Line data Source code
1 : /*
2 : * contrib/xml2/xpath.c
3 : *
4 : * Parser interface for DOM-based parser (libxml) rather than
5 : * stream-based SAX-type parser
6 : */
7 : #include "postgres.h"
8 :
9 : #include "access/htup_details.h"
10 : #include "executor/spi.h"
11 : #include "fmgr.h"
12 : #include "funcapi.h"
13 : #include "lib/stringinfo.h"
14 : #include "miscadmin.h"
15 : #include "utils/builtins.h"
16 : #include "utils/xml.h"
17 :
18 : /* libxml includes */
19 :
20 : #include <libxml/xpath.h>
21 : #include <libxml/tree.h>
22 : #include <libxml/xmlmemory.h>
23 : #include <libxml/xmlerror.h>
24 : #include <libxml/parserInternals.h>
25 :
26 2 : PG_MODULE_MAGIC;
27 :
28 : /* exported for use by xslt_proc.c */
29 :
30 : PgXmlErrorContext *pgxml_parser_init(PgXmlStrictness strictness);
31 :
32 : /* workspace for pgxml_xpath() */
33 :
34 : typedef struct
35 : {
36 : xmlDocPtr doctree;
37 : xmlXPathContextPtr ctxt;
38 : xmlXPathObjectPtr res;
39 : } xpath_workspace;
40 :
41 : /* local declarations */
42 :
43 : static xmlChar *pgxmlNodeSetToText(xmlNodeSetPtr nodeset,
44 : xmlChar *toptagname, xmlChar *septagname,
45 : xmlChar *plainsep);
46 :
47 : static text *pgxml_result_to_text(xmlXPathObjectPtr res, xmlChar *toptag,
48 : xmlChar *septag, xmlChar *plainsep);
49 :
50 : static xmlChar *pgxml_texttoxmlchar(text *textstring);
51 :
52 : static xmlXPathObjectPtr pgxml_xpath(text *document, xmlChar *xpath,
53 : xpath_workspace *workspace);
54 :
55 : static void cleanup_workspace(xpath_workspace *workspace);
56 :
57 :
58 : /*
59 : * Initialize for xml parsing.
60 : *
61 : * As with the underlying pg_xml_init function, calls to this MUST be followed
62 : * by a PG_TRY block that guarantees that pg_xml_done is called.
63 : */
64 : PgXmlErrorContext *
65 12 : pgxml_parser_init(PgXmlStrictness strictness)
66 : {
67 : PgXmlErrorContext *xmlerrcxt;
68 :
69 : /* Set up error handling (we share the core's error handler) */
70 12 : xmlerrcxt = pg_xml_init(strictness);
71 :
72 : /* Note: we're assuming an elog cannot be thrown by the following calls */
73 :
74 : /* Initialize libxml */
75 12 : xmlInitParser();
76 :
77 12 : xmlSubstituteEntitiesDefault(1);
78 12 : xmlLoadExtDtdDefaultValue = 1;
79 :
80 12 : return xmlerrcxt;
81 : }
82 :
83 :
84 : /*
85 : * Returns true if document is well-formed
86 : *
87 : * Note: this has been superseded by a core function. We still have to
88 : * have it in the contrib module so that existing SQL-level references
89 : * to the function won't fail; but in normal usage with up-to-date SQL
90 : * definitions for the contrib module, this won't be called.
