\r
#ifndef XMLTREE_H_\r
#define XMLTREE_H_\r
-#include "TextCollection/TextCollection.h"\r
+#include "TextCollection/TextCollectionBuilder.h"\r
#include <stdio.h>\r
#include <stdlib.h>\r
#include <cstring>\r
+\r
+//KIM : OJO\r
+//clash between TextCollection/Tools.h and libcds/includes/basics.h\r
+#undef W\r
+#undef WW\r
+#undef Wminusone\r
+\r
#include "bp.h"\r
#include <static_bitsequence.h>\r
#include <alphabet_mapper.h>\r
#include <static_sequence.h>\r
using SXSI::TextCollection;\r
+using SXSI::TextCollectionBuilder;\r
\r
\r
// this constant is used to efficiently compute the child operation in the tree\r
\r
#define NULLT -1\r
\r
+#define PERM_SAMPLE 10\r
+\r
// sets bit p in e\r
#define bitset(e,p) ((e)[(p)/W] |= (1<<((p)%W)))\r
// cleans bit p in e\r
} range;\r
\r
\r
+//KIM : OJO\r
+// I know this class implements the working draft that we have but the logics seem flawed here...\r
+// We should have two classes. One XMLTreeBuilder and one XMLTree.\r
+// XMLTreeBuilder would have OpenDocument, NewOpenTag,... and CloseDocument would return an XMLTree\r
+// XMLTree would have only an initialized structure. If find it really ugly to check (!finished) or (!initialized)\r
+// in every function (FirstChild....).\r
+\r
class XMLTree {\r
/** Balanced parentheses representation of the tree */\r
bp *Par;\r
static_bitsequence_rrr02 *EBVector; \r
\r
/** Tag sequence represented with a data structure for rank and select */\r
- static_sequence_wvtree *Tags;\r
+ static_sequence *Tags;\r
+ uint * tags_fix;\r
+ uint tags_blen, tags_len;\r
\r
/** The texts in the XML document */\r
+ TextCollectionBuilder *TextBuilder;\r
TextCollection *Text;\r
+\r
+ /** The texts in the XML document (cached for faster display) */\r
+ vector<string> CachedText;\r
\r
/** Flag indicating whether the whole data structure has been constructed or \r
* not. If the value is true, you cannot add more texts, nodes, etc. */\r
pb *par_aux;\r
TagType *tags_aux;\r
int npar;\r
+ int parArraySize;\r
int ntagnames;\r
unsigned int *empty_texts_aux;\r
+\r
+ // KIM : OJO\r
+ // I added those two. The TagName array should always contains two special tags\r
+ // <@> for attributes and <$> for PCDATA.\r
+ // <$> can never be in a document (since we handle the text differently)\r
+ // but <@> can be returned by the parser. This boolean is needed for the construction\r
+ // of the Tag bitmap to know if <@> must be taken into account or not\r
+ bool found_attributes;\r
+\r
+ // KIM : OJO\r
+ // Allows to disable the TextCollection for benchmarkin purposes\r
+ bool disable_tc;\r
\r
public:\r
+ void print_stats();\r
\r
/** Data structure constructor */\r
- XMLTree() {finished = false; initialized = false;}; \r
+ XMLTree() {finished = false; initialized = false; Text = 0; TextBuilder = 0; }; \r
\r
/** Data structure destructor */\r
~XMLTree();\r
\r
/** IsChild(x,y): returns whether node x is parent of node y. */\r
bool IsChild(treeNode x, treeNode y);\r
+\r
+ /** IsChild(x,y): returns whether node x is the first child of its parent */\r
+ bool IsFirstChild(treeNode x);\r
\r
/** NumChildren(x): number of children of node x. Constant time with the \r
* data structure of Sadakane. */\r
* Very fast in BP. */\r
treeNode FirstChild(treeNode x);\r
\r
+ /** LastChild(x): returns the last child of node x. \r
+ * Implemented by Kim naively. */\r
+ treeNode LastChild(treeNode x);\r
+\r
/** NextSibling(x): returns the next sibling of node x, assuming it \r
* exists. */\r
+\r
treeNode NextSibling(treeNode x);\r
\r
/** PrevSibling(x): returns the previous sibling of node x, assuming it \r
* is none. */\r
treeNode TaggedDesc(treeNode x, TagType tag);\r
\r
+\r
+ treeNode TaggedBelow(treeNode x, TagType *childtags, unsigned int ctlen,\r
+ TagType *desctags, unsigned int dtlen);\r
+ \r
+ treeNode TaggedNext(treeNode x, TagType *childtags, unsigned int ctlen,\r
+ TagType *folltags, unsigned int flen,treeNode root);\r
+\r
