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 <static_bitsequence.h>
37 #include <alphabet_mapper.h>
38 #include <static_sequence.h>
39 using SXSI::TextCollection;
40 using SXSI::TextCollectionBuilder;
43 // this constant is used to efficiently compute the child operation in the tree
48 #define PERM_SAMPLE 10
60 // Encoding of the XML Document :
61 // The following TAGs and IDs are fixed, "" is the tag of the root.
62 // a TextNode is represented by a leaf <<$>></<$>> The DocId in the TextCollection
63 // of that leaf is kept in a bit sequence.
64 // a TextNode below an attribute is likewise represented by a leaf <<@$>><</@$>>
65 // An element <e a1="v1" a2="v2" ... an="vn" > ...</e> the representation is:
66 // <e><<@>> <<@>a1> <<$@>>DocID(v1)</<$@>></<@>a1> ... </<@>> .... </e>
67 // Hence the attributes (if any) are always below the first child of their element,
68 // as the children of a fake node <@>.
71 #define DOCUMENT_OPEN_TAG ""
72 #define DOCUMENT_TAG_ID 0
73 #define ATTRIBUTE_OPEN_TAG "<@>"
74 #define ATTRIBUTE_TAG_ID 1
75 #define PCDATA_OPEN_TAG "<$>"
76 #define PCDATA_TAG_ID 2
77 #define ATTRIBUTE_DATA_OPEN_TAG "<@$>"
78 #define ATTRIBUTE_DATA_TAG_ID 3
79 #define CLOSING_TAG "</>"
80 #define CLOSING_TAG_ID 4
81 #define DOCUMENT_CLOSE_TAG "/"
82 #define ATTRIBUTE_CLOSE_TAG "/<@>"
83 #define PCDATA_CLOSE_TAG "/<$>"
84 #define ATTRIBUTE_DATA_CLOSE_TAG "/<@$>"
87 typedef std::unordered_set<int> TagIdSet;
88 typedef std::unordered_map<std::string,int> TagIdMap;
89 typedef TagIdMap::const_iterator TagIdMapIT;
91 #define REGISTER_TAG(v,h,t) do { (h)->insert(std::make_pair((t),(v)->size()));\
92 (v)->push_back(t); } while (false)
94 // returns NULLT if the test is true
95 #define NULLT_IF(x) do { if (x) return NULLT; } while (0)
97 // Direct calls to sarray library
99 #define BUFFER_ALLOC (8192 * 2)
100 #define BUFFER_SIZE (BUFFER_ALLOC / 2)
101 static inline int fast_find_open(bp *b,int s)
104 r = bwd_excess(b,s,0);
105 if (r >= -1) return r+1;
109 static inline int fast_find_close(bp *b,int s)
111 return fwd_excess(b,s,-1);
114 static inline int fast_find_parent_close(bp *b,int s)
116 return fwd_excess(b,s,-2);
120 static inline int fast_inspect(bp* Par,treeNode i)
125 return (Par->B[j] >> (D-1-l)) & 1;
128 static bool fast_isleaf(bp* Par,treeNode x){
129 return (fast_inspect(Par, x+1) == CP);
132 static treeNode fast_first_child(bp *Par, treeNode x)
135 return (fast_inspect(Par,x) == OP) ? x : NULLT;
138 inline static treeNode fast_next_sibling(bp* Par,treeNode x)
140 treeNode y = fast_find_close(Par,x)+1;
141 return (fast_inspect(Par, y) == OP) ? y : NULLT;
144 inline static bool fast_is_ancestor(bp * Par,treeNode x,treeNode y){
145 return (x <= y) && ((x==0) || (y <= fast_find_close(Par,x)));
148 // tag position -> tree node
149 static treeNode tagpos2node(int t)
153 // tree node -> tag position
154 static int node2tagpos(treeNode x)
160 class XMLTreeBuilder;
164 // Only the builder can access the constructor
165 friend class XMLTreeBuilder;
168 /** Balanced parentheses representation of the tree */
171 /** Mapping from tag identifer to tag name */
172 std::vector<std::string> *TagName;
175 /** Bit vector indicating with a 1 the positions of the non-empty texts. */
176 static_bitsequence *EBVector;
178 /** Tag sequence represented with a data structure for rank and select */
179 static_sequence *Tags;
181 uint tags_blen, tags_len;
183 /** The texts in the XML document */
184 TextCollection *Text;
186 // Allows to disable the TextCollection for benchmarkin purposes
188 SXSI::TextCollectionBuilder::index_type_t text_index_type;
191 std::vector<std::string *> *print_stack;
194 void _real_flush(int fd, size_t size) {
195 if (size == 0) return;
198 written = write(fd, buffer->data(), size);
199 if ((written < 0) && (errno == EAGAIN || errno == EINTR))
208 size_t size = buffer->size();
209 if (size < BUFFER_SIZE) return;
210 _real_flush(fd, size);
213 void _dput_str(std::string s, int fd){
218 void _dputs(const char* s, int fd){
223 void _dputc(const char c, int fd){
224 buffer->push_back(c);
227 size_t _dprintf(const char* s, int fd){
228 if (s == NULL) return 0;
231 for (i = 0; (c = s[i]); i++) {
234 _dputs(""", fd);
240 _dputs("'", fd);
256 void PrintNode(treeNode n, int fd);
257 /** Data structure constructors */
258 XMLTree(){ buffer = 0;};
260 // non const pointer are freed by this method.