91 : */
92 :
93 0 : PG_FUNCTION_INFO_V1(xml_is_well_formed);
94 :
95 : Datum
96 0 : xml_is_well_formed(PG_FUNCTION_ARGS)
97 : {
98 0 : text *t = PG_GETARG_TEXT_PP(0); /* document buffer */
99 0 : bool result = false;
100 0 : int32 docsize = VARSIZE_ANY_EXHDR(t);
101 : xmlDocPtr doctree;
102 : PgXmlErrorContext *xmlerrcxt;
103 :
104 0 : xmlerrcxt = pgxml_parser_init(PG_XML_STRICTNESS_LEGACY);
105 :
106 0 : PG_TRY();
107 : {
108 0 : doctree = xmlParseMemory((char *) VARDATA_ANY(t), docsize);
109 :
110 0 : result = (doctree != NULL);
111 :
112 0 : if (doctree != NULL)
113 0 : xmlFreeDoc(doctree);
114 : }
115 0 : PG_CATCH();
116 : {
117 0 : pg_xml_done(xmlerrcxt, true);
118 :
119 0 : PG_RE_THROW();
120 : }
121 0 : PG_END_TRY();
122 :
123 0 : pg_xml_done(xmlerrcxt, false);
124 :
125 0 : PG_RETURN_BOOL(result);
126 : }
127 :
128 :
129 : /* Encodes special characters (<, >, &, " and \r) as XML entities */
130 :
131 2 : PG_FUNCTION_INFO_V1(xml_encode_special_chars);
132 :
133 : Datum
134 0 : xml_encode_special_chars(PG_FUNCTION_ARGS)
135 : {
136 0 : text *tin = PG_GETARG_TEXT_PP(0);
137 : text *tout;
138 : xmlChar *ts,
139 : *tt;
140 :
141 0 : ts = pgxml_texttoxmlchar(tin);
142 :
143 0 : tt = xmlEncodeSpecialChars(NULL, ts);
144 :
145 0 : pfree(ts);
146 :
147 0 : tout = cstring_to_text((char *) tt);
148 :
149 0 : xmlFree(tt);
150 :
151 0 : PG_RETURN_TEXT_P(tout);
152 : }
153 :
154 : /*
155 : * Function translates a nodeset into a text representation
156 : *
157 : * iterates over each node in the set and calls xmlNodeDump to write it to
158 : * an xmlBuffer -from which an xmlChar * string is returned.
159 : *
160 : * each representation is surrounded by <tagname> ... </tagname>
161 : *
162 : * plainsep is an ordinary (not tag) separator - if used, then nodes are
163 : * cast to string as output method
164 : */
165 : static xmlChar *
166 0 : pgxmlNodeSetToText(xmlNodeSetPtr nodeset,
167 : xmlChar *toptagname,
168 : xmlChar *septagname,
169 : xmlChar *plainsep)
170 : {
171 : xmlBufferPtr buf;
172 : xmlChar *result;
173 : int i;
174 :
175 0 : buf = xmlBufferCreate();
176 :
177 0 : if ((toptagname != NULL) && (xmlStrlen(toptagname) > 0))
178 : {
179 0 : xmlBufferWriteChar(buf, "<");
180 0 : xmlBufferWriteCHAR(buf, toptagname);
181 0 : xmlBufferWriteChar(buf, ">");
182 : }
183 0 : if (nodeset != NULL)
184 : {
185 0 : for (i = 0; i < nodeset->nodeNr; i++)
186 : {
187 0 : if (plainsep != NULL)
188 : {
189 0 : xmlBufferWriteCHAR(buf,
190 0 : xmlXPathCastNodeToString(nodeset->nodeTab[i]));
191 :
192 : /* If this isn't the last entry, write the plain sep. */
193 0 : if (i < (nodeset->nodeNr) - 1)
194 0 : xmlBufferWriteChar(buf, (char *) plainsep);
195 : }
196 : else
197 : {
198 0 : if ((septagname != NULL) && (xmlStrlen(septagname) > 0))
199 : {
200 0 : xmlBufferWriteChar(buf, "<");
