1 /******************************************************************************
\r
2 * Copyright (C) 2008 by Diego Arroyuelo *
\r
3 * Interface for the in-memory XQuery/XPath engine *
\r
5 * This program is free software; you can redistribute it and/or modify *
\r
6 * it under the terms of the GNU Lesser General Public License as published *
\r
7 * by the Free Software Foundation; either version 2 of the License, or *
\r
8 * (at your option) any later version. *
\r
10 * This program is distributed in the hope that it will be useful, *
\r
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of *
\r
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *
\r
13 * GNU Lesser General Public License for more details. *
\r
15 * You should have received a copy of the GNU Lesser General Public License *
\r
16 * along with this program; if not, write to the *
\r
17 * Free Software Foundation, Inc., *
\r
18 * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. *
\r
19 ******************************************************************************/
\r
25 #include <unordered_set>
\r
26 #include <unordered_map>
\r
28 #include "TextCollection/TextCollectionBuilder.h"
\r
35 #include <libcds/includes/basics.h>
\r
36 #include <static_bitsequence.h>
\r
37 #include <alphabet_mapper.h>
\r
38 #include <static_sequence.h>
\r
39 using SXSI::TextCollection;
\r
40 using SXSI::TextCollectionBuilder;
\r
43 // this constant is used to efficiently compute the child operation in the tree
\r
48 #define PERM_SAMPLE 10
\r
51 typedef int treeNode;
\r
52 typedef int TagType;
\r
60 // Encoding of the XML Document :
\r
61 // The following TAGs and IDs are fixed, "" is the tag of the root.
\r
62 // a TextNode is represented by a leaf <<$>></<$>> The DocId in the TextCollection
\r
63 // of that leaf is kept in a bit sequence.
\r
64 // a TextNode below an attribute is likewise represented by a leaf <<@$>><</@$>>
\r
65 // An element <e a1="v1" a2="v2" ... an="vn" > ...</e> the representation is:
\r
66 // <e><<@>> <<@>a1> <<$@>>DocID(v1)</<$@>></<@>a1> ... </<@>> .... </e>
\r
67 // Hence the attributes (if any) are always below the first child of their element,
\r
68 // as the children of a fake node <@>.
\r
71 #define DOCUMENT_OPEN_TAG ""
\r
72 #define DOCUMENT_TAG_ID 0
\r
73 #define ATTRIBUTE_OPEN_TAG "<@>"
\r
74 #define ATTRIBUTE_TAG_ID 1
\r
75 #define PCDATA_OPEN_TAG "<$>"
\r
76 #define PCDATA_TAG_ID 2
\r
77 #define ATTRIBUTE_DATA_OPEN_TAG "<@$>"
\r
78 #define ATTRIBUTE_DATA_TAG_ID 3
\r
79 #define CLOSING_TAG "</>"
\r
80 #define CLOSING_TAG_ID 4
\r
81 #define DOCUMENT_CLOSE_TAG "/"
\r
82 #define ATTRIBUTE_CLOSE_TAG "/<@>"
\r
83 #define PCDATA_CLOSE_TAG "/<$>"
\r
84 #define ATTRIBUTE_DATA_CLOSE_TAG "/<@$>"
\r
87 typedef std::unordered_set<int> TagIdSet;
\r
88 typedef std::unordered_map<std::string,int> TagIdMap;
\r
89 typedef TagIdMap::const_iterator TagIdMapIT;
\r
91 #define REGISTER_TAG(v,h,t) do { (h)->insert(std::make_pair((t),(v)->size()));\
\r
92 (v)->push_back(t); } while (false)
\r
95 // returns NULLT if the test is true
\r
96 #define NULLT_IF(x) do { if (x) return NULLT; } while (0)
\r
103 class XMLTreeBuilder;
\r
107 // Only the builder can access the constructor
\r
108 friend class XMLTreeBuilder;
\r
111 /** Balanced parentheses representation of the tree */
\r
114 /** Mapping from tag identifer to tag name */
\r
115 std::vector<std::string> *TagName;
\r
118 /** Bit vector indicating with a 1 the positions of the non-empty texts. */
\r
119 static_bitsequence *EBVector;
\r
121 /** Tag sequence represented with a data structure for rank and select */
\r
122 static_sequence *Tags;
\r
124 uint tags_blen, tags_len;
\r
126 /** The texts in the XML document */
\r
127 TextCollection *Text;
\r
129 // Allows to disable the TextCollection for benchmarkin purposes
\r
134 std::string * buffer;
\r
135 void myfputs(const char* s, FILE * fp){
\r
137 if (buffer->size() >= 100000){
\r
138 fputs(buffer->c_str(),fp);
\r
143 void myfputc(const char c, FILE*fp){
\r
144 buffer->append(1,c);
\r
145 if (buffer->size() >= 100000){
\r
146 fputs(buffer->c_str(),fp);
\r
150 void mybufferflush(FILE* fp){
\r
151 fputs(buffer->c_str(), fp);
\r
155 size_t myfprintf(const char* s, FILE * fp){
\r
158 size_t i = buffer->size();
\r
160 size_t j = buffer->size();
\r
161 if (buffer->size() >= 100000){
\r
162 fputs(buffer->c_str(),fp);
\r
168 void PrintNode(treeNode n, int fd);
\r
169 /** Data structure constructors */
\r
170 XMLTree(){ buffer = 0;};
\r
172 // non const pointer are freed by this method.