261 XMLTree( pb * const par,
263 std::vector<std::string> * const TN,
264 TagIdMap * const tim, uint *empty_texts_bmp,
266 TextCollection * const TC, bool dis_tc,
267 TextCollectionBuilder::index_type_t _index_type );
270 /** Data structure destructor */
273 /** root(): returns the tree root. */
274 treeNode Root() { return 0; }
276 /** Size() : Number of parenthesis */
282 /** NumTags() : Number of distinct tags */
283 unsigned int NumTags() {
284 return TagName->size();
287 int TagsBinaryLength(){ return tags_blen; };
288 unsigned int TagStructLength(){ return uint_len(tags_blen,tags_len); };
289 unsigned int * TagStruct() { return tags_fix; };
292 /** SubtreeSize(x): the number of nodes (and attributes) in the subtree of
294 int SubtreeSize(treeNode x) { return subtree_size(Par, x); }
296 /** SubtreeTags(x,tag): the number of occurrences of tag within the subtree
298 int SubtreeTags(treeNode x, TagType tag){
299 //int s = x + 2*subtree_size(Par, x) - 1;
300 treeNode y = fast_find_close(Par, x);
304 for(int i = x; i < y; i++)
305 count += (Tag(i) == tag);
309 return (Tags->rank(tag, y) - Tags->rank(tag, x));
312 /** SubtreeElements(x) of element nodes in the subtree of x
314 int SubtreeElements(treeNode x);
316 /** IsLeaf(x): returns whether node x is leaf or not. In the succinct
317 * representation this is just a bit inspection. */
319 bool IsLeaf(treeNode x);
321 /** IsAncestor(x,y): returns whether node x is ancestor of node y. */
323 bool IsAncestor(treeNode x, treeNode y);
326 /** IsRigthDescendant returns true if y is a descendant of x and y is
327 not a descendant of the first child of x */
328 bool IsRightDescendant(treeNode x, treeNode y) {
329 if (x <= Root()) return false;
330 treeNode z = fast_find_parent_close(Par, x);
331 treeNode c = fast_find_close(Par, x);
332 return (y > c && y < z );
336 /** IsChild(x,y): returns whether node x is parent of node y. */
337 bool IsChild(treeNode x, treeNode y);
339 /** IsFirstChild(x): returns whether node x is the first child of its parent. */
341 bool IsFirstChild(treeNode x) {
342 return ((x != NULLT)&&(x==Root() || prev_sibling(Par,x) == (treeNode)-1));
345 /** NumChildren(x): number of children of node x. Constant time with the
346 * data structure of Sadakane. */
347 int NumChildren(treeNode x);
349 /** ChildNumber(x): returns i if node x is the i-th children of its
351 int ChildNumber(treeNode x);
353 /** Depth(x): depth of node x, a simple binary rank on the parentheses
355 int Depth(treeNode x);
357 /** Preorder(x): returns the preorder number of node x, just regarding tree
358 * nodes (and not texts). */
359 int Preorder(treeNode x);
361 /** Postorder(x): returns the postorder number of node x, just regarding
362 * tree nodes (and not texts). */
363 int Postorder(treeNode x);
366 /** DocIds(x): returns the range (i.e., a pair of integers) of document
367 * identifiers that descend from node x. */
368 range DocIds(treeNode x);
370 /** Parent(x): returns the parent node of node x. */
371 treeNode Parent(treeNode x) {
375 return parent(Par, x);
378 treeNode BinaryParent(treeNode x){
382 treeNode prev = x - 1;
383 return (fast_inspect(Par, prev) == OP) ? prev : fast_find_open(Par, prev);
387 /* Assumes x is neither 0 nor -1 */
389 /** Child(x,i): returns the i-th child of node x, assuming it exists. */
390 treeNode Child(treeNode x, int i);
394 /** LastChild(x): returns the last child of node x. */
395 treeNode LastChild(treeNode x) {
396 NULLT_IF(x == NULLT || fast_isleaf(Par,x));