201 0 : xmlBufferWriteCHAR(buf, septagname);
202 0 : xmlBufferWriteChar(buf, ">");
203 : }
204 0 : xmlNodeDump(buf,
205 0 : nodeset->nodeTab[i]->doc,
206 0 : nodeset->nodeTab[i],
207 : 1, 0);
208 :
209 0 : if ((septagname != NULL) && (xmlStrlen(septagname) > 0))
210 : {
211 0 : xmlBufferWriteChar(buf, "</");
212 0 : xmlBufferWriteCHAR(buf, septagname);
213 0 : xmlBufferWriteChar(buf, ">");
214 : }
215 : }
216 : }
217 : }
218 :
219 0 : if ((toptagname != NULL) && (xmlStrlen(toptagname) > 0))
220 : {
221 0 : xmlBufferWriteChar(buf, "</");
222 0 : xmlBufferWriteCHAR(buf, toptagname);
223 0 : xmlBufferWriteChar(buf, ">");
224 : }
225 0 : result = xmlStrdup(buf->content);
226 0 : xmlBufferFree(buf);
227 0 : return result;
228 : }
229 :
230 :
231 : /* Translate a PostgreSQL "varlena" -i.e. a variable length parameter
232 : * into the libxml2 representation
233 : */
234 : static xmlChar *
235 0 : pgxml_texttoxmlchar(text *textstring)
236 : {
237 0 : return (xmlChar *) text_to_cstring(textstring);
238 : }
239 :
240 : /* Publicly visible XPath functions */
241 :
242 : /*
243 : * This is a "raw" xpath function. Check that it returns child elements
244 : * properly
245 : */
246 2 : PG_FUNCTION_INFO_V1(xpath_nodeset);
247 :
248 : Datum
249 0 : xpath_nodeset(PG_FUNCTION_ARGS)
250 : {
251 0 : text *document = PG_GETARG_TEXT_PP(0);
252 0 : text *xpathsupp = PG_GETARG_TEXT_PP(1); /* XPath expression */
253 0 : xmlChar *toptag = pgxml_texttoxmlchar(PG_GETARG_TEXT_PP(2));
254 0 : xmlChar *septag = pgxml_texttoxmlchar(PG_GETARG_TEXT_PP(3));
255 : xmlChar *xpath;
256 : text *xpres;
257 : xmlXPathObjectPtr res;
258 : xpath_workspace workspace;
259 :
260 0 : xpath = pgxml_texttoxmlchar(xpathsupp);
261 :
262 0 : res = pgxml_xpath(document, xpath, &workspace);
263 :
264 0 : xpres = pgxml_result_to_text(res, toptag, septag, NULL);
265 :
266 0 : cleanup_workspace(&workspace);
267 :
268 0 : pfree(xpath);
269 :
270 0 : if (xpres == NULL)
271 0 : PG_RETURN_NULL();
272 0 : PG_RETURN_TEXT_P(xpres);
273 : }
274 :
275 : /*
276 : * The following function is almost identical, but returns the elements in
277 : * a list.
278 : */
279 2 : PG_FUNCTION_INFO_V1(xpath_list);
280 :
281 : Datum
282 0 : xpath_list(PG_FUNCTION_ARGS)
283 : {
284 0 : text *document = PG_GETARG_TEXT_PP(0);
285 0 : text *xpathsupp = PG_GETARG_TEXT_PP(1); /* XPath expression */
286 0 : xmlChar *plainsep = pgxml_texttoxmlchar(PG_GETARG_TEXT_PP(2));
287 : xmlChar *xpath;
288 : text *xpres;
289 : xmlXPathObjectPtr res;
290 : xpath_workspace workspace;
291 :
292 0 : xpath = pgxml_texttoxmlchar(xpathsupp);
293 :
294 0 : res = pgxml_xpath(document, xpath, &workspace);
295 :
296 0 : xpres = pgxml_result_to_text(res, NULL, NULL, plainsep);
297 :
298 0 : cleanup_workspace(&workspace);
299 :
300 0 : pfree(xpath);
301 :
302 0 : if (xpres == NULL)
303 0 : PG_RETURN_NULL();
