1 /******************************************************************************
2 * Copyright (C) 2008 by Diego Arroyuelo *
3 * Interface for the in-memory XQuery/XPath engine *
5 * This program is free software; you can redistribute it and/or modify *
6 * it under the terms of the GNU Lesser General Public License as published *
7 * by the Free Software Foundation; either version 2 of the License, or *
8 * (at your option) any later version. *
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 Lesser General Public License for more details. *
15 * You should have received a copy of the GNU Lesser General Public License *
16 * along with this program; if not, write to the *
17 * Free Software Foundation, Inc., *
18 * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. *
19 ******************************************************************************/
25 #include <unordered_set>
26 #include <unordered_map>
28 #include <TextCollection/TextCollectionBuilder.h>
35 #include <libcds/includes/basics.h>
36 #include <libcds/includes/static_bitsequence.h>
37 #include <libcds/includes/alphabet_mapper.h>
38 #include <libcds/includes/static_sequence.h>
40 using SXSI::TextCollection;
41 using SXSI::TextCollectionBuilder;
44 // this constant is used to efficiently compute the child operation in the tree
49 #define PERM_SAMPLE 10
61 // Encoding of the XML Document :
62 // The following TAGs and IDs are fixed, "" is the tag of the root.
63 // a TextNode is represented by a leaf <<$>></<$>> The DocId in the TextCollection
64 // of that leaf is kept in a bit sequence.
65 // a TextNode below an attribute is likewise represented by a leaf <<@$>><</@$>>
66 // An element <e a1="v1" a2="v2" ... an="vn" > ...</e> the representation is:
67 // <e><<@>> <<@>a1> <<$@>>DocID(v1)</<$@>></<@>a1> ... </<@>> .... </e>
68 // Hence the attributes (if any) are always below the first child of their element,
69 // as the children of a fake node <@>.
72 #define DOCUMENT_OPEN_TAG ""
73 #define DOCUMENT_TAG_ID 0
74 #define ATTRIBUTE_OPEN_TAG "<@>"
75 #define ATTRIBUTE_TAG_ID 1
76 #define PCDATA_OPEN_TAG "<$>"
77 #define PCDATA_TAG_ID 2
78 #define ATTRIBUTE_DATA_OPEN_TAG "<@$>"
79 #define ATTRIBUTE_DATA_TAG_ID 3
80 #define CLOSING_TAG "</>"
81 #define CLOSING_TAG_ID 4
82 #define DOCUMENT_CLOSE_TAG "/"
83 #define ATTRIBUTE_CLOSE_TAG "/<@>"
84 #define PCDATA_CLOSE_TAG "/<$>"
85 #define ATTRIBUTE_DATA_CLOSE_TAG "/<@$>"
88 typedef std::unordered_set<int> TagIdSet;
89 typedef std::unordered_map<std::string,int> TagIdMap;
90 typedef TagIdMap::const_iterator TagIdMapIT;
92 #define REGISTER_TAG(v,h,t) do { (h)->insert(std::make_pair((t),(v)->size()));\
93 (v)->push_back(t); } while (false)
95 // returns NULLT if the test is true
96 #define NULLT_IF(x) do { if (x) return NULLT; } while (0)
98 // Direct calls to sarray library
100 #define BUFFER_ALLOC (8192 * 2)
101 #define BUFFER_SIZE (BUFFER_ALLOC / 2)
103 // tag position -> tree node
104 static treeNode tagpos2node(int t)
108 // tree node -> tag position
109 static int node2tagpos(treeNode x)
115 class XMLTreeBuilder;
119 // Only the builder can access the constructor
120 friend class XMLTreeBuilder;
123 /** Balanced parentheses representation of the tree */
126 /** Mapping from tag identifer to tag name */
127 std::vector<std::string> *TagName;
130 /** Bit vector indicating with a 1 the positions of the non-empty texts. */
131 static_bitsequence *EBVector;
133 /** Tag sequence represented with a data structure for rank and select */
134 static_sequence *Tags;
136 uint tags_blen, tags_len;
138 /** The texts in the XML document */
139 TextCollection *Text;
141 // Allows to disable the TextCollection for benchmarkin purposes
143 SXSI::TextCollectionBuilder::index_type_t text_index_type;
146 std::vector<std::string *> *print_stack;
149 void _real_flush(int fd, size_t size) {
150 if (size == 0) return;
153 written = write(fd, buffer->data(), size);
154 if ((written < 0) && (errno == EAGAIN || errno == EINTR))
163 size_t size = buffer->size();
164 if (size < BUFFER_SIZE) return;
165 _real_flush(fd, size);
168 void _dput_str(std::string s, int fd){
173 void _dputs(const char* s, int fd){
178 void _dputc(const char c, int fd){
179 buffer->push_back(c);
182 size_t _dprintf(const char* s, int fd){
183 if (s == NULL) return 0;
186 for (i = 0; (c = s[i]); i++) {
189 _dputs(""", fd);
195 _dputs("'", fd);
211 void PrintNode(treeNode n, int fd);
212 /** Data structure constructors */
213 XMLTree(){ buffer = 0;};
215 // non const pointer are freed by this method.
