source: trunk/gsdl/src/colservr/mgsearch.cpp@ 539

/**********************************************************************
 *
 * mgsearch.cpp -- 
 * Copyright (C) 1999  The New Zealand Digital Library Project
 *
 * A component of the Greenstone digital library software
 * from the New Zealand Digital Library Project at the
 * University of Waikato, New Zealand.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 *
 * $Id: mgsearch.cpp 539 1999-09-07 22:52:52Z rjmcnab $
 *
 *********************************************************************/

/*
   $Log$
   Revision 1.19  1999/09/07 22:52:52  rjmcnab
   Seems to be an error in mg for retrieving documents using a paragraph
   based index for some cases. Just added a work around (loads the default
   index every time).

   Revision 1.18  1999/09/07 04:57:22  sjboddie
   added gpl notice

   Revision 1.17  1999/08/31 22:42:41  rjmcnab
   A couple of minor things.

   Revision 1.16  1999/08/25 04:51:06  sjboddie
   small change to allow for searching using boolean operators

   Revision 1.15  1999/07/16 08:35:03  rjmcnab
   Fixed a weird bug to do with a faulty case statement.

   Revision 1.14  1999/07/16 03:42:22  sjboddie
   changed isApprox

   Revision 1.13  1999/07/16 00:12:46  sjboddie
   removed all the old post-processing stuff

   Revision 1.12  1999/07/07 06:17:47  rjmcnab
   broke search_index into index+subcollection+language
   within mgsearch

   Revision 1.11  1999/07/05 21:06:43  rjmcnab
   Disabled quoted strings.

   Revision 1.10  1999/07/01 09:29:19  rjmcnab
   Changes for better reporting of number documents which match a query. Changes
   should still work as before with older versions of mg.

   Revision 1.9  1999/07/01 03:54:48  rjmcnab
   Added code to plug in the equivalent terms of each of the query terms.
   Also added a function to get a raw utf8 encoded mg document (for speeding
   up a phrase matching function)

   Revision 1.8  1999/06/30 04:04:12  rjmcnab
   made stemming functions available from mgsearch and made the stems
   for the query terms available in queryinfo

   Revision 1.7  1999/06/27 22:07:27  sjboddie
   got rid of all the old functions for dealing with dir indexes

   Revision 1.6  1999/06/09 00:41:32  sjboddie
   phrase searching now uses case-folding if it's turned on

   Revision 1.5  1999/02/21 22:31:35  rjmcnab

   Removed locateinfo.

   Revision 1.4  1999/02/03 01:13:27  sjboddie

   Got interface to handle subcollections and language subcollections -
   committed changes made to some of the collections

   Revision 1.3  1999/01/19 01:38:17  rjmcnab

   Made the source more portable.

   Revision 1.2  1999/01/12 01:51:02  rjmcnab

   Standard header.

   Revision 1.1  1999/01/08 09:02:16  rjmcnab

   Moved from src/library.

 */


#include "gsdlconf.h"
#include "mgsearch.h"
#include "fileutil.h"

#include <string.h>
#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>

#if defined(GSDL_USE_OBJECTSPACE)
#  include <ospace\std\iostream>
#elif defined(GSDL_USE_IOS_H)
#  include <iostream.h>
#else
#  include <iostream>
#endif

#if defined(__WIN32__)
// gdbm stuff
#  include "autoconf.h"
#  include "systems.h"
#  include "gdbmconst.h"
#  include "gdbm.h"
#else
#  include <gdbm.h>
#endif

  
#include <assert.h>

#include "mgq.h"
// #include "locateinfo.h"
#include "gsdlunicode.h"
#include "unitool.h"


/////////////
// globals //
/////////////

static char *tempdoc = NULL;
static int templen = 0;


//////////////////////
// useful functions //
//////////////////////


// input and output are in utf8
text_t mgsearch_stemword (const text_t &word) {
  // allocate working stem space
  int maxstemlen = mgq_getmaxstemlen ();
  unsigned char *word_stem = new unsigned char [maxstemlen + 2];
  if (word_stem == NULL) return "";

  // copy word to word_stem
  int len = 0;
  text_t::const_iterator here = word.begin();
  text_t::const_iterator end = word.end();
  while (len < maxstemlen && here != end) {
    word_stem[len+1] = (unsigned char)(*here);
    len++; here++;
  }
  word_stem[len+1] = '\0';
  word_stem[0] = len;

  mgq_stemword (word_stem);