304 0 : PG_RETURN_TEXT_P(xpres);
305 : }
306 :
307 :
308 4 : PG_FUNCTION_INFO_V1(xpath_string);
309 :
310 : Datum
311 2 : xpath_string(PG_FUNCTION_ARGS)
312 : {
313 2 : text *document = PG_GETARG_TEXT_PP(0);
314 2 : text *xpathsupp = PG_GETARG_TEXT_PP(1); /* XPath expression */
315 : xmlChar *xpath;
316 : int32 pathsize;
317 : text *xpres;
318 : xmlXPathObjectPtr res;
319 : xpath_workspace workspace;
320 :
321 2 : pathsize = VARSIZE_ANY_EXHDR(xpathsupp);
322 :
323 : /*
324 : * We encapsulate the supplied path with "string()" = 8 chars + 1 for NUL
325 : * at end
326 : */
327 : /* We could try casting to string using the libxml function? */
328 :
329 2 : xpath = (xmlChar *) palloc(pathsize + 9);
330 2 : memcpy((char *) xpath, "string(", 7);
331 2 : memcpy((char *) (xpath + 7), VARDATA_ANY(xpathsupp), pathsize);
332 2 : xpath[pathsize + 7] = ')';
333 2 : xpath[pathsize + 8] = '\0';
334 :
335 2 : res = pgxml_xpath(document, xpath, &workspace);
336 :
337 2 : xpres = pgxml_result_to_text(res, NULL, NULL, NULL);
338 :
339 2 : cleanup_workspace(&workspace);
340 :
341 2 : pfree(xpath);
342 :
343 2 : if (xpres == NULL)
344 2 : PG_RETURN_NULL();
345 0 : PG_RETURN_TEXT_P(xpres);
346 : }
347 :
348 :
349 2 : PG_FUNCTION_INFO_V1(xpath_number);
350 :
351 : Datum
352 0 : xpath_number(PG_FUNCTION_ARGS)
353 : {
354 0 : text *document = PG_GETARG_TEXT_PP(0);
355 0 : text *xpathsupp = PG_GETARG_TEXT_PP(1); /* XPath expression */
356 : xmlChar *xpath;
357 : float4 fRes;
358 : xmlXPathObjectPtr res;
359 : xpath_workspace workspace;
360 :
361 0 : xpath = pgxml_texttoxmlchar(xpathsupp);
362 :
363 0 : res = pgxml_xpath(document, xpath, &workspace);
364 :
365 0 : pfree(xpath);
366 :
367 0 : if (res == NULL)
368 0 : PG_RETURN_NULL();
369 :
370 0 : fRes = xmlXPathCastToNumber(res);
371 :
372 0 : cleanup_workspace(&workspace);
373 :
374 0 : if (xmlXPathIsNaN(fRes))
375 0 : PG_RETURN_NULL();
376 :
377 0 : PG_RETURN_FLOAT4(fRes);
378 : }
379 :
380 :
381 2 : PG_FUNCTION_INFO_V1(xpath_bool);
382 :
383 : Datum
384 0 : xpath_bool(PG_FUNCTION_ARGS)
385 : {
386 0 : text *document = PG_GETARG_TEXT_PP(0);
387 0 : text *xpathsupp = PG_GETARG_TEXT_PP(1); /* XPath expression */
388 : xmlChar *xpath;
389 : int bRes;
390 : xmlXPathObjectPtr res;
391 : xpath_workspace workspace;
392 :
393 0 : xpath = pgxml_texttoxmlchar(xpathsupp);
394 :
395 0 : res = pgxml_xpath(document, xpath, &workspace);
396 :
397 0 : pfree(xpath);
398 :
399 0 : if (res == NULL)
400 0 : PG_RETURN_BOOL(false);
401 :
402 0 : bRes = xmlXPathCastToBoolean(res);
403 :
404 0 : cleanup_workspace(&workspace);
405 :
406 0 : PG_RETURN_BOOL(bRes);
407 : }
408 :
409 :
410 :
411 : /* Core function to evaluate XPath query */
412 :
413 : static xmlXPathObjectPtr
414 2 : pgxml_xpath(text *document, xmlChar *xpath, xpath_workspace *workspace)
415 : {