397 return fast_find_open(Par, fast_find_close(Par, x)-1);
400 /** PrevSibling(x): returns the previous sibling of node x, assuming it
403 treeNode PrevSibling(treeNode x)
406 return prev_sibling(Par, x);
410 /** TaggedChild(x,tag): returns the first child of node x tagged tag, or
411 * NULLT if there is none. Because of the balanced-parentheses representation
412 * of the tree, this operation is not supported efficiently, just iterating
413 * among the children of node x until finding the desired child. */
416 treeNode SelectChild(treeNode x, TagIdSet * tags);
418 /** TaggedFollowingSibling(x,tag): returns the first sibling of node x tagged tag, or
419 * NULLT if there is none. */
421 treeNode SelectFollowingSibling(treeNode x, TagIdSet * tags);
426 treeNode SelectDescendant(treeNode x, TagIdSet * tags) {
427 NULLT_IF (x == NULLT);
429 if (fast_inspect(Par, fc) == CP) return NULLT;
430 treeNode min = NULLT;
433 TagIdSet::const_iterator tagit;
434 for (tagit = tags->begin(); tagit != tags->end(); ++tagit) {
435 aux = TaggedDescendant(x, (TagType) *tagit);
436 if (((unsigned int) aux) < ((unsigned int) min)) min = aux;
441 /** TaggedPrec(x,tag): returns the first node tagged tag with smaller
442 * preorder than x and not an ancestor of x. Returns NULLT if there
444 treeNode TaggedPreceding(treeNode x, TagType tag);
446 /** TaggedFoll(x,tag): returns the first node tagged tag with larger
447 * preorder than x and not in the subtree of x. Returns NULLT if there
449 treeNode TaggedFollowing(treeNode x, TagType tag);
453 treeNode SelectFollowingBelow(treeNode x, TagIdSet * tags, treeNode ancestor);
455 // treeNode TaggedFollowingBefore(treeNode x, TagType tag,treeNode closing);
457 treeNode SelectFollowingBefore(treeNode x, TagIdSet * tags, treeNode ancestor_closing)
462 treeNode close = fast_find_close(Par,x);
465 treeNode min = NULLT;
469 TagIdSet::const_iterator tagit;
470 for (tagit = tags->begin(); tagit != tags->end(); ++tagit) {
472 aux = tagpos2node(Tags->select_next(*tagit, close));
474 if (((unsigned int) aux) < ((unsigned int) min)) min = aux;
478 return (min < ancestor_closing) ? min : NULLT;
482 /** TaggedAncestor(x, tag): returns the closest ancestor of x tagged
483 * tag. Return NULLT is there is none. */
484 treeNode TaggedAncestor(treeNode x, TagType tag);
486 /** PrevText(x): returns the document identifier of the text to the left of
487 * node x, or NULLT if x is the root node. */
488 DocID PrevText(treeNode x);
490 /** NextText(x): returns the document identifier of the text to the right of
491 * node x, or NULLT if x is the root node. */
492 DocID NextText(treeNode x);
494 /** MyText(x): returns the document identifier of the text below node x, or
495 * NULLT if x is not a leaf node. */
496 DocID MyText(treeNode x);
497 DocID MyTextUnsafe(treeNode x);
499 /** TextXMLId(d): returns the preorder of document with identifier d in the
500 * tree consisting of all tree nodes and all text nodes. */
501 int TextXMLId(DocID d);
503 /** NodeXMLId(x): returns the preorder of node x in the tree consisting of
504 * all tree nodes and all text nodes. */
505 int NodeXMLId(treeNode x);
507 /** ParentNode(d): returns the parent node of document identifier d. */
508 treeNode ParentNode(DocID d);
510 treeNode PrevNode(DocID d);
512 /** GetTagId(tagname): returns the tag identifier corresponding to a given
513 * tag name. Returns NULLT in case that the tag name does not exists. */
514 TagType GetTagId(unsigned char *tagname);
516 /** GetTagName(tagid): returns the tag name of a given tag identifier.