\r
173 XMLTree( pb * const par, uint npar, std::vector<std::string> * const TN, TagIdMap * const tim, uint *empty_texts_bmp, TagType *tags,
\r
174 TextCollection * const TC, bool dis_tc);
\r
177 /** Data structure destructor */
\r
180 /** root(): returns the tree root. */
\r
181 treeNode Root() { return 0; }
\r
183 /** Size() : Number of parenthesis */
\r
184 unsigned int Size(){
\r
188 /** SubtreeSize(x): the number of nodes (and attributes) in the subtree of
\r
190 int SubtreeSize(treeNode x);
\r
192 /** SubtreeTags(x,tag): the number of occurrences of tag within the subtree
\r
194 int SubtreeTags(treeNode x, TagType tag);
\r
196 /** SubtreeElements(x) of element nodes in the subtree of x
\r
198 int SubtreeElements(treeNode x);
\r
200 /** IsLeaf(x): returns whether node x is leaf or not. In the succinct
\r
201 * representation this is just a bit inspection. */
\r
203 bool IsLeaf(treeNode x);
\r
205 /** IsAncestor(x,y): returns whether node x is ancestor of node y. */
\r
207 bool IsAncestor(treeNode x, treeNode y);
\r
209 /** IsChild(x,y): returns whether node x is parent of node y. */
\r
210 bool IsChild(treeNode x, treeNode y);
\r
212 /** IsFirstChild(x): returns whether node x is the first child of its parent. */
\r
214 bool IsFirstChild(treeNode x);
\r
216 /** NumChildren(x): number of children of node x. Constant time with the
\r
217 * data structure of Sadakane. */
\r
218 int NumChildren(treeNode x);
\r
220 /** ChildNumber(x): returns i if node x is the i-th children of its
\r
222 int ChildNumber(treeNode x);
\r
224 /** Depth(x): depth of node x, a simple binary rank on the parentheses
\r
226 int Depth(treeNode x);
\r
228 /** Preorder(x): returns the preorder number of node x, just regarding tree
\r
229 * nodes (and not texts). */
\r
230 int Preorder(treeNode x);
\r
232 /** Postorder(x): returns the postorder number of node x, just regarding
\r
233 * tree nodes (and not texts). */
\r
234 int Postorder(treeNode x);
\r
236 /** Tag(x): returns the tag identifier of node x. */
\r
237 TagType Tag(treeNode x) {
\r
238 if (tags_blen == 8)
\r
239 return (TagType) (((uchar*)tags_fix)[(int) x]);
\r
241 return (TagType) get_field(tags_fix,tags_blen, (int) x);
\r
244 /** DocIds(x): returns the range (i.e., a pair of integers) of document
\r
245 * identifiers that descend from node x. */
\r
246 range DocIds(treeNode x);
\r
248 /** Parent(x): returns the parent node of node x. */
\r
249 treeNode Parent(treeNode x);
\r
250 /* Assumes x is neither 0 nor -1 */
\r
252 /** Child(x,i): returns the i-th child of node x, assuming it exists. */
\r
253 treeNode Child(treeNode x, int i);
\r
255 /** FirstChild(x): returns the first child of node x, or NULLT if the node is a leaf
\r
257 treeNode FirstChild(treeNode x);
\r
259 /** FirstElement(x): returns the first non text, non attribute child of node x, or NULLT
\r
262 treeNode FirstElement(treeNode x);
\r
264 /** LastChild(x): returns the last child of node x. */
\r
265 treeNode LastChild(treeNode x);
\r
267 /** NextSibling(x): returns the next sibling of node x, or NULLT if none
\r
270 treeNode NextSibling(treeNode x);
\r
272 /** NextElement(x): returns the first non text, non attribute sibling of node x, or NULLT
\r
275 treeNode NextElement(treeNode x);
\r
277 /** PrevSibling(x): returns the previous sibling of node x, assuming it
\r
280 treeNode PrevSibling(treeNode x);
\r
282 /** TaggedChild(x,tag): returns the first child of node x tagged tag, or
\r
283 * NULLT if there is none. Because of the balanced-parentheses representation
\r
284 * of the tree, this operation is not supported efficiently, just iterating