416 2 : int32 docsize = VARSIZE_ANY_EXHDR(document);
417 : PgXmlErrorContext *xmlerrcxt;
418 : xmlXPathCompExprPtr comppath;
419 :
420 2 : workspace->doctree = NULL;
421 2 : workspace->ctxt = NULL;
422 2 : workspace->res = NULL;
423 :
424 2 : xmlerrcxt = pgxml_parser_init(PG_XML_STRICTNESS_LEGACY);
425 :
426 2 : PG_TRY();
427 : {
428 2 : workspace->doctree = xmlParseMemory((char *) VARDATA_ANY(document),
429 : docsize);
430 2 : if (workspace->doctree != NULL)
431 : {
432 0 : workspace->ctxt = xmlXPathNewContext(workspace->doctree);
433 0 : workspace->ctxt->node = xmlDocGetRootElement(workspace->doctree);
434 :
435 : /* compile the path */
436 0 : comppath = xmlXPathCompile(xpath);
437 0 : if (comppath == NULL)
438 0 : xml_ereport(xmlerrcxt, ERROR, ERRCODE_EXTERNAL_ROUTINE_EXCEPTION,
439 : "XPath Syntax Error");
440 :
441 : /* Now evaluate the path expression. */
442 0 : workspace->res = xmlXPathCompiledEval(comppath, workspace->ctxt);
443 :
444 0 : xmlXPathFreeCompExpr(comppath);
445 : }
446 : }
447 0 : PG_CATCH();
448 : {
449 0 : cleanup_workspace(workspace);
450 :
451 0 : pg_xml_done(xmlerrcxt, true);
452 :
453 0 : PG_RE_THROW();
454 : }
455 2 : PG_END_TRY();
456 :
457 2 : if (workspace->res == NULL)
458 2 : cleanup_workspace(workspace);
459 :
460 2 : pg_xml_done(xmlerrcxt, false);
461 :
462 2 : return workspace->res;
463 : }
464 :
465 : /* Clean up after processing the result of pgxml_xpath() */
466 : static void
467 4 : cleanup_workspace(xpath_workspace *workspace)
468 : {
469 4 : if (workspace->res)
470 0 : xmlXPathFreeObject(workspace->res);
471 4 : workspace->res = NULL;
472 4 : if (workspace->ctxt)
473 0 : xmlXPathFreeContext(workspace->ctxt);
474 4 : workspace->ctxt = NULL;
475 4 : if (workspace->doctree)
476 0 : xmlFreeDoc(workspace->doctree);
477 4 : workspace->doctree = NULL;
478 4 : }
479 :
480 : static text *
481 2 : pgxml_result_to_text(xmlXPathObjectPtr res,
482 : xmlChar *toptag,
483 : xmlChar *septag,
484 : xmlChar *plainsep)
485 : {
486 : xmlChar *xpresstr;
487 : text *xpres;
488 :
489 2 : if (res == NULL)
490 2 : return NULL;
491 :
492 0 : switch (res->type)
493 : {
494 : case XPATH_NODESET:
495 0 : xpresstr = pgxmlNodeSetToText(res->nodesetval,
496 : toptag,
497 : septag, plainsep);
498 0 : break;
499 :
500 : case XPATH_STRING:
501 0 : xpresstr = xmlStrdup(res->stringval);
502 0 : break;
503 :
504 : default:
505 0 : elog(NOTICE, "unsupported XQuery result: %d", res->type);
506 0 : xpresstr = xmlStrdup((const xmlChar *) "<unsupported/>");
507 : }
508 :
509 : /* Now convert this result back to text */
510 0 : xpres = cstring_to_text((char *) xpresstr);
511 :
512 : /* Free various storage */
513 0 : xmlFree(xpresstr);
514 :
515 0 : return xpres;
516 : }
517 :
518 : /*
519 : * xpath_table is a table function. It needs some tidying (as do the
520 : * other functions here!
