFRM_Materialize; |
440 |
|
|
441 |
|
per_query_ctx = fcinfo->flinfo->fn_mcxt; |
442 |
|
oldcontext = MemoryContextSwitchTo(per_query_ctx); |
443 |
|
|
444 |
|
/* initialize our tuplestore */ |
445 |
|
tupstore = tuplestore_begin_heap(true, false, SortMem); |
446 |
|
|
447 |
|
|
448 |
|
/* take rest of arguments from function */ |
449 |
|
|
450 |
|
/* node URL */ |
451 |
|
if (PG_ARGISNULL(0)) { |
452 |
|
ereport(ERROR, |
453 |
|
(errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
454 |
|
errmsg("node URL can't be null"), |
455 |
|
errdetail("Node URL must be valid URL to HyperEstraier node"))); |
456 |
|
} |
457 |
|
node_url = _textout(PG_GETARG_TEXT_P(0)); |
458 |
|
|
459 |
|
/* login and password */ |
460 |
|
if (PG_ARGISNULL(1) || PG_ARGISNULL(2)) { |
461 |
|
ereport(ERROR, |
462 |
|
(errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
463 |
|
errmsg("username and password can't be NULL"), |
464 |
|
errdetail("You must specify valid username and password to HyperEstraier node"))); |
465 |
|
} |
466 |
|
user = _textout(PG_GETARG_TEXT_P(1)); |
467 |
|
passwd = _textout(PG_GETARG_TEXT_P(2)); |
468 |
|
|
469 |
|
/* query string */ |
470 |
|
if (PG_ARGISNULL(3)) { |
471 |
|
query = ""; |
472 |
|
} else { |
473 |
|
query = _textout(PG_GETARG_TEXT_P(3)); |
474 |
|
} |
475 |
|
|
476 |
|
/* atribute filter */ |
477 |
|
if (PG_ARGISNULL(4)) { |
478 |
|
attr = ""; |
479 |
|
} else { |
480 |
|
attr = _textout(PG_GETARG_TEXT_P(4)); |
481 |
|
} |
482 |
|
|
483 |
|
/* sort order */ |
484 |
|
if (PG_ARGISNULL(5)) { |
485 |
|
order = ""; |
486 |
|
} else { |
487 |
|
order = _textout(PG_GETARG_TEXT_P(5)); |
488 |
|
} |
489 |
|
|
490 |
|
|
491 |
|
/* limit */ |
492 |
|
if (PG_ARGISNULL(6)) { |
493 |
|
limit = 0; |
494 |
|
} else { |
495 |
|
limit = PG_GETARG_INT32(6); |
496 |
|
} |
497 |
|
|
498 |
|
/* offset */ |
499 |
|
if (PG_ARGISNULL(7)) { |
500 |
|
offset = 0; |
501 |
|
} else { |
502 |
|
offset = PG_GETARG_INT32(7); |
503 |
|
} |
504 |
|
|
505 |
|
/* initialize the network environment */ |
506 |
|
if(!est_init_net_env()){ |
507 |
|
ereport(ERROR, (errcode(ERRCODE_QUERY_CANCELED), |
508 |
|
errmsg("pgest_node: can't create network enviroment"))); |
509 |
|
} |
510 |
|
|
511 |
|
/* create the node connection object */ |
512 |
|
elog(DEBUG1, "pgest_node: est_node_new(%s) as %s", node_url, user); |
513 |
|
node = est_node_new(node_url); |
514 |
|
est_node_set_auth(node, user, passwd); |
515 |
|
|
516 |
|
elog(DEBUG1, "pgest_node: query[%s] attr[%s] limit %d offset %d", query, (PG_ARGISNULL(4) ? "NULL" : attr), limit, offset); |
517 |
|
|
518 |
|
/* create a search condition object */ |
519 |
|
if (!(cond = est_cond_new())) { |
520 |
|
ereport(ERROR, (errcode(ERRCODE_QUERY_CANCELED), |
521 |
|
errmsg("pgest_node: est_cond_new failed"))); |
522 |
|
} |
523 |
|
|
524 |
|
/* set the search phrase to the search condition object */ |
525 |
|
if (! PG_ARGISNULL(3) && strlen(query) > 0) |
526 |
|
est_cond_set_phrase(cond, query); |
527 |
|
|
528 |
|
/* minimum valid attribute length is 10: @a STREQ a */ |
529 |
|
if (! PG_ARGISNULL(4) && strlen(attr) >= 10) { |
530 |
|
elog(DEBUG1,"attributes: %s", attr); |
531 |
|
char *curr_attr; |
532 |
|
curr_attr = strtok(attr, ATTR_DELIMITER); |
533 |
|
while (curr_attr) { |
534 |
|
elog(DEBUG1,"est_cond_add_attr(%s)", curr_attr); |
535 |
|
est_cond_add_attr(cond, curr_attr); |
536 |
|
curr_attr = strtok(NULL, ATTR_DELIMITER); |
537 |
|
} |
538 |
|
} |
539 |
|
|
540 |
|
/* set the search phrase to the search condition object */ |
541 |
|
if (! PG_ARGISNULL(5) && strlen(order) > 0) { |
542 |
|
elog(DEBUG1,"est_cond_set_order(%s)", order); |
543 |
|
est_cond_set_order(cond, order); |
544 |
|
} |
545 |
|
|
546 |
|
if (limit) { |
547 |
|
elog(DEBUG1,"est_cond_set_max(%d)", limit + offset); |
548 |
|
est_cond_set_max(cond, limit + offset); |
549 |
|
} |
550 |
|
|
551 |
|
/* get the result of search */ |
552 |
|
/* FIXME: allow user to specify depath of search */ |