  // copy word_stem back to tempstr
  text_t tempstr;
  tempstr.setcarr((char *)(&word_stem[1]), word_stem[0]);

  delete [] word_stem;
  
  return tempstr;
}



////////////////////////
// callback functions //
////////////////////////

// This routine is called for each document found in a search
// it assumes that cache_num is set up correctly to point to
// a suitable result cache
int ourquerycallback(char * /*UDoc*/, int /*ULen*/, int DocNum, 
             float Weight, void *info) {

  
  queryresultsclass *queryresults = (queryresultsclass * )info;

  // append this entry to the document results
  docresultclass docresult;
  docresult.docnum = DocNum;
  docresult.num_query_terms_matched = (int)(Weight/100.0); // will always be 0 on some versions of mg...
  docresult.docweight = Weight - docresult.num_query_terms_matched*100;
  
  queryresults->docs.docset[DocNum] = docresult;
  queryresults->docs.docorder.push_back(DocNum);
  
  return 0;
}

int termequivcallback(char *Word, int ULen,  int /*Freq*/, 
              float /*Weight*/,  void *info) {
  text_tset *equivterms = (text_tset *)info;
  if (equivterms == NULL) return 0;

  text_t thisterm;
  thisterm.setcarr(Word, ULen);

  equivterms->insert(thisterm);
  
  return 0;
}


void mgsearch_equivterms (const text_t &word, text_tset &equivterms) {
  // allocate working stem space
  int maxstemlen = mgq_getmaxstemlen ();
  unsigned char *word_stem = new unsigned char [maxstemlen + 2];
  if (word_stem == NULL) return;

  // copy word to word_stem
  int len = 0;
  text_t::const_iterator here = word.begin();
  text_t::const_iterator end = word.end();
  while (len < maxstemlen && here != end) {
    word_stem[len+1] = (unsigned char)(*here);
    len++; here++;
  }
  word_stem[len+1] = '\0';
  word_stem[0] = len;

  // get the equivalent terms
  mgq_equivterms (word_stem, termequivcallback, (void *)(&equivterms));
  
  delete [] word_stem;

  return;
}

  text_tset utf8equivterms; // kept as utf8 string for fast matching


// This callback is called once for each term in the query
int termfreqcallback(char *Word, int ULen,  int Freq, 
             float /*Weight*/,  void *info) {
  queryresultsclass *queryresults = (queryresultsclass *)info;
  if (queryresults == NULL) return 0;

  text_t term;
  term.setcarr(Word, ULen);
  termfreqclass termfreq;

  termfreq.termstr = to_uni(term);
  text_t utf8termstem = mgsearch_stemword (term);
  termfreq.termstemstr = to_uni (utf8termstem);

  mgsearch_equivterms (utf8termstem, termfreq.utf8equivterms);
  
  termfreq.termfreq = Freq;
  queryresults->orgterms.push_back(termfreq);
  
  return 0;
}

// this callback is called once for each variation of each term
int termvariantscallback(char *Word, int ULen, int /*Freq*/,
             float /*Weight*/, void *info) {

  text_t term;
  term.setcarr(Word, ULen);
  queryresultsclass *queryresults = (queryresultsclass *)info;
  queryresults->termvariants.insert(to_uni(term));

  return 0;
}

// This callback is for getting document text
int doctextcallback(char *Doc, int ULen,  int /*Freq*/, 
            float /*Weight*/,  void * /*info*/) {
  tempdoc = Doc;
  templen = ULen;
  
  return 0;
}


static text_t getindexsuffix (const text_t &collection, 
                  const text_t &index) {

  text_t indexsuffix = "index";  
  indexsuffix = filename_cat (indexsuffix, index);
  indexsuffix = filename_cat (indexsuffix, collection);
  return indexsuffix;
}




////////////////////
// mgsearch class //
////////////////////

mgsearchclass::mgsearchclass ()
{
  cache = new querycache (RESULTCACHESIZE);
}

mgsearchclass::~mgsearchclass () 
{
  if (cache != NULL) 
    {
      delete cache;
      cache = NULL;
    }
}


void mgsearchclass::setcollectdir (const text_t &thecollectdir)
{
  collectdir = thecollectdir;
}