\r
285 * among the children of node x until finding the desired child. */
\r
286 treeNode TaggedChild(treeNode x, TagType tag);
\r
288 treeNode SelectChild(treeNode x, TagIdSet * tags);
\r
290 /** TaggedFollowingSibling(x,tag): returns the first sibling of node x tagged tag, or
\r
291 * NULLT if there is none. */
\r
292 treeNode TaggedFollowingSibling(treeNode x, TagType tag);
\r
294 treeNode SelectFollowingSibling(treeNode x, TagIdSet * tags);
\r
296 /** TaggedDesc(x,tag): returns the first node tagged tag with larger
\r
297 * preorder than x and within the subtree of x. Returns NULT if there
\r
299 treeNode TaggedDescendant(treeNode x, TagType tag);
\r
301 treeNode SelectDescendant(treeNode x, TagIdSet * tags);
\r
303 /** TaggedPrec(x,tag): returns the first node tagged tag with smaller
\r
304 * preorder than x and not an ancestor of x. Returns NULLT if there
\r
306 treeNode TaggedPreceding(treeNode x, TagType tag);
\r
308 /** TaggedFoll(x,tag): returns the first node tagged tag with larger
\r
309 * preorder than x and not in the subtree of x. Returns NULLT if there
\r
311 treeNode TaggedFollowing(treeNode x, TagType tag);
\r
313 treeNode TaggedFollowingBelow(treeNode x, TagType tag,treeNode ancestor);
\r
315 treeNode SelectFollowingBelow(treeNode x, TagIdSet * tags, treeNode ancestor);
\r
317 treeNode TaggedFollowingBefore(treeNode x, TagType tag,treeNode closing);
\r
319 treeNode SelectFollowingBefore(treeNode x, TagIdSet * tags, treeNode closing);
\r
321 /** TaggedAncestor(x, tag): returns the closest ancestor of x tagged
\r
322 * tag. Return NULLT is there is none. */
\r
323 treeNode TaggedAncestor(treeNode x, TagType tag);
\r
325 /** PrevText(x): returns the document identifier of the text to the left of
\r
326 * node x, or NULLT if x is the root node. */
\r
327 DocID PrevText(treeNode x);
\r
329 /** NextText(x): returns the document identifier of the text to the right of
\r
330 * node x, or NULLT if x is the root node. */
\r
331 DocID NextText(treeNode x);
\r
333 /** MyText(x): returns the document identifier of the text below node x, or
\r
334 * NULLT if x is not a leaf node. */
\r
335 DocID MyText(treeNode x);
\r
336 DocID MyTextUnsafe(treeNode x);
\r
338 /** TextXMLId(d): returns the preorder of document with identifier d in the
\r
339 * tree consisting of all tree nodes and all text nodes. */
\r
340 int TextXMLId(DocID d);
\r
342 /** NodeXMLId(x): returns the preorder of node x in the tree consisting of
\r
343 * all tree nodes and all text nodes. */
\r
344 int NodeXMLId(treeNode x);
\r
346 /** ParentNode(d): returns the parent node of document identifier d. */
\r
347 treeNode ParentNode(DocID d);
\r
349 treeNode PrevNode(DocID d);
\r
351 /** GetTagId(tagname): returns the tag identifier corresponding to a given
\r
352 * tag name. Returns NULLT in case that the tag name does not exists. */
\r
353 TagType GetTagId(unsigned char *tagname);
\r
355 /** GetTagName(tagid): returns the tag name of a given tag identifier.
\r
356 * Returns NULL in case that the tag identifier is not valid.*/
\r
357 unsigned char *GetTagName(TagType tagid);
\r
359 /** GetTagName(tagid): returns the tag name of a given tag identifier.
\r
360 * The result is just a reference and should not be freed by the caller.
\r
362 const unsigned char *GetTagNameByRef(TagType tagid);
\r
364 /** RegisterTag adds a new tag to the tag collection this is needed
\r
365 * if the query contains a tag which is not in the document, we need
\r
366 * to give this new tag a fresh id and store it somewhere. A logical
\r
368 * We might want to use a hashtable instead of an array though.