521 : */
522 4 : PG_FUNCTION_INFO_V1(xpath_table);
523 :
524 : Datum
525 10 : xpath_table(PG_FUNCTION_ARGS)
526 : {
527 : /* Function parameters */
528 10 : char *pkeyfield = text_to_cstring(PG_GETARG_TEXT_PP(0));
529 10 : char *xmlfield = text_to_cstring(PG_GETARG_TEXT_PP(1));
530 10 : char *relname = text_to_cstring(PG_GETARG_TEXT_PP(2));
531 10 : char *xpathset = text_to_cstring(PG_GETARG_TEXT_PP(3));
532 10 : char *condition = text_to_cstring(PG_GETARG_TEXT_PP(4));
533 :
534 : /* SPI (input tuple) support */
535 : SPITupleTable *tuptable;
536 : HeapTuple spi_tuple;
537 : TupleDesc spi_tupdesc;
538 :
539 : /* Output tuple (tuplestore) support */
540 10 : Tuplestorestate *tupstore = NULL;
541 : TupleDesc ret_tupdesc;
542 : HeapTuple ret_tuple;
543 :
544 10 : ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
545 : AttInMetadata *attinmeta;
546 : MemoryContext per_query_ctx;
547 : MemoryContext oldcontext;
548 :
549 : char **values;
550 : xmlChar **xpaths;
551 : char *pos;
552 10 : const char *pathsep = "|";
553 :
554 : int numpaths;
555 : int ret;
556 : uint64 proc;
557 : int j;
558 : int rownr; /* For issuing multiple rows from one original
559 : * document */
560 : bool had_values; /* To determine end of nodeset results */
561 : StringInfoData query_buf;
562 : PgXmlErrorContext *xmlerrcxt;
563 10 : volatile xmlDocPtr doctree = NULL;
564 :
565 : /* We only have a valid tuple description in table function mode */
566 10 : if (rsinfo == NULL || !IsA(rsinfo, ReturnSetInfo))
567 0 : ereport(ERROR,
568 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
569 : errmsg("set-valued function called in context that cannot accept a set")));
570 10 : if (rsinfo->expectedDesc == NULL)
571 0 : ereport(ERROR,
572 : (errcode(ERRCODE_SYNTAX_ERROR),
573 : errmsg("xpath_table must be called as a table function")));
574 :
575 : /*
576 : * We want to materialise because it means that we don't have to carry
577 : * libxml2 parser state between invocations of this function
578 : */
579 10 : if (!(rsinfo->allowedModes & SFRM_Materialize))
580 0 : ereport(ERROR,
581 : (errcode(ERRCODE_SYNTAX_ERROR),
582 : errmsg("xpath_table requires Materialize mode, but it is not "
583 : "allowed in this context")));
584 :
585 : /*
586 : * The tuplestore must exist in a higher context than this function call
587 : * (per_query_ctx is used)
588 : */
589 10 : per_query_ctx = rsinfo->econtext->ecxt_per_query_memory;
590 10 : oldcontext = MemoryContextSwitchTo(per_query_ctx);
591 :
592 : /*
593 : * Create the tuplestore - work_mem is the max in-memory size before a
594 : * file is created on disk to hold it.
595 : */
596 10 : tupstore =
597 10 : tuplestore_begin_heap(rsinfo->allowedModes & SFRM_Materialize_Random,
598 : false, work_mem);
599 :
600 10 : MemoryContextSwitchTo(oldcontext);
601 :
602 : /* get the requested return tuple description */
603 10 : ret_tupdesc = CreateTupleDescCopy(rsinfo->expectedDesc);
604 :
605 : /* must have at least one output column (for the pkey) */
606 10 : if (ret_tupdesc->natts < 1)
607 0 : ereport(ERROR,
608 : (errcode(ERRCODE_SYNTAX_ERROR),
609 : errmsg("xpath_table must have at least one output column")));
610 :
611 : /*
612 : * At the moment we assume that the returned attributes make sense for the
613 : * XPath specified (i.e. we trust the caller). It's not fatal if they get
614 : * it wrong - the input function for the column type will raise an error
615 : * if the path result can't be converted into the correct binary
616 : * representation.
617 : */
618 :
619 10 : attinmeta = TupleDescGetAttInMetadata(ret_tupdesc);
620 :
621 : /* Set return mode and allocate value space. */
622 10 : rsinfo->returnMode = SFRM_Materialize;
623 10 : rsinfo->setDesc = ret_tupdesc;
624 :
625 10 : values = (char **) palloc(ret_tupdesc->natts * sizeof(char *));
626 10 : xpaths = (xmlChar **) palloc(ret_tupdesc->natts * sizeof(xmlChar *));
627 :
628 : /*
629 : * Split XPaths. xpathset is a writable CString.