216 XMLTree( pb * const par,
218 std::vector<std::string> * const TN,
219 TagIdMap * const tim, uint *empty_texts_bmp,
221 TextCollection * const TC, bool dis_tc,
222 TextCollectionBuilder::index_type_t _index_type );
225 /** Data structure destructor */
228 /** root(): returns the tree root. */
229 treeNode Root() { return 0; }
231 /** Size() : Number of parenthesis */
237 /** NumTags() : Number of distinct tags */
238 unsigned int NumTags() {
239 return TagName->size();
242 int TagsBinaryLength(){ return tags_blen; };
243 unsigned int TagStructLength(){ return uint_len(tags_blen,tags_len); };
244 unsigned int * TagStruct() { return tags_fix; };
247 /** SubtreeSize(x): the number of nodes (and attributes) in the subtree of
249 int SubtreeSize(treeNode x) { return bp_subtree_size(Par, x); }
251 /** SubtreeTags(x,tag): the number of occurrences of tag within the subtree
253 int SubtreeTags(treeNode x, TagType tag){
254 //int s = x + 2*subtree_size(Par, x) - 1;
255 treeNode y = bp_find_close(Par, x);
259 for(int i = x; i < y; i++)
260 count += (Tag(i) == tag);
264 return (Tags->rank(tag, y) - Tags->rank(tag, x));
267 /** SubtreeElements(x) of element nodes in the subtree of x
269 int SubtreeElements(treeNode x);
271 /** IsLeaf(x): returns whether node x is leaf or not. In the succinct
272 * representation this is just a bit inspection. */
274 bool IsLeaf(treeNode x);
276 /** IsAncestor(x,y): returns whether node x is ancestor of node y. */
278 bool IsAncestor(treeNode x, treeNode y);
281 /** IsRigthDescendant returns true if y is a descendant of x and y is
282 not a descendant of the first child of x */
283 bool IsRightDescendant(treeNode x, treeNode y) {
284 if (x <= Root()) return false;
285 treeNode z = bp_parent_close(Par, x);
286 treeNode c = bp_find_close(Par, x);
287 return (y > c && y < z );
291 /** IsChild(x,y): returns whether node x is parent of node y. */
292 bool IsChild(treeNode x, treeNode y);
294 /** IsFirstChild(x): returns whether node x is the first child of its parent. */
296 bool IsFirstChild(treeNode x) {
297 return ((x != NULLT)&&(x==Root() || bp_prev_sibling(Par,x) == (treeNode)-1));
300 /** NumChildren(x): number of children of node x. Constant time with the
301 * data structure of Sadakane. */
302 int NumChildren(treeNode x);
304 /** ChildNumber(x): returns i if node x is the i-th children of its
306 int ChildNumber(treeNode x);
308 /** Depth(x): depth of node x, a simple binary rank on the parentheses
310 int Depth(treeNode x);
312 /** Preorder(x): returns the preorder number of node x, just regarding tree
313 * nodes (and not texts). */
314 int Preorder(treeNode x);
316 /** Postorder(x): returns the postorder number of node x, just regarding
317 * tree nodes (and not texts). */
318 int Postorder(treeNode x);
321 /** DocIds(x): returns the range (i.e., a pair of integers) of document
322 * identifiers that descend from node x. */
323 range DocIds(treeNode x);
325 /** Parent(x): returns the parent node of node x. */
326 treeNode Parent(treeNode x) {
327 return (x == Root()) ? NULLT : bp_parent(Par, x);
330 treeNode BinaryParent(treeNode x){
334 treeNode prev = x - 1;
335 return (bp_inspect(Par, prev) == OP) ? prev : bp_find_open(Par, prev);
339 /* Assumes x is neither 0 nor -1 */
341 /** Child(x,i): returns the i-th child of node x, assuming it exists. */
342 treeNode Child(treeNode x, int i);
346 /** LastChild(x): returns the last child of node x. */
347 treeNode LastChild(treeNode x) {
348 NULLT_IF(x == NULLT || bp_isleaf(Par,x));