517 * Returns NULL in case that the tag identifier is not valid.*/
518 unsigned char *GetTagName(TagType tagid);
520 /** GetTagName(tagid): returns the tag name of a given tag identifier.
521 * The result is just a reference and should not be freed by the caller.
523 const unsigned char *GetTagNameByRef(TagType tagid);
525 /** RegisterTag adds a new tag to the tag collection this is needed
526 * if the query contains a tag which is not in the document, we need
527 * to give this new tag a fresh id and store it somewhere. A logical
529 * We might want to use a hashtable instead of an array though.
531 TagType RegisterTag(unsigned char *tagname);
533 bool EmptyText(DocID i) {
534 return Text->EmptyText(i);
537 /** Prefix(s): search for texts prefixed by string s. */
538 TextCollection::document_result Prefix(uchar const *s) {
539 return Text->Prefix(s);
542 /** Suffix(s): search for texts having string s as a suffix. */
543 TextCollection::document_result Suffix(uchar const *s) {
544 return Text->Suffix(s);
547 /** Equal(s): search for texts equal to string s. */
548 TextCollection::document_result Equals(uchar const *s) {
549 return Text->Equal(s);
552 /** Contains(s): search for texts containing string s. */
553 TextCollection::document_result Contains(uchar const *s) {
554 return Text->Contains(s);
557 /** LessThan(s): returns document identifiers for the texts that
558 * are lexicographically smaller than string s. */
559 TextCollection::document_result LessThan(uchar const *s) {
560 return Text->LessThan(s);
563 /** IsPrefix(x): returns true if there is a text prefixed by string s. */
564 bool IsPrefix(uchar const *s) {
565 return Text->IsPrefix(s);
568 /** IsSuffix(s): returns true if there is a text having string s as a
570 bool IsSuffix(uchar const *s) {
571 return Text->IsSuffix(s);
574 /** IsEqual(s): returns true if there is a text that equals given
576 bool IsEqual(uchar const *s) {
577 return Text->IsEqual(s);
580 /** IsContains(s): returns true if there is a text containing string s. */
581 bool IsContains(uchar const *s) {
582 return Text->IsContains(s);
585 /** IsLessThan(s): returns true if there is at least a text that is
586 * lexicographically smaller than string s. */
587 bool IsLessThan(uchar const *s) {
588 return Text->IsLessThan(s);
591 /** Count(s): Global counting */
592 unsigned Count(uchar const *s) {
593 return Text->Count(s);
596 /** CountPrefix(s): counting version of Prefix(s). */
597 unsigned CountPrefix(uchar const *s) {
598 return Text->CountPrefix(s);
601 /** CountSuffix(s): counting version of Suffix(s). */
602 unsigned CountSuffix(uchar const *s) {
603 return Text->CountSuffix(s);
606 /** CountEqual(s): counting version of Equal(s). */
607 unsigned CountEqual(uchar const *s) {
608 return Text->CountEqual(s);
611 /** CountContains(s): counting version of Contains(s). */
612 unsigned CountContains(uchar const *s) {
613 return Text->CountContains(s);
616 /** CountLessThan(s): counting version of LessThan(s). */
617 unsigned CountLessThan(uchar const *s) {
618 return Text->CountLessThan(s);
621 /** GetText(d): returns the text corresponding to document with
623 uchar* GetText(DocID d) {
625 uchar * s = Text->GetText(d);
626 return (s[0] == 1 ? (s+1) : s);
629 /** GetText(i, j): returns the texts corresponding to documents with
630 * ids i, i+1, ..., j. Texts are separated by '\0' character. */
631 // uchar* GetText(DocID i, DocID j) {
632 // uchar * s = Text->GetText(i, j);
633 // return (s[0] == 1 ? (uchar*)"" : s);
636 TextCollection *getTextCollection() {
640 /** Save: saves XML tree data structure to file. */