630 : *
631 : * Note that we stop splitting once we've done all needed for tupdesc
632 : */
633 10 : numpaths = 0;
634 10 : pos = xpathset;
635 24 : while (numpaths < (ret_tupdesc->natts - 1))
636 : {
637 10 : xpaths[numpaths++] = (xmlChar *) pos;
638 10 : pos = strstr(pos, pathsep);
639 10 : if (pos != NULL)
640 : {
641 4 : *pos = '\0';
642 4 : pos++;
643 : }
644 : else
645 6 : break;
646 : }
647 :
648 : /* Now build query */
649 10 : initStringInfo(&query_buf);
650 :
651 : /* Build initial sql statement */
652 10 : appendStringInfo(&query_buf, "SELECT %s, %s FROM %s WHERE %s",
653 : pkeyfield,
654 : xmlfield,
655 : relname,
656 : condition);
657 :
658 10 : if ((ret = SPI_connect()) < 0)
659 0 : elog(ERROR, "xpath_table: SPI_connect returned %d", ret);
660 :
661 10 : if ((ret = SPI_exec(query_buf.data, 0)) != SPI_OK_SELECT)
662 0 : elog(ERROR, "xpath_table: SPI execution failed for query %s",
663 : query_buf.data);
664 :
665 10 : proc = SPI_processed;
666 10 : tuptable = SPI_tuptable;
667 10 : spi_tupdesc = tuptable->tupdesc;
668 :
669 : /* Switch out of SPI context */
670 10 : MemoryContextSwitchTo(oldcontext);
671 :
672 : /*
673 : * Check that SPI returned correct result. If you put a comma into one of
674 : * the function parameters, this will catch it when the SPI query returns
675 : * e.g. 3 columns.
676 : */
677 10 : if (spi_tupdesc->natts != 2)
678 : {
679 0 : ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
680 : errmsg("expression returning multiple columns is not valid in parameter list"),
681 : errdetail("Expected two columns in SPI result, got %d.", spi_tupdesc->natts)));
682 : }
683 :
684 : /*
685 : * Setup the parser. This should happen after we are done evaluating the
686 : * query, in case it calls functions that set up libxml differently.
687 : */
688 10 : xmlerrcxt = pgxml_parser_init(PG_XML_STRICTNESS_LEGACY);
689 :
690 10 : PG_TRY();
691 : {
692 : /* For each row i.e. document returned from SPI */
693 : uint64 i;
694 :
695 20 : for (i = 0; i < proc; i++)
696 : {
697 : char *pkey;
698 : char *xmldoc;
699 : xmlXPathContextPtr ctxt;
700 : xmlXPathObjectPtr res;
701 : xmlChar *resstr;
702 : xmlXPathCompExprPtr comppath;
703 :
704 : /* Extract the row data as C Strings */
705 10 : spi_tuple = tuptable->vals[i];
706 10 : pkey = SPI_getvalue(spi_tuple, spi_tupdesc, 1);
707 10 : xmldoc = SPI_getvalue(spi_tuple, spi_tupdesc, 2);
708 :
709 : /*
710 : * Clear the values array, so that not-well-formed documents
711 : * return NULL in all columns. Note that this also means that
712 : * spare columns will be NULL.