349 return bp_find_open(Par, bp_find_close(Par, x)-1);
352 /** PrevSibling(x): returns the previous sibling of node x, assuming it
355 treeNode PrevSibling(treeNode x)
358 return bp_prev_sibling(Par, x);
362 /** TaggedChild(x,tag): returns the first child of node x tagged tag, or
363 * NULLT if there is none. Because of the balanced-parentheses representation
364 * of the tree, this operation is not supported efficiently, just iterating
365 * among the children of node x until finding the desired child. */
368 treeNode SelectChild(treeNode x, TagIdSet * tags);
370 /** TaggedFollowingSibling(x,tag): returns the first sibling of node x tagged tag, or
371 * NULLT if there is none. */
373 treeNode SelectFollowingSibling(treeNode x, TagIdSet * tags);
378 treeNode SelectDescendant(treeNode x, TagIdSet * tags) {
379 NULLT_IF (x == NULLT);
381 if (bp_inspect(Par, fc) == CP) return NULLT;
382 treeNode min = NULLT;
385 TagIdSet::const_iterator tagit;
386 for (tagit = tags->begin(); tagit != tags->end(); ++tagit) {
387 aux = TaggedDescendant(x, (TagType) *tagit);
388 if (((unsigned int) aux) < ((unsigned int) min)) min = aux;
393 /** TaggedPrec(x,tag): returns the first node tagged tag with smaller
394 * preorder than x and not an ancestor of x. Returns NULLT if there
396 treeNode TaggedPreceding(treeNode x, TagType tag);
398 /** TaggedFoll(x,tag): returns the first node tagged tag with larger
399 * preorder than x and not in the subtree of x. Returns NULLT if there
401 treeNode TaggedFollowing(treeNode x, TagType tag);
405 treeNode SelectFollowingBelow(treeNode x, TagIdSet * tags, treeNode ancestor);
407 // treeNode TaggedFollowingBefore(treeNode x, TagType tag,treeNode closing);
409 treeNode SelectFollowingBefore(treeNode x, TagIdSet * tags, treeNode ancestor_closing)
414 treeNode close = bp_find_close(Par,x);
417 treeNode min = NULLT;
421 TagIdSet::const_iterator tagit;
422 for (tagit = tags->begin(); tagit != tags->end(); ++tagit) {
424 aux = tagpos2node(Tags->select_next(*tagit, close));
426 if (((unsigned int) aux) < ((unsigned int) min)) min = aux;
430 return (min < ancestor_closing) ? min : NULLT;
434 /** TaggedAncestor(x, tag): returns the closest ancestor of x tagged
435 * tag. Return NULLT is there is none. */
436 treeNode TaggedAncestor(treeNode x, TagType tag);
438 /** PrevText(x): returns the document identifier of the text to the left of
439 * node x, or NULLT if x is the root node. */
440 DocID PrevText(treeNode x);
442 /** NextText(x): returns the document identifier of the text to the right of
443 * node x, or NULLT if x is the root node. */
444 DocID NextText(treeNode x);
446 /** MyText(x): returns the document identifier of the text below node x, or
447 * NULLT if x is not a leaf node. */
448 DocID MyText(treeNode x);
449 DocID MyTextUnsafe(treeNode x);
451 /** TextXMLId(d): returns the preorder of document with identifier d in the
452 * tree consisting of all tree nodes and all text nodes. */
453 int TextXMLId(DocID d);
455 /** NodeXMLId(x): returns the preorder of node x in the tree consisting of
456 * all tree nodes and all text nodes. */
457 int NodeXMLId(treeNode x);
459 /** ParentNode(d): returns the parent node of document identifier d. */
460 treeNode ParentNode(DocID d);
462 treeNode PrevNode(DocID d);
464 /** GetTagId(tagname): returns the tag identifier corresponding to a given
465 * tag name. Returns NULLT in case that the tag name does not exists. */
466 TagType GetTagId(unsigned char *tagname);
468 /** GetTagName(tagid): returns the tag name of a given tag identifier.
