Changeset 1279 for branches/New_Config_Format-branch/gsdl/src/mgpp/text/invf.h

Timestamp:

2000-07-13T10:21:53+12:00 (24 years ago)

Author:

sjboddie

Message:

merged changes to trunk into New_Config_Format branch

File:

• branches/New_Config_Format-branch/gsdl/src/mgpp/text/invf.h (modified) (6 diffs)

branches/New_Config_Format-branch/gsdl/src/mgpp/text/invf.h

@@ -31 +31 @@
 
 // NOTE: This does not include the magic number
+// header info for .invf.dict file
 struct invf_dict_header {
   unsigned long lookback;
@@ -91 +92 @@
 // this version of the blocked dictionary uses a fixed number
 // of entries per block, not a fixed block size
+// info for .invf.dict.blocked file
+// blocked dict has a heap of blocks, some for words, some for tags
+// and an index into each set of blocks. The index has pointers to
+// the first entry in each block. Can do a binary search on the index
+// to find out which block an elemnet is in
 struct block_dict_header : public invf_dict_header {
   // note: word_dict_start and tag_dict_start are undefined
   // for blocked dictionaries
 
-  unsigned long entries_per_wblk;
+  unsigned long entries_per_wblk; // word blocks
   unsigned long num_wblks;
   unsigned long max_wblk_size;
@@ -101 +107 @@
   unsigned long wblk_idx_start;
   
-  unsigned long entries_per_tblk;
+  unsigned long entries_per_tblk; // tag blocks
   unsigned long num_tblks;
   unsigned long max_tblk_size;
@@ -117 +123 @@
 struct block_dict_el {
   UCArray el; // word or tag
-  unsigned long frag_occur; // # entries in invf file
-  unsigned long freq;
-  unsigned long invf_ptr;
+  unsigned long frag_occur; // # entries in invf file - if have a
+  // word level index, this is the same as freq, otherwise, its the number
+  // of fragments containing this word
+  unsigned long freq; // # of times this word occurs
+  unsigned long invf_ptr; // pointer into inverted file
 
   virtual void Clear ();
@@ -133 +141 @@
 
 struct word_block_dict_el : public block_dict_el {
-  unsigned long *levelFreqs;
+  unsigned long *levelFreqs; // freq of the word at each level
 
   void Clear ();
@@ -209 +217 @@
 #define SKIP_MODE_NO_SKIPS 0
 
+// invf file - has a list of frags for each word, but the word is not 
+//  stored in the invf file - the dictionaries store the words, along
+// with num entries, and a pointer into invf file
 struct invf_file_header {
   unsigned long no_of_words;