X-Git-Url: http://git.nguyen.vg/gitweb/?a=blobdiff_plain;f=XMLTree.cpp;h=cdfe95296d6e4310ac04356d18e374af0688600d;hb=39ee9d4c866fd5e4ba478ef65dc7e714e80c0b91;hp=a696d7f62507ce8a10b1ea256b3c0902ebc3d168;hpb=2220c6ca1fb8ebefd48d6f5d9a6be3afc7325da5;p=SXSI%2FXMLTree.git

diff --git a/XMLTree.cpp b/XMLTree.cpp
index a696d7f..cdfe952 100644
--- a/XMLTree.cpp
+++ b/XMLTree.cpp
@@ -1,7 +1,7 @@
 #include "XMLTree.h"
 #include <cstring>
 
- // functions to convert tag positions to the corresponding tree node and viceversa. 
+// functions to convert tag positions to the corresponding tree node and viceversa. 
 // These are implemented in order to be able to change the tree and Tags representations, 
 // without affecting the code so much.
 // Current implementation corresponds to balanced-parentheses representation for
@@ -17,7 +17,8 @@ inline int node2tagpos(treeNode x) {
    return (int)x;
 }
 
-// to prevent suprious "unused result" warnings
+
+//KIM OJO to prevent suprious "unused result" warnings
 
 inline void ufread(void *ptr, size_t size, size_t nmemb, FILE *stream){
   size_t res;
@@ -27,6 +28,7 @@ inline void ufread(void *ptr, size_t size, size_t nmemb, FILE *stream){
 
   return;
 }
+
 inline void ufwrite(const void *ptr, size_t size, size_t nmemb, FILE *stream){
   size_t res;
   res = fwrite(ptr,size,nmemb,stream);
@@ -35,6 +37,36 @@ inline void ufwrite(const void *ptr, size_t size, size_t nmemb, FILE *stream){
   return;
 }
 
+// OJO to fail cleanly while doing a realloc
+// if we can't realloc we are pretty much screwed anyway but
+// it makes the code clearer to not have a bunch of if (!ptr) { printf("..."); exit(1); };
+inline void * urealloc(void *ptr, size_t size){
+
+  void * dest = realloc(ptr,size);
+  //don't fail if we requested size 0
+  if (dest == NULL && size > 0 )
+    throw std::bad_alloc();
+  return dest;
+
+}
+
+inline void * ucalloc(size_t nmemb, size_t size){
+
+  void * dest = calloc(nmemb,size);
+  //don't fail if we requested size 0
+  if (dest == NULL && nmemb > 0 && size > 0 )
+    throw std::bad_alloc();
+  return dest;
+
+}
+
+inline void * umalloc(size_t size){
+  void * dest = malloc(size);
+  if (dest == NULL && size > 0)
+    throw std::bad_alloc();
+  return dest;
+}
+
 // Save: saves XML tree data structure to file. 
 void XMLTree::Save(unsigned char *filename) 
  {
@@ -73,7 +105,15 @@ void XMLTree::Save(unsigned char *filename)
     // stores the texts   
     if (!disable_tc)
       Text->Save(fp);
-
+    if (!disable_tc){
+    int st = CachedText.size();
+    ufwrite(&st, sizeof(int),1,fp);
+    for (int i = 0; i< CachedText.size(); ++i){
+      st = CachedText.at(i).size();
+      ufwrite(&st, sizeof(int),1,fp);
+      ufwrite(CachedText.at(i).c_str(),sizeof(char),(1+strlen(CachedText.at(i).c_str())),fp);
+    };
+    };
     fclose(fp);
 
  }
@@ -85,66 +125,111 @@ XMLTree *XMLTree::Load(unsigned char *filename, int sample_rate_text)
  {
 