// you only need to use this function before doing any stemming
// casefolding and stemming will be set if values for them are
// provided (0 or 1).
// makeindexcurrent returns true if it was able to load the database
bool mgsearchclass::makeindexcurrent (const text_t &index,
                      const text_t &subcollection,
                      const text_t &language,
                      const text_t &collection,
                      int casefolding,
                      int stemming) {
  bool databaseloaded = true;

  // get the names of the collection, index and text suffixes
  char *ccollection = collection.getcstr();
  assert (ccollection != NULL);
  char *idxsuffix = (getindexsuffix (collection, (index+subcollection+language))).getcstr();
  assert (idxsuffix != NULL);
  char *txtsuffix = (getindexsuffix (collection, "text")).getcstr();
  assert (txtsuffix != NULL);

#ifdef __WIN32__
  char *ccollectdir = (collectdir+"\\").getcstr(); assert (ccollectdir != NULL);
#else
  char *ccollectdir = collectdir.getcstr(); assert (ccollectdir != NULL);
#endif

  if (load_database(ccollection, ccollectdir, idxsuffix, txtsuffix)) {
    if (casefolding == 0) mgq_ask(".set casefold off");
    else if (casefolding > 0) mgq_ask(".set casefold on");
    if (stemming == 0) mgq_ask(".set stem off");
    else if (stemming > 0) mgq_ask(".set stem on");
    
  } else databaseloaded = false;

  // free up the c strings
  delete ccollection;
  delete idxsuffix;
  delete txtsuffix;
  delete ccollectdir;

  return databaseloaded;
}


// stem word uses the values set in the last call to makeindexcurrent
// to stem the word. It is assumed that word is in unicode
text_t mgsearchclass::stemword (const text_t &word) {
  return to_uni (mgsearch_stemword (to_utf8 (word)));
}

text_t mgsearchclass::stemword (text_t::const_iterator here, text_t::const_iterator end) {
  return to_uni (mgsearch_stemword (to_utf8 (here, end)));
}


bool mgsearchclass::search(const queryparamclass &queryparams, 
               queryresultsclass &queryresults) {
  assert (cache != NULL);

  queryresults.clear();

  // first check the cache
  if (cache->find(queryparams, queryresults)) return true;

  // make sure there is a query to be processed
  text_t::const_iterator queryhere = queryparams.querystring.begin();
  text_t::const_iterator queryend = queryparams.querystring.end();
  while (queryhere != queryend) {
    if (is_unicode_letdig (*queryhere)) break;
    queryhere++;
  }

  // if we reached the end of the query string without finding
  // any alphanumeric characters then return no results (and say
  // the database was loaded)
  if (queryhere == queryend) return true;

  if (makeindexcurrent (queryparams.index, queryparams.subcollection,
            queryparams.language, queryparams.collection)) {
    setsearchmode (queryparams);
    submitquery (queryparams);
    getresults (queryparams, queryresults);
    return true;
  }

  return false;
}


void mgsearchclass::setsearchmode (const queryparamclass &queryparams)
{
  mgq_ask(".set expert true");
  mgq_ask(".set sorted_terms true");
  mgq_ask(".set accumulator_method list");
  mgq_ask(".set max_accumulators 500000");
  mgq_ask(".set maxparas 500000");
  mgq_ask(".set verbatim true");
  mgq_ask(".unset skip_dump");
  mgq_ask(".set mode docnums");

  switch (queryparams.search_type) 
    {
    case 0: mgq_ask(".set query boolean");  break;
    case 1:  mgq_ask(".set query ranked"); break;
    }
  switch (queryparams.casefolding) 
    {
    case 1: mgq_ask(".set casefold on");  break;
    case 0: mgq_ask(".set casefold off"); break;
    }
  switch (queryparams.stemming) 
    {
    case 1: mgq_ask(".set stem on");  break;
    case 0: mgq_ask(".set stem off"); break;
    }
  mgq_ask(".set heads_length 150");
  
  if (queryparams.maxdocs == -1) {
    mgq_ask(".set maxdocs all");
  } else {
    char maxdocstr[32];
    sprintf(maxdocstr, ".set maxdocs %i", queryparams.maxdocs);
    mgq_ask(maxdocstr);
  }
}


void mgsearchclass::submitquery (const queryparamclass &queryparams)
{
  // sort out the query string
  text_t ttquerystring = queryparams.querystring;
  filterquery (ttquerystring);
  char *querystring = to_utf8(ttquerystring).getcstr();
  
