2 /******************************************************************************
\r
3 * Copyright (C) 2008 by Diego Arroyuelo *
\r
4 * Interface for the in-memory XQuery/XPath engine *
\r
6 * This program is free software; you can redistribute it and/or modify *
\r
7 * it under the terms of the GNU Lesser General Public License as published *
\r
8 * by the Free Software Foundation; either version 2 of the License, or *
\r
9 * (at your option) any later version. *
\r
11 * This program is distributed in the hope that it will be useful, *
\r
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of *
\r
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *
\r
14 * GNU Lesser General Public License for more details. *
\r
16 * You should have received a copy of the GNU Lesser General Public License *
\r
17 * along with this program; if not, write to the *
\r
18 * Free Software Foundation, Inc., *
\r
19 * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. *
\r
20 ******************************************************************************/
\r
24 #include "TextCollection/TextCollection.h"
\r
30 //clash between TextCollection/Tools.h and 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
42 // this constant is used to efficiently compute the child operation in the tree
\r
47 #define PERM_SAMPLE 10
\r
50 #define bitset(e,p) ((e)[(p)/W] |= (1<<((p)%W)))
\r
51 // cleans bit p in e
\r
52 #define bitclean(e,p) ((e)[(p)/W] &= ~(1<<((p)%W)))
\r
55 typedef int treeNode;
\r
56 typedef int TagType;
\r
66 // I know this class implements the working draft that we have but the logics seem flawed here...
\r
67 // We should have two classes. One XMLTreeBuilder and one XMLTree.
\r
68 // XMLTreeBuilder would have OpenDocument, NewOpenTag,... and CloseDocument would return an XMLTree
\r
69 // XMLTree would have only an initialized structure. If find it really ugly to check (!finished) or (!initialized)
\r
70 // in every function (FirstChild....).
\r
73 /** Balanced parentheses representation of the tree */
\r
76 /** Mapping from tag identifer to tag name */
\r
77 unsigned char **TagName;
\r
79 /** boolean flag indicating whether we are indexing empty texts or not */
\r
80 bool indexing_empty_texts;
\r
82 /** Bit vector indicating with a 1 the positions of the non-empty texts. */
\r
83 static_bitsequence_rrr02 *EBVector;
\r
85 /** Tag sequence represented with a data structure for rank and select */
\r
86 static_sequence *Tags;
\r
88 /** The texts in the XML document */
\r
89 TextCollection *Text;
\r
90 /** The texts in the XML document (cached for faster display) */
\r
91 vector<string> CachedText;
\r
93 /** Flag indicating whether the whole data structure has been constructed or
\r
94 * not. If the value is true, you cannot add more texts, nodes, etc. */
\r
97 /** Flag indicating whether the construction of the data structure has been
\r
98 * initialized or not (by calling method OpenDocument()). If this is true,
\r
99 * you cannot insert new tags or texts. */
\r
102 /* the following components are used for construction purposes */
\r
108 unsigned int *empty_texts_aux;
\r
111 // I added those two. The TagName array should always contains two special tags
\r
112 // <@> for attributes and <$> for PCDATA.