     FILE *fp;
-    char filenameaux[1024];
+    char buffer[1024];
     XMLTree *XML_Tree;
     int i;
-    
+    size_t s_tree = 0;
+    long s_text = 0;
+    size_t s_tags = 0;
+
     // first load the tree topology
-    sprintf(filenameaux, "%s.srx", filename);
-    fp = fopen(filenameaux, "r");
+    sprintf(buffer, "%s.srx", filename);
+    fp = fopen(buffer, "r");
     if (fp == NULL) {
-       printf("Error: cannot open file %s to load the tree structure of XML collection\n", filenameaux);
+       printf("Error: cannot open file %s to load the tree structure of XML collection\n", buffer);
        exit(1);
     } 
 
     XML_Tree = new XMLTree();
 
-    XML_Tree->Par = (bp *)malloc(sizeof(bp));
+    XML_Tree->Par = (bp *)umalloc(sizeof(bp));
 
     loadTree(XML_Tree->Par, fp); 
-    
+
+    s_tree += sizeof(bp);
+
     // stores the table with tag names
     ufread(&XML_Tree->ntagnames, sizeof(int), 1, fp);
-    XML_Tree->TagName = (unsigned char **)malloc(XML_Tree->ntagnames*sizeof(unsigned char *));
+    
+    s_tree += sizeof(int);
+
+    XML_Tree->TagName = (unsigned char **)umalloc(XML_Tree->ntagnames*sizeof(unsigned char *));
+    
+    s_tags += sizeof(unsigned char*)*XML_Tree->ntagnames;
+
 
     for (i=0; i<XML_Tree->ntagnames;i++) {
       
-      // Kim is it needed ?
+      // OJO Kim is it needed ?
       int k = feof(fp);
-       // fscanf chokes ont "\n" which is the case for the root element
-       char * r = fgets(filenameaux,1023,fp);
-       //       int r = fscanf(fp, "<%s>\n",filenameaux);
+
+      
+       // fscanf chokes on "\n" which is the case for the root element
+       char * r = fgets(buffer,1023,fp);
+       //       int r = fscanf(fp, "%s\n",buffer);
        if (r==NULL)
 	 throw "Cannot read tag list";
 
-
-       int len = strlen((const char*)filenameaux);
-       XML_Tree->TagName[i] = (unsigned char *)calloc(len,sizeof(char));
-
-       //XML_Tree->TagName[i] = (unsigned char *)malloc(sizeof(unsigned char)*(strlen((const char *)filenameaux)+1));
-       //the - 1 removes the trailing \n
-       strncpy((char *)XML_Tree->TagName[i], (const char *)filenameaux,len - 1);
+       // strlen is actually the right size, since there is a trailing '\n'
+       int len = strlen((const char*)buffer);
+       XML_Tree->TagName[i] = (unsigned char *)ucalloc(len,sizeof(char));
+       strncpy((char *)XML_Tree->TagName[i], (const char *)buffer,len - 1);
+       s_tags+= len*sizeof(char);
     }
 	
     // loads the flags
+
     ufread(&(XML_Tree->indexing_empty_texts), sizeof(bool), 1, fp);
     ufread(&(XML_Tree->initialized), sizeof(bool), 1, fp);
     ufread(&(XML_Tree->finished), sizeof(bool), 1, fp);
     ufread(&(XML_Tree->disable_tc), sizeof(bool), 1, fp);
-    if (!(XML_Tree->indexing_empty_texts)) XML_Tree->EBVector = static_bitsequence_rrr02::load(fp);
+    
+    s_tree+=sizeof(bool)*4;
 
+    if (!(XML_Tree->indexing_empty_texts)) XML_Tree->EBVector = static_bitsequence_rrr02::load(fp);
+    
+    s_tree+= XML_Tree->EBVector->size();
+    
     // loads the tags
     XML_Tree->Tags = static_sequence::load(fp);
+    s_tree+= XML_Tree->Tags->size();
+
+    s_text = ftell(fp);
 