469 * Returns NULL in case that the tag identifier is not valid.*/
470 unsigned char *GetTagName(TagType tagid);
472 /** GetTagName(tagid): returns the tag name of a given tag identifier.
473 * The result is just a reference and should not be freed by the caller.
475 const unsigned char *GetTagNameByRef(TagType tagid);
477 /** RegisterTag adds a new tag to the tag collection this is needed
478 * if the query contains a tag which is not in the document, we need
479 * to give this new tag a fresh id and store it somewhere. A logical
481 * We might want to use a hashtable instead of an array though.
483 TagType RegisterTag(unsigned char *tagname);
485 bool EmptyText(DocID i) {
486 return Text->EmptyText(i);
489 /** Prefix(s): search for texts prefixed by string s. */
490 TextCollection::document_result Prefix(uchar const *s) {
491 return Text->Prefix(s);
494 /** Suffix(s): search for texts having string s as a suffix. */
495 TextCollection::document_result Suffix(uchar const *s) {
496 return Text->Suffix(s);
499 /** Equal(s): search for texts equal to string s. */
500 TextCollection::document_result Equals(uchar const *s) {
501 return Text->Equal(s);
504 /** Contains(s): search for texts containing string s. */
505 TextCollection::document_result Contains(uchar const *s) {
506 return Text->Contains(s);
509 /** LessThan(s): returns document identifiers for the texts that
510 * are lexicographically smaller than string s. */
511 TextCollection::document_result LessThan(uchar const *s) {
512 return Text->LessThan(s);
515 /** IsPrefix(x): returns true if there is a text prefixed by string s. */
516 bool IsPrefix(uchar const *s) {
517 return Text->IsPrefix(s);
520 /** IsSuffix(s): returns true if there is a text having string s as a
522 bool IsSuffix(uchar const *s) {
523 return Text->IsSuffix(s);
526 /** IsEqual(s): returns true if there is a text that equals given
528 bool IsEqual(uchar const *s) {
529 return Text->IsEqual(s);
532 /** IsContains(s): returns true if there is a text containing string s. */
533 bool IsContains(uchar const *s) {
534 return Text->IsContains(s);
537 /** IsLessThan(s): returns true if there is at least a text that is
538 * lexicographically smaller than string s. */
539 bool IsLessThan(uchar const *s) {
540 return Text->IsLessThan(s);
543 /** Count(s): Global counting */
544 unsigned Count(uchar const *s) {
545 return Text->Count(s);
548 /** CountPrefix(s): counting version of Prefix(s). */
549 unsigned CountPrefix(uchar const *s) {
550 return Text->CountPrefix(s);
553 /** CountSuffix(s): counting version of Suffix(s). */
554 unsigned CountSuffix(uchar const *s) {
555 return Text->CountSuffix(s);
558 /** CountEqual(s): counting version of Equal(s). */
559 unsigned CountEqual(uchar const *s) {
560 return Text->CountEqual(s);
563 /** CountContains(s): counting version of Contains(s). */
564 unsigned CountContains(uchar const *s) {
565 return Text->CountContains(s);
568 /** CountLessThan(s): counting version of LessThan(s). */
569 unsigned CountLessThan(uchar const *s) {
570 return Text->CountLessThan(s);
573 /** GetText(d): returns the text corresponding to document with
575 uchar* GetText(DocID d) {
577 uchar * s = Text->GetText(d);
578 return (s[0] == 1 ? (s+1) : s);
581 /** GetText(i, j): returns the texts corresponding to documents with
582 * ids i, i+1, ..., j. Texts are separated by '\0' character. */
583 // uchar* GetText(DocID i, DocID j) {
584 // uchar * s = Text->GetText(i, j);
585 // return (s[0] == 1 ? (uchar*)"" : s);
588 TextCollection *getTextCollection() {
592 /** Save: saves XML tree data structure to file. */
593 void Save(int fd, char* name );
595 /** Load: loads XML tree data structure from file. sample_rate_text
596 * indicates the sample rate for the text search data structure. */