+ treeNode TaggedDescOnly(treeNode x, TagType *desctags, unsigned int dtlen);\r
+ \r
+ treeNode TaggedDescOrFollOnly(treeNode x, TagType *folltags, unsigned int flen,\r
+ treeNode root);\r
+\r
+ treeNode TaggedFollOnly(treeNode x, TagType *folltags, unsigned int flen,\r
+ treeNode root);\r
+ \r
+\r
/** TaggedPrec(x,tag): returns the first node tagged tag with smaller \r
* preorder than x and not an ancestor of x. Returns NULLT if there \r
* is none. */\r
* preorder than x and not in the subtree of x. Returns NULLT if there \r
* is none. */\r
treeNode TaggedFoll(treeNode x, TagType tag);\r
- \r
+\r
+ treeNode TaggedFollBelow(treeNode x, TagType tag,treeNode root); \r
+ \r
+ /** TaggedFollowingSibling(x,tag) */\r
+ treeNode TaggedFollowingSibling(treeNode x, TagType tag);\r
+\r
+ /** TaggedAncestor(x, tag): returns the closest ancestor of x tagged \r
+ * tag. Return NULLT is there is none. */\r
+ treeNode TaggedAncestor(treeNode x, TagType tag);\r
+ \r
/** PrevText(x): returns the document identifier of the text to the left of \r
* node x, or NULLT if x is the root node. */\r
DocID PrevText(treeNode x);\r
\r
/** ParentNode(d): returns the parent node of document identifier d. */\r
treeNode ParentNode(DocID d);\r
+ treeNode PrevNode(DocID d);\r
\r
- /** OpenDocument(empty_texts,sample_rate_text): initilizes the construction\r
+ /** OpenDocument(empty_texts,sample_rate_text,dtc): initilizes the construction\r
* of the data structure for the XML document. Parameter empty_texts \r
* indicates whether we index empty texts in document or not. Parameter \r
* sample_rate_text indicates the sampling rate for the text searching data\r
* structures (small values get faster searching but a bigger space \r
- * requirement). Returns a non-zero value upon success, NULLT in case of \r
+ * requirement). dtc disable the use of the TextCollection\r
+ * (i.e. everything is considered an empty text *)\r
+ * Returns a non-zero value upon success, NULLT in case of \r
* error. */\r
- int OpenDocument(bool empty_texts, int sample_rate_text);\r
+\r
+ int OpenDocument(bool empty_texts, int sample_rate_text, bool dtc);\r
\r
/** CloseDocument(): finishes the construction of the data structure for \r
* the XML document. Tree and tags are represented in the final form, \r
* Returns NULL in case that the tag identifier is not valid.*/\r
unsigned char *GetTagName(TagType tagid);\r
\r
+\r
+ // OJO\r
+ /** GetTagName(tagid): returns the tag name of a given tag identifier. \r
+ * The result is just a reference and should not be freed by the caller.\r
+ */\r
+ const unsigned char *GetTagNameByRef(TagType tagid);\r
+\r
+ //OJO\r
+ /** RegisterTag adds a new tag to the tag collection this is needed\r
+ * if the query contains a tag which is not in the document, we need\r
+ * to give this new tag a fresh id and store it somewhere. A logical\r
+ * choice is here.\r
+ * We might want to use a hashtable instead of an array though.\r
+ */\r
+ TagType RegisterTag(unsigned char *tagname);\r
+\r
+ bool EmptyText(DocID i) {\r
+ return Text->EmptyText(i);\r
+ }\r
/** Prefix(s): search for texts prefixed by string s. */\r
TextCollection::document_result Prefix(uchar const *s) {\r
return Text->Prefix(s);\r
bool IsLessThan(uchar const *s) {\r
return Text->IsLessThan(s);\r
}\r
+ \r
+ /** Count(s): Global counting */\r
+ unsigned Count(uchar const *s) {\r
+ return Text->Count(s);\r
+ }\r
\r
/** CountPrefix(s): counting version of Prefix(s). */\r
unsigned CountPrefix(uchar const *s) {\r
/** GetText(d): returns the text corresponding to document with\r
* id d. */\r
uchar* GetText(DocID d) {\r
- return Text->GetText(d);\r
+ return Text->GetText(d);\r
+ }\r
+\r
+ uchar* GetCachedText(DocID d) {\r
+ uchar * str = (uchar*) calloc(sizeof(char),(CachedText.at(d).size() + 1));\r
+ strcpy((char*) str,(const char*) CachedText.at(d).c_str());\r
+ return (uchar*) (str);\r
}\r
\r
TextCollection *getTextCollection() {\r
projects / SXSI / XMLTree.git / blobdiff