     // loads the texts
     if (!XML_Tree->disable_tc){
       XML_Tree->Text = TextCollection::InitTextCollection(sample_rate_text);
       XML_Tree->Text->Load(fp,sample_rate_text);
+      int sst;
+      int st;
+      ufread(&sst, sizeof(int),1,fp);
+      for (int i=0;i<sst;i++){
+	ufread(&st, sizeof(int),1,fp);
+	char* str = (char*) malloc(sizeof(char)*st+1);
+	ufread(str,sizeof(char),st+1,fp);
+	string cppstr = str;
+	XML_Tree->CachedText.push_back(cppstr);
+	free(str);
+      };
+
     }
     else
       XML_Tree->Text = NULL;
 
+    s_text = ftell(fp) - s_text;
+
+    
+
+
     fclose(fp);
+
+    std::cerr << "Tree part is " << s_tree/1024 << " Kbytes,\n"
+	      << "with node->tagid part " << XML_Tree->Tags->size()/1024 << "Kbytes \n"
+	      << "size of Tag part : " << XML_Tree->Tags->length () << " elements\n"
+	      << "sizof(unsigned int)* " <<  XML_Tree->Tags->length () << " = " << 
+      sizeof(unsigned int) * XML_Tree->Tags->length () / 1024 << " Kbytes\n"
+	      << "Tag part is " << s_tags/1024 << " Kbytes,\n"
+	      << "Text collection is " << s_text/1024 << " Kbytes \n";
     return XML_Tree;
  }
 
@@ -207,6 +292,9 @@ int XMLTree::SubtreeTags(treeNode x, TagType tag)
        fprintf(stderr, "Error: data structure has not been constructed properly\n");
        exit(1);
     }
+    if (x == Root())
+      x = first_child(Par,x);
+    
 
     int s = x + 2*subtree_size(Par, x) - 1;
  
@@ -297,7 +385,6 @@ int XMLTree::Postorder(treeNode x)
        fprintf(stderr, "Error: data structure has not been constructed properly\n");
        exit(1);
     }
-
     return postorder_rank(Par, x);
  }
 
@@ -308,7 +395,7 @@ TagType XMLTree::Tag(treeNode x)
        fprintf(stderr, "Error: data structure has not been constructed properly\n");
        exit(1);
     }
-
+    
     return Tags->access(node2tagpos(x));
  }
 
@@ -436,6 +523,9 @@ treeNode XMLTree::TaggedDesc(treeNode x, TagType tag)
 
     int r, s;
     treeNode y;
+    if (isleaf(Par,x))
+      return NULLT;
+
     r = (int) Tags->rank(tag, node2tagpos(x));
     s = (int) Tags->select(tag, r+1);
     if (s == -1) return NULLT; // there is no such node
@@ -444,6 +534,28 @@ treeNode XMLTree::TaggedDesc(treeNode x, TagType tag)
     else return y;
  }
 
+// TaggedNext(x,tag): returns the first node tagged tag with larger preorder than x 
+// Returns NULLT if there is none.
+treeNode XMLTree::TaggedNext(treeNode x, TagType tag) 
+ {
+    if (!finished) {
+       fprintf(stderr, "Error: data structure has not been constructed properly\n");
+       exit(1);
+    }
+
+    int r, s;
+    treeNode y;
+    if (x==NULLT)
+      return NULLT;
+
+    r = (int) Tags->rank(tag, node2tagpos(x));
+    s = (int) Tags->select(tag, r+1);
+    if (s == -1) return NULLT; // there is no such node
+    y = tagpos2node(s); // transforms the tag position into a node position  
+    return (y<=x ? NULLT : y);
+ }
+
+
 // TaggedPrec(x,tag): returns the first node tagged tag with smaller preorder than x and not an
 // ancestor of x. Returns NULLT if there is none.
 treeNode XMLTree::TaggedPrec(treeNode x, TagType tag) 
@@ -479,6 +591,9 @@ treeNode XMLTree::TaggedFoll(treeNode x, TagType tag)
     }
 