\r
113 // <$> can never be in a document (since we handle the text differently)
\r
114 // but <@> can be returned by the parser. This boolean is needed for the construction
\r
115 // of the Tag bitmap to know if <@> must be taken into account or not
\r
116 bool found_attributes;
\r
119 // Allows to disable the TextCollection for benchmarkin purposes
\r
124 /** Data structure constructor */
\r
125 XMLTree() {finished = false; initialized = false;};
\r
127 /** Data structure destructor */
\r
130 /** root(): returns the tree root. */
\r
133 /** SubtreeSize(x): the number of nodes (and attributes) in the subtree of
\r
135 int SubtreeSize(treeNode x);
\r
137 /** SubtreeTags(x,tag): the number of occurrences of tag within the subtree
\r
139 int SubtreeTags(treeNode x, TagType tag);
\r
141 /** IsLeaf(x): returns whether node x is leaf or not. In the succinct
\r
142 * representation this is just a bit inspection. */
\r
143 bool IsLeaf(treeNode x);
\r
145 /** IsAncestor(x,y): returns whether node x is ancestor of node y. */
\r
146 bool IsAncestor(treeNode x, treeNode y);
\r
148 /** IsChild(x,y): returns whether node x is parent of node y. */
\r
149 bool IsChild(treeNode x, treeNode y);
\r
151 /** NumChildren(x): number of children of node x. Constant time with the
\r
152 * data structure of Sadakane. */
\r
153 int NumChildren(treeNode x);
\r
155 /** ChildNumber(x): returns i if node x is the i-th children of its
\r
157 inline int ChildNumber(treeNode x);
\r
159 /** Depth(x): depth of node x, a simple binary rank on the parentheses
\r
161 int Depth(treeNode x);
\r
163 /** Preorder(x): returns the preorder number of node x, just regarding tree
\r
164 * nodes (and not texts). */
\r
165 int Preorder(treeNode x);
\r
167 /** Postorder(x): returns the postorder number of node x, just regarding
\r
168 * tree nodes (and not texts). */
\r
169 int Postorder(treeNode x);
\r
171 /** Tag(x): returns the tag identifier of node x. */
\r
172 TagType Tag(treeNode x);
\r
174 /** DocIds(x): returns the range (i.e., a pair of integers) of document
\r
175 * identifiers that descend from node x. */
\r
176 range DocIds(treeNode x);
\r
178 /** Parent(x): returns the parent node of node x. */
\r
179 treeNode Parent(treeNode x);
\r
181 /** Child(x,i): returns the i-th child of node x, assuming it exists. */
\r
182 treeNode Child(treeNode x, int i);
\r
184 /** FirstChild(x): returns the first child of node x, assuming it exists.
\r
185 * Very fast in BP. */
\r
186 treeNode FirstChild(treeNode x);
\r
188 /** NextSibling(x): returns the next sibling of node x, assuming it
\r
190 treeNode NextSibling(treeNode x);
\r
192 /** PrevSibling(x): returns the previous sibling of node x, assuming it
\r
194 treeNode PrevSibling(treeNode x);
\r
196 /** TaggedChild(x,i,tag): returns the i-th child of node x tagged tag, or
\r
197 * NULLT if there is none. Because of the balanced-parentheses representation
\r
198 * of the tree, this operation is not supported efficiently, just iterating
\r
199 * among the children of node x until finding the desired child. */
\r
200 treeNode TaggedChild(treeNode x, int i, TagType tag);
\r
202 /** TaggedDesc(x,tag): returns the first node tagged tag with larger
\r
203 * preorder than x and within the subtree of x. Returns NULT if there
\r
205 treeNode TaggedDesc(treeNode x, TagType tag);
\r
207 /** TaggedNext(x,tag): returns the first node tagged tag with larger
\r
208 * preorder than x. Returns NULT if there is none. */
\r
209 treeNode TaggedNext(treeNode x, TagType tag);
\r
211 /** TaggedPrec(x,tag): returns the first node tagged tag with smaller
\r
212 * preorder than x and not an ancestor of x. Returns NULLT if there
\r
214 treeNode TaggedPrec(treeNode x, TagType tag);
\r
216 /** TaggedFoll(x,tag): returns the first node tagged tag with larger
\r
217 * preorder than x and not in the subtree of x. Returns NULLT if there
\r
219 treeNode TaggedFoll(treeNode x, TagType tag);
\r
221 /** TaggedFollowingSibling(x,tag) */
\r