\r
370 TagType RegisterTag(unsigned char *tagname);
\r
372 bool EmptyText(DocID i) {
\r
373 return Text->EmptyText(i);
\r
376 /** Prefix(s): search for texts prefixed by string s. */
\r
377 TextCollection::document_result Prefix(uchar const *s) {
\r
378 return Text->Prefix(s);
\r
381 /** Suffix(s): search for texts having string s as a suffix. */
\r
382 TextCollection::document_result Suffix(uchar const *s) {
\r
383 return Text->Suffix(s);
\r
386 /** Equal(s): search for texts equal to string s. */
\r
387 TextCollection::document_result Equals(uchar const *s) {
\r
388 return Text->Equal(s);
\r
391 /** Contains(s): search for texts containing string s. */
\r
392 TextCollection::document_result Contains(uchar const *s) {
\r
393 return Text->Contains(s);
\r
396 /** LessThan(s): returns document identifiers for the texts that
\r
397 * are lexicographically smaller than string s. */
\r
398 TextCollection::document_result LessThan(uchar const *s) {
\r
399 return Text->LessThan(s);
\r
402 /** IsPrefix(x): returns true if there is a text prefixed by string s. */
\r
403 bool IsPrefix(uchar const *s) {
\r
404 return Text->IsPrefix(s);
\r
407 /** IsSuffix(s): returns true if there is a text having string s as a
\r
409 bool IsSuffix(uchar const *s) {
\r
410 return Text->IsSuffix(s);
\r
413 /** IsEqual(s): returns true if there is a text that equals given
\r
415 bool IsEqual(uchar const *s) {
\r
416 return Text->IsEqual(s);
\r
419 /** IsContains(s): returns true if there is a text containing string s. */
\r
420 bool IsContains(uchar const *s) {
\r
421 return Text->IsContains(s);
\r
424 /** IsLessThan(s): returns true if there is at least a text that is
\r
425 * lexicographically smaller than string s. */
\r
426 bool IsLessThan(uchar const *s) {
\r
427 return Text->IsLessThan(s);
\r
430 /** Count(s): Global counting */
\r
431 unsigned Count(uchar const *s) {
\r
432 return Text->Count(s);
\r
435 /** CountPrefix(s): counting version of Prefix(s). */
\r
436 unsigned CountPrefix(uchar const *s) {
\r
437 return Text->CountPrefix(s);
\r
440 /** CountSuffix(s): counting version of Suffix(s). */
\r
441 unsigned CountSuffix(uchar const *s) {
\r
442 return Text->CountSuffix(s);
\r
445 /** CountEqual(s): counting version of Equal(s). */
\r
446 unsigned CountEqual(uchar const *s) {
\r
447 return Text->CountEqual(s);
\r
450 /** CountContains(s): counting version of Contains(s). */
\r
451 unsigned CountContains(uchar const *s) {
\r
452 return Text->CountContains(s);
\r
455 /** CountLessThan(s): counting version of LessThan(s). */
\r
456 unsigned CountLessThan(uchar const *s) {
\r
457 return Text->CountLessThan(s);
\r
460 /** GetText(d): returns the text corresponding to document with
\r
462 uchar* GetText(DocID d) {
\r
464 uchar * s = Text->GetText(d);
\r
465 return (s[0] == 1 ? (s+1) : s);
\r
468 /** GetText(i, j): returns the texts corresponding to documents with
\r
469 * ids i, i+1, ..., j. Texts are separated by '\0' character. */
\r
470 // uchar* GetText(DocID i, DocID j) {
\r
471 // uchar * s = Text->GetText(i, j);
\r
472 // return (s[0] == 1 ? (uchar*)"" : s);
\r
475 TextCollection *getTextCollection() {
\r
479 /** Save: saves XML tree data structure to file. */
\r
480 void Save(int fd, char *filename);
\r
482 /** Load: loads XML tree data structure from file. sample_rate_text
\r
483 * indicates the sample rate for the text search data structure. */
\r
484 static XMLTree *Load(int fd, char *filename, bool load_tc, int sample_factor);
\r
486 void insertTag(TagType tag, uint position);
\r
488 void print_stats();
\r
491 /** Parenthesis functions */
\r
492 treeNode Closing(treeNode x);
\r
494 bool IsOpen(treeNode x);
\r
497 /** Print procedure */
\r
498 void Print(int fd,treeNode x, bool no_text);
\r
499 void Print(int fd,treeNode x) { Print(fd,x,false); }
\r