     int r, s;
+    if (x ==NULLT || x == Root()|| (next_sibling(Par,x) == -1 ))
+      return NULLT;
+
     r = (int) Tags->rank(tag, node2tagpos(next_sibling(Par, x))-1);
     s = (int) Tags->select(tag, r+1);  // select returns -1 in case that there is no r+1-th tag.
     if (s==-1) return NULLT;
@@ -591,10 +706,16 @@ treeNode XMLTree::ParentNode(DocID d)
        fprintf(stderr, "Error: data structure has not been constructed properly\n");
        exit(1);
     }
-
+    
+    if (d == NULLT)
+      return NULLT;
+    
     int s;
+    // OJO : Kim : I added the d+1. before that, else branch was 
+    // EBVector->select1(d)
+    // and gave wrong results (I'm really poking a bear with a stick here).
     if (indexing_empty_texts) s = d;
-    else s = EBVector->select1(d);
+    else s = EBVector->select1(d+1);
     
     if (inspect(Par,s) == CP) // is a closing parenthesis
        return parent(Par, find_open(Par, s));
@@ -619,48 +740,25 @@ int XMLTree::OpenDocument(bool empty_texts, int sample_rate_text,bool dtc)
     
     indexing_empty_texts = empty_texts;
     
-    par_aux = (pb *)malloc(sizeof(pb)*parArraySize);
-    if (!par_aux) {
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;
-    }
+    par_aux = (pb *)umalloc(sizeof(pb)*parArraySize);
     
-    tags_aux = (TagType *) malloc(sizeof(TagType));
-    if (!tags_aux) {
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;
-    }
+    tags_aux = (TagType *) umalloc(sizeof(TagType));
     
-    TagName = (unsigned char **) malloc(2*sizeof(unsigned char*));
-    if (!TagName){
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;
-    }
+    TagName = (unsigned char **) umalloc(2*sizeof(unsigned char*));
 
-    TagName[0] = (unsigned char *) malloc(4*sizeof(unsigned char));
-    strcpy((char *) TagName[0], "<@>");
+    TagName[0] = (unsigned char *) umalloc(4*sizeof(unsigned char));
 
-    if (!TagName[0]){
-      fprintf(stderr, "Error: not enough memory\n");
-      return NULLT;
-    }
+    strcpy((char *) TagName[0], "<@>");
 
-    TagName[1] = (unsigned char *) malloc(4*sizeof(unsigned char));
-    if (!TagName[1]){
-      fprintf(stderr, "Error: not enough memory\n");
-      return NULLT;
-    }
+    TagName[1] = (unsigned char *) umalloc(4*sizeof(unsigned char));
 
     strcpy((char *) TagName[1], "<$>");
 
 
-    if (!indexing_empty_texts) {
-       empty_texts_aux = (unsigned int *)malloc(sizeof(unsigned int));
-       if (!empty_texts_aux) {
-          fprintf(stderr, "Error: not enough memory\n");
-          return NULLT;
-       }
-    }
+    if (!indexing_empty_texts) 
+      empty_texts_aux = (unsigned int *)umalloc(sizeof(unsigned int));
+       
+    
     
     Text = TextCollection::InitTextCollection((unsigned)sample_rate_text);
     
@@ -679,14 +777,10 @@ int XMLTree::CloseDocument()
     }
     
     // closing parenthesis for the tree root
-    par_aux = (pb *)realloc(par_aux, sizeof(pb)*(1+npar/(8*sizeof(pb))));
-    if (!par_aux) {
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;    
-    }
+    par_aux = (pb *)urealloc(par_aux, sizeof(pb)*(1+npar/(8*sizeof(pb))));
     
     // creates the data structure for the tree topology
-    Par = (bp *)malloc(sizeof(bp));
+    Par = (bp *)umalloc(sizeof(bp));
     bp_construct(Par, npar, par_aux, OPT_DEGREE|0);    
     // creates structure for tags
     static_bitsequence_builder * bmb = new static_bitsequence_builder_brw32(20);
@@ -714,7 +808,9 @@ int XMLTree::CloseDocument()
        empty_texts_aux = NULL;
     }
    