  // submit the query
  mgq_ask(querystring);

  delete querystring;
}


void mgsearchclass::getresults (const queryparamclass &queryparams,
                queryresultsclass &queryresults) {

  mgq_results(result_docnums, 0, MAXNUMDOCS, 
          ourquerycallback, (void *)(&queryresults));
  
  // get the term frequencies
  mgq_results(result_termfreqs, 0, MAXNUMTERMS, 
          termfreqcallback, (void *)(&queryresults));
  queryresults.sortuniqqueryterms();

  // get term variants 
  mgq_results(result_terms, 0, MAXNUMTERMS,
              termvariantscallback, (void *)(&queryresults));

  // get the number of documents retrieved
  int total_retrieved = 0, is_approx = 0;
  mgq_docsretrieved (&total_retrieved, &is_approx);

  if (total_retrieved == 0) {
    // not available (or really was zero)
    queryresults.docs_matched = queryresults.docs.docset.size();
    if (queryresults.docs_matched < queryparams.maxdocs)
      queryresults.is_approx = Exact;
    else
      queryresults.is_approx = MoreThan;
  } else {
    queryresults.docs_matched = total_retrieved;
    if (is_approx) queryresults.is_approx = Approximate;
    else queryresults.is_approx = Exact;
  }
}

void mgsearchclass::filterquery (text_t &ttquerystring) {
  text_t::iterator ithere = ttquerystring.begin ();
  text_t::iterator itend = ttquerystring.end ();
  
  // remove all non alphanumeric characters (except
  // boolean operators
  while (ithere != itend) {
    if ((!is_unicode_letdig(*ithere)) && (*ithere != '!') &&
    (*ithere != '&') && (*ithere != '|') && (*ithere != '(') &&
    (*ithere != ')')) (*ithere) = ' ';
    ithere++;
  }
}


// the document text for 'docnum' is placed in 'output'
// docTargetDocument returns 'true' if it was able to
// try to get a document
// collection is needed to see if an index from the 
// collection is loaded. If no index has been loaded
// defaultindex is needed to load one
bool mgsearchclass::docTargetDocument(const text_t &defaultindex,
                      const text_t &defaultsubcollection,
                      const text_t &defaultlanguage,
                      const text_t &collection,
                      int docnum,
                      text_t &output) {
  output.clear();

  // get the mg version of the document
  char *mgdoc = NULL;
  int doclen = 0;
  if (!mgdocument (defaultindex, defaultsubcollection, defaultlanguage,
           collection, docnum, mgdoc, doclen)) return false;
  if (mgdoc == NULL) return false;

  // replace all control-Cs with spaces
  char *mgdoc_here = mgdoc;
  char *mgdoc_end = mgdoc + doclen;
  while (mgdoc_here < mgdoc_end) {
    if (*mgdoc_here == '\x3') *mgdoc_here = ' ';
    mgdoc_here++;
  }

  // convert this document to unicode
  utf8inconvertclass inconvert;
  convertclass::status_t status;
  inconvert.reset ();
  inconvert.setinput (mgdoc, doclen);
  inconvert.convert (output, status);

  return true;
}


bool mgsearchclass::mgdocument (const text_t &defaultindex, 
                const text_t &defaultsubcollection,
                const text_t &defaultlanguage,
                const text_t &collection,
                int docnum,
                char *&UDoc, int &ULen) {
  int databaseloaded = 0;

  UDoc = NULL; ULen = 0;
  
  // see if we can make an appropriate database current
//    char *ccollection = collection.getcstr();
//    assert (ccollection != NULL);
//    databaseloaded = load_text_database (ccollection);
//    delete ccollection;
  
  // try and load the database
//    if (!databaseloaded)
  databaseloaded = makeindexcurrent (defaultindex, defaultsubcollection,
                     defaultlanguage, collection);
  
  if (databaseloaded) {
    // retrieve the document from mg
    char docstr[32];
    sprintf(docstr, "%i", docnum);
    
    mgq_ask(".set mode text");
    mgq_ask(".set query docnums");
    mgq_ask(docstr);

    tempdoc = NULL;
    templen = 0;
    mgq_results (result_docs, 0, 1, doctextcallback, (void *)NULL);
    UDoc = tempdoc;
    ULen = templen;
  }

  return (bool)databaseloaded;
}