597 static XMLTree *Load(int fd, bool load_tc, int sample_factor, char * name);
599 void insertTag(TagType tag, uint position);
604 /** Parenthesis functions */
605 treeNode Closing(treeNode x);
607 bool IsOpen(treeNode x);
610 /** Print procedure */
611 void Print(int fd,treeNode x, bool no_text);
612 void Print(int fd,treeNode x) { Print(fd,x,false); }
613 void Flush(int fd){ if (buffer) _real_flush(fd, buffer->size()); }
615 // The following are inlined here for speed
616 /** Tag(x): returns the tag identifier of node x. */
618 inline TagType Tag(treeNode x) const throw () {
620 return (TagType) (((uchar*)tags_fix)[(int) x]);
622 return get_field(tags_fix, tags_blen, x);
625 size_t idxlen = x * tags_blen;
626 size_t j = idxlen % W;
627 size_t i = idxlen / W;
628 size_t offset = W - tags_blen;
629 size_t offset2 = offset - j;
630 size_t w = tags_fix[i];
631 return (offset2 >= 0)
632 ? ( w << offset2 ) >> offset
633 : ( w >> j) | (tags_fix[i+1] << (W+offset2)) >> offset;
638 /** FirstChild(x): returns the first child of node x, or NULLT if the node is a leaf
640 treeNode FirstChild(treeNode x) {
642 return bp_first_child(Par, x);
646 /** FirstElement(x): returns the first non text, non attribute child of node x, or NULLT
649 treeNode FirstElement(treeNode node){
651 NULLT_IF(node == NULLT);
652 treeNode x = bp_first_child(Par, node);
653 NULLT_IF(x == NULLT);
656 case ATTRIBUTE_TAG_ID:
657 x = bp_next_sibling(Par,x);
658 if (x == NULLT || Tag(x) != PCDATA_TAG_ID) return x;
662 return (bp_inspect(Par,x)==OP)? x : NULLT;
670 /** NextSibling(x): returns the next sibling of node x, or NULLT if none
673 treeNode NextSibling(treeNode x) {
675 return bp_next_sibling(Par, x);
678 /** NextElement(x): returns the first non text, non attribute sibling of node x, or NULLT
681 treeNode NextElement(treeNode x)
684 x = bp_next_sibling(Par, x);
685 NULLT_IF(x == NULLT);
686 if (Tag(x) == PCDATA_TAG_ID){
688 return (bp_inspect(Par, y) == OP) ? y : NULLT;
692 /** TaggedDesc(x,tag): returns the first node tagged tag with larger
693 * preorder than x and within the subtree of x. Returns NULT if there
695 inline treeNode TaggedNext(treeNode x, TagType tag)
697 return tagpos2node(Tags->select_next(tag,node2tagpos(x)));
699 inline treeNode TaggedDescendant(treeNode x, TagType tag)
702 int s = (int) Tags->select_next(tag,node2tagpos(x));
705 treeNode y = tagpos2node(s); // transforms the tag position into a node position
707 return (bp_is_ancestor(Par,x,y) ? y : NULLT);
710 inline treeNode TaggedFollowingBelow(treeNode x, TagType tag, treeNode ancestor)
712 treeNode close = bp_find_close(Par, x);
713 treeNode s = tagpos2node(Tags->select_next(tag, close));
715 if (ancestor == Root() || s == NULLT || s < bp_find_close(Par,ancestor)) return s;
719 inline treeNode TaggedFollowingBefore(treeNode x, TagType tag, treeNode ancestor_closing)
721 treeNode close = bp_find_close(Par, x);
722 treeNode s = tagpos2node(Tags->select_next(tag, close));
724 if (ancestor_closing == Root() || s == NULLT || s < ancestor_closing) return s;
728 inline treeNode NextNodeBefore(treeNode x, treeNode ancestor_closing)
730 treeNode close = bp_find_close(Par, x);
731 int rank = bp_rank_open(Par, close);
732 treeNode y = bp_select_open(Par, rank+1);
733 return (y < ancestor_closing) ? y : NULLT;
736 // TaggedSibling(x,tag): returns the first sibling of node x tagged tag, or NULLT if there is none.
737 treeNode TaggedFollowingSibling(treeNode x, TagType tag)
740 treeNode sibling = x;
742 while ((sibling = bp_next_sibling(Par, sibling)) != NULLT) {
744 if (ctag == tag) return sibling;
749 treeNode TaggedChild(treeNode x, TagType tag)
752 NULLT_IF(x==NULLT || bp_isleaf(Par,x));
754 child = bp_first_child(Par, x);
756 if (Tag(child) == tag)
759 return TaggedFollowingSibling(child, tag);