+    // OJO was leaked before, found by valgrind
     free(tags_aux);
+
     tags_aux = NULL;
 
     finished = true;
@@ -737,15 +833,10 @@ int XMLTree::NewOpenTag(unsigned char *tagname)
     
     // inserts a new opening parentheses in the bit sequence
     if (sizeof(pb)*8*parArraySize == npar) { // no space left for the new parenthesis
-       par_aux = (pb *)realloc(par_aux, sizeof(pb)*2*parArraySize);
+       par_aux = (pb *)urealloc(par_aux, sizeof(pb)*2*parArraySize);
        parArraySize *= 2;
     }
     
-    if (!par_aux) {
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;    
-    }
-
     setbit(par_aux,npar,OP);  // marks a new opening parenthesis
 
     // transforms the tagname into a tag identifier. If the tag is new, we insert
@@ -759,7 +850,7 @@ int XMLTree::NewOpenTag(unsigned char *tagname)
       found_attributes=true;
 
     if (i==ntagnames) { // the tag is a new one, then we insert it
-       TagName = (unsigned char **)realloc(TagName, sizeof(char *)*(ntagnames+1));
+       TagName = (unsigned char **)urealloc(TagName, sizeof(char *)*(ntagnames+1));
        
        if (!TagName) {
           fprintf(stderr, "Error: not enough memory\n");
@@ -767,19 +858,15 @@ int XMLTree::NewOpenTag(unsigned char *tagname)
        }
        
        ntagnames++;
-       TagName[i] = (unsigned char *)malloc(sizeof(unsigned char)*(strlen((const char *)tagname)+1));
+       TagName[i] = (unsigned char *)umalloc(sizeof(unsigned char)*(strlen((const char *)tagname)+1));
        strcpy((char *)TagName[i], (const char *)tagname);
     } 
-    tags_aux = (TagType *) realloc(tags_aux, sizeof(TagType)*(npar + 1));
-    if (!tags_aux) {
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;
-    }
+    tags_aux = (TagType *) urealloc(tags_aux, sizeof(TagType)*(npar + 1));
 
     tags_aux[npar] = i; // inserts the new tag id within the preorder sequence of tags
     
     npar++;
-
+    
     return 1;
     
  }
@@ -799,14 +886,10 @@ int XMLTree::NewClosingTag(unsigned char *tagname)
     
     // inserts a new closing parentheses in the bit sequence
     if (sizeof(pb)*8*parArraySize == npar) { // no space left for the new parenthesis
-       par_aux = (pb *)realloc(par_aux, sizeof(pb)*2*parArraySize);
+       par_aux = (pb *)urealloc(par_aux, sizeof(pb)*2*parArraySize);
        parArraySize *= 2;
     }
     
-    if (!par_aux) {
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;    
-    }
     setbit(par_aux,npar,CP);  // marks a new closing opening parenthesis
 
     // transforms the tagname into a tag identifier. If the tag is new, we insert
@@ -815,25 +898,16 @@ int XMLTree::NewClosingTag(unsigned char *tagname)
        if (strcmp((const char *)tagname,(const char *)TagName[i])==0) break;
  
     if (i==ntagnames) { // the tag is a new one, then we insert it
-       TagName = (unsigned char **)realloc(TagName, sizeof(char *)*(ntagnames+1));
+       TagName = (unsigned char **)urealloc(TagName, sizeof(char *)*(ntagnames+1));
        
-       if (!TagName) {
-          fprintf(stderr, "Error: not enough memory\n");
-          return NULLT;
-       }
        
        ntagnames++;
-       TagName[i] = (unsigned char *)malloc(sizeof(char)*(strlen((const char *)tagname)+2));
+       TagName[i] = (unsigned char *)umalloc(sizeof(char)*(strlen((const char *)tagname)+2));
        TagName[i][0] = '/';
        strcpy((char *)&(TagName[i][1]), (const char *)tagname);
     } 
 