713 : */
714 30 : for (j = 0; j < ret_tupdesc->natts; j++)
715 20 : values[j] = NULL;
716 :
717 : /* Insert primary key */
718 10 : values[0] = pkey;
719 :
720 : /* Parse the document */
721 10 : if (xmldoc)
722 10 : doctree = xmlParseMemory(xmldoc, strlen(xmldoc));
723 : else /* treat NULL as not well-formed */
724 0 : doctree = NULL;
725 :
726 10 : if (doctree == NULL)
727 : {
728 : /* not well-formed, so output all-NULL tuple */
729 0 : ret_tuple = BuildTupleFromCStrings(attinmeta, values);
730 0 : tuplestore_puttuple(tupstore, ret_tuple);
731 0 : heap_freetuple(ret_tuple);
732 : }
733 : else
734 : {
735 : /* New loop here - we have to deal with nodeset results */
736 10 : rownr = 0;
737 :
738 : do
739 : {
740 : /* Now evaluate the set of xpaths. */
741 16 : had_values = false;
742 36 : for (j = 0; j < numpaths; j++)
743 : {
744 20 : ctxt = xmlXPathNewContext(doctree);
745 20 : ctxt->node = xmlDocGetRootElement(doctree);
746 :
747 : /* compile the path */
748 20 : comppath = xmlXPathCompile(xpaths[j]);
749 20 : if (comppath == NULL)
750 0 : xml_ereport(xmlerrcxt, ERROR,
751 : ERRCODE_EXTERNAL_ROUTINE_EXCEPTION,
752 : "XPath Syntax Error");
753 :
754 : /* Now evaluate the path expression. */
755 20 : res = xmlXPathCompiledEval(comppath, ctxt);
756 20 : xmlXPathFreeCompExpr(comppath);
757 :
758 20 : if (res != NULL)
759 : {
760 20 : switch (res->type)
761 : {
762 : case XPATH_NODESET:
763 : /* We see if this nodeset has enough nodes */
764 40 : if (res->nodesetval != NULL &&
765 20 : rownr < res->nodesetval->nodeNr)
766 : {
767 8 : resstr = xmlXPathCastNodeToString(res->nodesetval->nodeTab[rownr]);
768 8 : had_values = true;
769 : }
770 : else
771 12 : resstr = NULL;
772 :
773 20 : break;
774 :
775 : case XPATH_STRING:
776 0 : resstr = xmlStrdup(res->stringval);
777 0 : break;
778 :
779 : default:
780 0 : elog(NOTICE, "unsupported XQuery result: %d", res->type);
781 0 : resstr = xmlStrdup((const xmlChar *) "<unsupported/>");
782 : }
783 :
784 : /*
785 : * Insert this into the appropriate column in the
786 : * result tuple.
787 : */
788 20 : values[j + 1] = (char *) resstr;
789 : }
790 20 : xmlXPathFreeContext(ctxt);
791 : }
792 :
793 : /* Now add the tuple to the output, if there is one. */
794 16 : if (had_values)
795 : {
796 6 : ret_tuple = BuildTupleFromCStrings(attinmeta, values);
797 6 : tuplestore_puttuple(tupstore, ret_tuple);
798 6 : heap_freetuple(ret_tuple);
799 : }
800 :
801 16 : rownr++;
802 16 : } while (had_values);
803 : }
804 :
805 10 : if (doctree != NULL)
806 10 : xmlFreeDoc(doctree);
807 10 : doctree = NULL;
808 :
809 10 : if (pkey)
810 10 : pfree(pkey);
811 10 : if (xmldoc)
812 10 : pfree(xmldoc);
813 : }
814 : }
815 0 : PG_CATCH();
816 : {
817 0 : if (doctree != NULL)
818 0 : xmlFreeDoc(doctree);
819 :
820 0 : pg_xml_done(xmlerrcxt, true);
821 :
822 0 : PG_RE_THROW();
823 : }
824 10 : PG_END_TRY();
825 :
826 10 : if (doctree != NULL)
827 0 : xmlFreeDoc(doctree);
828 :
829 10 : pg_xml_done(xmlerrcxt, false);
830 :
831 : tuplestore_donestoring(tupstore);
832 :
833 10 : SPI_finish();
834 :
835 10 : rsinfo->setResult = tupstore;
836 :
837 : /*
838 : * SFRM_Materialize mode expects us to return a NULL Datum. The actual
839 : * tuples are in our tuplestore and passed back through rsinfo->setResult.
840 : * rsinfo->setDesc is set to the tuple description that we actually used
841 : * to build our tuples with, so the caller can verify we did what it was
842 : * expecting.
843 : */
844 10 : return (Datum) 0;
845 : }
|