222 treeNode TaggedFollowingSibling(treeNode x, TagType tag);
\r
224 /** TaggedAncestor(x, tag): returns the closest ancestor of x tagged
\r
225 * tag. Return NULLT is there is none. */
\r
226 treeNode TaggedAncestor(treeNode x, TagType tag);
\r
228 /** PrevText(x): returns the document identifier of the text to the left of
\r
229 * node x, or NULLT if x is the root node. */
\r
230 DocID PrevText(treeNode x);
\r
232 /** NextText(x): returns the document identifier of the text to the right of
\r
233 * node x, or NULLT if x is the root node. */
\r
234 DocID NextText(treeNode x);
\r
236 /** MyText(x): returns the document identifier of the text below node x, or
\r
237 * NULLT if x is not a leaf node. */
\r
238 DocID MyText(treeNode x);
\r
240 /** TextXMLId(d): returns the preorder of document with identifier d in the
\r
241 * tree consisting of all tree nodes and all text nodes. */
\r
242 int TextXMLId(DocID d);
\r
244 /** NodeXMLId(x): returns the preorder of node x in the tree consisting of
\r
245 * all tree nodes and all text nodes. */
\r
246 int NodeXMLId(treeNode x);
\r
248 /** ParentNode(d): returns the parent node of document identifier d. */
\r
249 treeNode ParentNode(DocID d);
\r
251 /** OpenDocument(empty_texts,sample_rate_text,dtc): initilizes the construction
\r
252 * of the data structure for the XML document. Parameter empty_texts
\r
253 * indicates whether we index empty texts in document or not. Parameter
\r
254 * sample_rate_text indicates the sampling rate for the text searching data
\r
255 * structures (small values get faster searching but a bigger space
\r
256 * requirement). dtc disable the use of the TextCollection
\r
257 * (i.e. everything is considered an empty text *)
\r
258 * Returns a non-zero value upon success, NULLT in case of
\r
261 int OpenDocument(bool empty_texts, int sample_rate_text, bool dtc);
\r
263 /** CloseDocument(): finishes the construction of the data structure for
\r
264 * the XML document. Tree and tags are represented in the final form,
\r
265 * dynamic data structures are made static, and the flag "finished" is set
\r
266 * to true. After that, the data structure can be queried. */
\r
267 int CloseDocument();
\r
269 /** NewOpenTag(tagname): indicates the event of finding a new opening tag
\r
270 * in the document. Tag name is given. Returns a non-zero value upon
\r
271 * success, and returns NULLT in case of error. */
\r
272 int NewOpenTag(unsigned char *tagname);
\r
274 /** NewClosingTag(tagname): indicates the event of finding a new closing tag
\r
275 * in the document. Tag name is given. Returns a non-zero value upon
\r
276 * success, and returns NULLT in case of error. */
\r
277 int NewClosingTag(unsigned char *tagname);
\r
279 /** NewText(s): indicates the event of finding a new (non-empty) text s in
\r
280 * the document. The new text is inserted within the text collection.
\r
281 * Returns a non-zero value upon success, NULLT in case of error. */
\r
282 int NewText(unsigned char *s);
\r
284 /** NewEmptyText(): indicates the event of finding a new empty text in the
\r
285 * document. In case of indexing empty and non-empty texts, we insert the
\r
286 * empty texts into the text collection. In case of indexing only non-empty
\r
287 * texts, it just indicates an empty text in the bit vector of empty texts.
\r
288 * Returns a non-zero value upon success, NULLT in case of error. */
\r
289 int NewEmptyText();
\r
291 /** GetTagId(tagname): returns the tag identifier corresponding to a given
\r
292 * tag name. Returns NULLT in case that the tag name does not exists. */
\r
293 TagType GetTagId(unsigned char *tagname);
\r
295 /** GetTagName(tagid): returns the tag name of a given tag identifier.
\r
296 * Returns NULL in case that the tag identifier is not valid.*/
\r
297 unsigned char *GetTagName(TagType tagid);
\r
301 /** GetTagName(tagid): returns the tag name of a given tag identifier.
\r
302 * The result is just a reference and should not be freed by the caller.