-    tags_aux = (TagType *)realloc(tags_aux, sizeof(TagType)*(npar + 1));
-
-    if (!tags_aux) {
-       fprintf(stderr, "Error: not enough memory\n");
-       return NULLT;
-    }
+    tags_aux = (TagType *)urealloc(tags_aux, sizeof(TagType)*(npar + 1));
 
     tags_aux[npar] = i; // inserts the new tag id within the preorder sequence of tags
     
@@ -860,16 +934,13 @@ int XMLTree::NewText(unsigned char *s)
     };
 
     if (!indexing_empty_texts) {
-       empty_texts_aux = (unsigned int *)realloc(empty_texts_aux, sizeof(pb)*(1+(npar-1)/(8*sizeof(pb))));
-       if (!empty_texts_aux) {
-          fprintf(stderr, "Error: not enough memory\n");
-          return NULLT;
-       }
-       
-       bitset(empty_texts_aux, npar-1);  // marks the non-empty text with a 1 in the bit vector
+       empty_texts_aux = (unsigned int *)urealloc(empty_texts_aux, sizeof(pb)*(1+(npar-1)/(8*sizeof(pb))));
+              bitset(empty_texts_aux, npar-1);  // marks the non-empty text with a 1 in the bit vector
     }
     
     Text->InsertText(s);
+    string cpps = (char*) s;
+    CachedText.push_back(cpps); 
     
     return 1; // success
  }
@@ -888,11 +959,7 @@ int XMLTree::NewEmptyText()
     }
 
     if (!indexing_empty_texts) {
-       empty_texts_aux = (unsigned int *)realloc(empty_texts_aux, sizeof(pb)*(1+(npar-1)/(8*sizeof(pb))));
-       if (!empty_texts_aux) {
-          fprintf(stderr, "Error: not enough memory\n");
-          return NULLT;
-       }
+       empty_texts_aux = (unsigned int *)urealloc(empty_texts_aux, sizeof(pb)*(1+(npar-1)/(8*sizeof(pb))));
        
        bitclean(empty_texts_aux, npar-1);  // marks the empty text with a 0 in the bit vector
     }
@@ -922,23 +989,33 @@ unsigned char *XMLTree::GetTagName(TagType tagid)
     unsigned char *s;
 
     if (tagid >= ntagnames) return NULL; // invalid tag identifier
-    s = (unsigned char *)malloc((strlen((const char *)TagName[tagid])+1)*sizeof(unsigned char));
+    s = (unsigned char *)umalloc((strlen((const char *)TagName[tagid])+1)*sizeof(unsigned char));
     strcpy((char *)s, (const char *)TagName[tagid]);
     return s;
  }
 
 
+//KIM : OJO need the two following methods
+
+const unsigned char *XMLTree::GetTagNameByRef(TagType tagid)
+ {
+    if (tagid >= ntagnames) return NULL; // invalid tag identifier
+    return ((const unsigned char*)  TagName[tagid]);
+ }
+
+
+
 TagType XMLTree::RegisterTag(unsigned char *tagname)
 {
   if (!finished)
     return NULLT;
   
-
   TagType id = XMLTree::GetTagId(tagname);
   if (id == NULLT){
     id = ntagnames;
     ntagnames = ntagnames + 1;    
-    TagName = (unsigned char **) realloc(TagName,ntagnames*(sizeof(unsigned char*)));
+    TagName = (unsigned char **) urealloc(TagName,ntagnames*(sizeof(unsigned char*)));
+    TagName[id] = (unsigned char *) umalloc(sizeof(unsigned char)*strlen( (const char*) tagname)+1);
     strcpy((char*)TagName[id], (const char *)tagname);  
   };