553 |
|
nres = est_node_search(node, cond, 0); |
554 |
|
|
555 |
|
if (! nres) { |
556 |
|
int status = est_node_status(node); |
557 |
|
est_cond_delete(cond); |
558 |
|
est_node_delete(node); |
559 |
|
est_free_net_env(); |
560 |
|
ereport(ERROR, (errcode(ERRCODE_QUERY_CANCELED), |
561 |
|
errmsg("pgest_node: search failed, node status %d", status))); |
562 |
|
} |
563 |
|
|
564 |
|
/* get number of results */ |
565 |
|
resnum = est_noderes_doc_num(nres); |
566 |
|
|
567 |
|
/* check if results exists */ |
568 |
|
if ( 0 == resnum ) { |
569 |
|
elog(INFO, "pgest_node: no results for: %s", query ); |
570 |
|
} |
571 |
|
|
572 |
|
/* total number of tuples to be returned */ |
573 |
|
if (limit && limit < resnum) { |
574 |
|
nrows = limit; |
575 |
|
} else { |
576 |
|
nrows = resnum - offset; |
577 |
|
} |
578 |
|
|
579 |
|
|
580 |
|
elog(DEBUG1, "pgest_node: found %d hits for %s", resnum, query); |
581 |
|
|
582 |
|
|
583 |
|
values = (char **) palloc(ncols * sizeof(char *)); |
584 |
|
|
585 |
|
for (i = 0; i < nrows; i++) |
586 |
|
{ |
587 |
|
|
588 |
|
/* get result from estraier */ |
589 |
|
if (! ( rdoc = est_noderes_get_doc(nres, i + offset) )) { |
590 |
|
elog(INFO, "can't find result %d", i + offset); |
591 |
|
} else { |
592 |
|
elog(DEBUG1, "URI: %s\n Title: %s\n", |
593 |
|
est_resdoc_attr(rdoc, "@uri"), |
594 |
|
est_resdoc_attr(rdoc, "@title") |
595 |
|
); |
596 |
|
} |
597 |
|
|
598 |
|
/* iterate over results */ |
599 |
|
for (j = 0; j < ncols; j++) |
600 |
|
{ |
601 |
|
bool isnull; |
602 |
|
|
603 |
|
/* array value of this position */ |
604 |
|
indx[0] = j + attr_dim_lower_bounds[0]; |
605 |
|
|
606 |
|
dvalue = array_ref(attr_arr, attr_ndims, indx, -1, attr_len, attr_byval, attr_align, &isnull); |
607 |
|
|
608 |
|
if (!isnull && rdoc) |
609 |
|
values[j] = DatumGetCString( |
610 |
|
node_attr2text(rdoc, |
611 |
|
(char *)DirectFunctionCall1(textout, dvalue) |
612 |
|
)); |
613 |
|
else |
614 |
|
values[j] = NULL; |
615 |
|
} |
616 |
|
/* construct the tuple */ |
617 |
|
tuple = BuildTupleFromCStrings(attinmeta, values); |
618 |
|
|
619 |
|
/* now store it */ |
620 |
|
tuplestore_puttuple(tupstore, tuple); |
621 |
|
|
622 |
|
} |
623 |
|
|
624 |
|
tuplestore_donestoring(tupstore); |
625 |
|
rsinfo->setResult = tupstore; |
626 |
|
|
627 |
|
/* |
628 |
|
* SFRM_Materialize mode expects us to return a NULL Datum. The actual |
629 |
|
* tuples are in our tuplestore and passed back through |
630 |
|
* rsinfo->setResult. rsinfo->setDesc is set to the tuple description |
631 |
|
* that we actually used to build our tuples with, so the caller can |
632 |
|
* verify we did what it was expecting. |
633 |
|
*/ |
634 |
|
rsinfo->setDesc = tupdesc; |
635 |
|
MemoryContextSwitchTo(oldcontext); |
636 |
|
|
637 |
|
/* delete the node result object */ |
638 |
|
est_noderes_delete(nres); |
639 |
|
|
640 |
|
/* destroy the search condition object */ |
641 |
|
est_cond_delete(cond); |
642 |
|
|
643 |
|
/* destroy the node object */ |
644 |
|
est_node_delete(node); |
645 |
|
|
646 |
|
/* free the networking environment */ |
647 |
|
est_free_net_env(); |
648 |
|
|
649 |
|
return (Datum) 0; |
650 |
|
} |
651 |
|
|
652 |
|
/* make text var from node attr */ |
653 |
|
char *node_attr2text(ESTRESDOC *rdoc, char *attr) { |
654 |
|
char *val; |
655 |
|
const char *attrval; |
656 |
|
int len; |
657 |
|
int attrlen; |
658 |
|
|
659 |
|
elog(DEBUG1, "doc: %08x, attr: %s", rdoc, attr); |
660 |
|
|
661 |
|
if ( (attrval = est_resdoc_attr(rdoc, attr)) && (attrlen = strlen(attrval)) ) { |
662 |
|
val = (char *) palloc(attrlen * sizeof(char)); |
663 |
|
} else { |
664 |
|
return (Datum) NULL; |
665 |
|
} |
666 |
|
|
667 |
|
len = strlen(attrval); |
668 |
|
elog(DEBUG1, "node_attr2text(%s) = '%s' %d bytes", attr, attrval, len); |
669 |
|
|
670 |
|
len++; |
671 |
|
len *= sizeof(char); |
672 |
|
|
673 |
elog(DEBUG2, "palloc(%d)", len); |
elog(DEBUG2, "palloc(%d)", len); |
674 |
|
|