\r
304 const unsigned char *GetTagNameByRef(TagType tagid);
\r
307 /** RegisterTag adds a new tag to the tag collection this is needed
\r
308 * if the query contains a tag which is not in the document, we need
\r
309 * to give this new tag a fresh id and store it somewhere. A logical
\r
311 * We might want to use a hashtable instead of an array though.
\r
313 TagType RegisterTag(unsigned char *tagname);
\r
315 bool EmptyText(DocID i) {
\r
316 return Text->EmptyText(i);
\r
318 /** Prefix(s): search for texts prefixed by string s. */
\r
319 TextCollection::document_result Prefix(uchar const *s) {
\r
320 return Text->Prefix(s);
\r
323 /** Suffix(s): search for texts having string s as a suffix. */
\r
324 TextCollection::document_result Suffix(uchar const *s) {
\r
325 return Text->Suffix(s);
\r
328 /** Equal(s): search for texts equal to string s. */
\r
329 TextCollection::document_result Equal(uchar const *s) {
\r
330 return Text->Equal(s);
\r
333 /** Contains(s): search for texts containing string s. */
\r
334 TextCollection::document_result Contains(uchar const *s) {
\r
335 return Text->Contains(s);
\r
338 /** LessThan(s): returns document identifiers for the texts that
\r
339 * are lexicographically smaller than string s. */
\r
340 TextCollection::document_result LessThan(uchar const *s) {
\r
341 return Text->LessThan(s);
\r
344 /** IsPrefix(x): returns true if there is a text prefixed by string s. */
\r
345 bool IsPrefix(uchar const *s) {
\r
346 return Text->IsPrefix(s);
\r
349 /** IsSuffix(s): returns true if there is a text having string s as a
\r
351 bool IsSuffix(uchar const *s) {
\r
352 return Text->IsSuffix(s);
\r
355 /** IsEqual(s): returns true if there is a text that equals given
\r
357 bool IsEqual(uchar const *s) {
\r
358 return Text->IsEqual(s);
\r
361 /** IsContains(s): returns true if there is a text containing string s. */
\r
362 bool IsContains(uchar const *s) {
\r
363 return Text->IsContains(s);
\r
366 /** IsLessThan(s): returns true if there is at least a text that is
\r
367 * lexicographically smaller than string s. */
\r
368 bool IsLessThan(uchar const *s) {
\r
369 return Text->IsLessThan(s);
\r
372 /** CountPrefix(s): counting version of Prefix(s). */
\r
373 unsigned CountPrefix(uchar const *s) {
\r
374 return Text->CountPrefix(s);
\r
377 /** CountSuffix(s): counting version of Suffix(s). */
\r
378 unsigned CountSuffix(uchar const *s) {
\r
379 return Text->CountSuffix(s);
\r
382 /** CountEqual(s): counting version of Equal(s). */
\r
383 unsigned CountEqual(uchar const *s) {
\r
384 return Text->CountEqual(s);
\r
387 /** CountContains(s): counting version of Contains(s). */
\r
388 unsigned CountContains(uchar const *s) {
\r
389 return Text->CountContains(s);
\r
392 /** CountLessThan(s): counting version of LessThan(s). */
\r
393 unsigned CountLessThan(uchar const *s) {
\r
394 return CountLessThan(s);
\r
397 /** GetText(d): returns the text corresponding to document with
\r
399 uchar* GetText(DocID d) {
\r
400 return Text->GetText(d);
\r
403 uchar* GetCachedText(DocID d) {
\r
404 uchar * str = (uchar*) calloc(sizeof(char),(CachedText.at(d).size() + 1));
\r
405 strcpy((char*) str,(const char*) CachedText.at(d).c_str());
\r
406 return (uchar*) (str);
\r
409 TextCollection *getTextCollection() {
\r
412 /** Save: saves XML tree data structure to file. */
\r
413 void Save(unsigned char *filename);
\r
415 /** Load: loads XML tree data structure from file. sample_rate_text
\r
416 * indicates the sample rate for the text search data structure. */
\r
417 static XMLTree *Load(unsigned char *filename, int sample_rate_text);
\r