641 void Save(int fd, char* name );
643 /** Load: loads XML tree data structure from file. sample_rate_text
644 * indicates the sample rate for the text search data structure. */
645 static XMLTree *Load(int fd, bool load_tc, int sample_factor, char * name);
647 void insertTag(TagType tag, uint position);
652 /** Parenthesis functions */
653 treeNode Closing(treeNode x);
655 bool IsOpen(treeNode x);
658 /** Print procedure */
659 void Print(int fd,treeNode x, bool no_text);
660 void Print(int fd,treeNode x) { Print(fd,x,false); }
661 void Flush(int fd){ if (buffer) _real_flush(fd, buffer->size()); }
663 // The following are inlined here for speed
664 /** Tag(x): returns the tag identifier of node x. */
666 inline TagType Tag(treeNode x) const throw () {
668 return (TagType) (((uchar*)tags_fix)[(int) x]);
670 return get_field(tags_fix, tags_blen, x);
673 size_t idxlen = x * tags_blen;
674 size_t j = idxlen % W;
675 size_t i = idxlen / W;
676 size_t offset = W - tags_blen;
677 size_t offset2 = offset - j;
678 size_t w = tags_fix[i];
679 return (offset2 >= 0)
680 ? ( w << offset2 ) >> offset
681 : ( w >> j) | (tags_fix[i+1] << (W+offset2)) >> offset;
686 /** FirstChild(x): returns the first child of node x, or NULLT if the node is a leaf
688 treeNode FirstChild(treeNode x) {
690 return fast_first_child(Par, x);
694 /** FirstElement(x): returns the first non text, non attribute child of node x, or NULLT
697 treeNode FirstElement(treeNode node){
699 NULLT_IF(node == NULLT);
700 treeNode x = fast_first_child(Par, node);
701 NULLT_IF(x == NULLT);
704 case ATTRIBUTE_TAG_ID:
705 x = fast_next_sibling(Par,x);
706 if (x == NULLT || Tag(x) != PCDATA_TAG_ID) return x;
710 return (fast_inspect(Par,x)==OP)? x : NULLT;
718 /** NextSibling(x): returns the next sibling of node x, or NULLT if none
721 treeNode NextSibling(treeNode x) {
723 return fast_next_sibling(Par, x);
726 /** NextElement(x): returns the first non text, non attribute sibling of node x, or NULLT
729 treeNode NextElement(treeNode x)
732 x = fast_next_sibling(Par, x);
733 NULLT_IF(x == NULLT);
734 if (Tag(x) == PCDATA_TAG_ID){
736 return (fast_inspect(Par, y) == OP) ? y : NULLT;
740 /** TaggedDesc(x,tag): returns the first node tagged tag with larger
741 * preorder than x and within the subtree of x. Returns NULT if there
743 inline treeNode TaggedDescendant(treeNode x, TagType tag)
746 int s = (int) Tags->select_next(tag,node2tagpos(x));
749 treeNode y = tagpos2node(s); // transforms the tag position into a node position
751 return (fast_is_ancestor(Par,x,y) ? y : NULLT);
754 inline treeNode TaggedFollowingBelow(treeNode x, TagType tag, treeNode ancestor)
756 treeNode close = fast_find_close(Par, x);
757 treeNode s = tagpos2node(Tags->select_next(tag, close));
759 if (ancestor == Root() || s == NULLT || s < fast_find_close(Par,ancestor)) return s;
763 inline treeNode TaggedFollowingBefore(treeNode x, TagType tag, treeNode ancestor_closing)
765 treeNode close = fast_find_close(Par, x);
766 treeNode s = tagpos2node(Tags->select_next(tag, close));
768 if (ancestor_closing == Root() || s == NULLT || s < ancestor_closing) return s;
772 // TaggedSibling(x,tag): returns the first sibling of node x tagged tag, or NULLT if there is none.
773 treeNode TaggedFollowingSibling(treeNode x, TagType tag)
776 treeNode sibling = x;
778 while ((sibling = fast_next_sibling(Par, sibling)) != NULLT) {
780 if (ctag == tag) return sibling;
785 treeNode TaggedChild(treeNode x, TagType tag)
788 NULLT_IF(x==NULLT || fast_isleaf(Par,x));
790 child = fast_first_child(Par, x);
792 if (Tag(child) == tag)
795 return TaggedFollowingSibling(child, tag);