source: gsdl/trunk/runtime-src/src/recpt/formattools.cpp@ 19058

/**********************************************************************
 *
 * formattools.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.
 *
 *********************************************************************/

#include "formattools.h"
#include "cgiutils.h"
#include "recptprototools.h"
#include "OIDtools.h"
#include "summarise.h"

#include <assert.h>

// a few function prototypes

static text_t format_string (const text_t& collection, recptproto* collectproto,
                 ResultDocInfo_t &docinfo, displayclass &disp, 
                 format_t *formatlistptr, text_tmap &options,
                 ostream& logout);

static bool parse_action (text_t::const_iterator &here, const text_t::const_iterator &end, 
              format_t *formatlistptr, text_tset &metadata, bool &getParents);

static text_t format_summary (const text_t& collection, recptproto* collectproto,
                  ResultDocInfo_t &docinfo, displayclass &disp, 
                  text_tmap &options, ostream& logout);
static text_t format_text (const text_t& collection, recptproto* collectproto,
                  ResultDocInfo_t &docinfo, displayclass &disp, 
                  text_tmap &options, ostream& logout);

static text_t expand_metadata(const text_t &metavalue, const text_t& collection,
                  recptproto* collectproto, ResultDocInfo_t &docinfo,
                  displayclass &disp, text_tmap &options,
                  ostream &logout);


void metadata_t::clear() {
  metaname.clear();
  metacommand = mNone;
  mqualifier.parent  = pNone;
  mqualifier.sibling = sNone;
  mqualifier.child   = cNone;
  pre_tree_traverse.clear();
  parentoptions.clear();
  siblingoptions.clear();
  childoptions.clear();
}

void decision_t::clear() {
  command = dMeta;
  meta.clear();
  text.clear();
}

void format_t::clear() {
  command = comText;
  decision.clear();
  text.clear();
  meta.clear();
  nextptr = NULL;
  ifptr = NULL;
  elseptr = NULL;
  orptr = NULL;
}

void formatinfo_t::clear() {
  DocumentImages = false;
  DocumentTitles = true;
  DocumentHeading = "{Or}{[parent(Top):Title],[Title],untitled}<br>";
  DocumentContents = true;
  DocumentArrowsBottom = true; 
  DocumentArrowsTop = false; 
  DocumentSearchResultLinks = false;
  DocumentButtons.erase (DocumentButtons.begin(), DocumentButtons.end());
  //  DocumentButtons.push_back ("Expand Text");
  //  DocumentButtons.push_back ("Expand Contents");
  DocumentButtons.push_back ("Detach");
  DocumentButtons.push_back ("Highlight");
  RelatedDocuments = "";
  DocumentText = "[Text]";
  formatstrings.erase (formatstrings.begin(), formatstrings.end());
  DocumentUseHTML = false;
  AllowExtendedOptions = false;
}

// simply checks to see if formatstring begins with a <td> tag
bool is_table_content (const text_t &formatstring) {
  text_t::const_iterator here = formatstring.begin();
  text_t::const_iterator end = formatstring.end();
  
  while (here != end) {
    if (*here != ' ') {
      if ((*here == '<') && ((here+3) < end)) {
    if ((*(here+1) == 't' || *(here+1) == 'T') &&
        (*(here+2) == 'd' || *(here+2) == 'D') &&
        (*(here+3) == '>' || *(here+3) == ' '))
      return true;
      } else return false;
    }
    ++here;
  }
  return false;
}

bool is_table_content (const format_t *formatlistptr) {

  if (formatlistptr == NULL) return false;
  
  if (formatlistptr->command == comText)
    return is_table_content (formatlistptr->text);
    
  return false;
}

// returns false if key isn't in formatstringmap
bool get_formatstring (const text_t &key, const text_tmap &formatstringmap,
               text_t &formatstring) {

  formatstring.clear();
  text_tmap::const_iterator it = formatstringmap.find(key);
  if (it == formatstringmap.end()) return false;
  formatstring = (*it).second;
  return true;
}

// tries to find "key1key2" then "key1" then "key2"
bool get_formatstring (const text_t &key1, const text_t &key2,  
               const text_tmap &formatstringmap,
               text_t &formatstring) {

  formatstring.clear();
  text_tmap::const_iterator it = formatstringmap.find(key1 + key2);
  if (it != formatstringmap.end()) {
    formatstring = (*it).second;
    return true;
  }
  it = formatstringmap.find(key1);
  if (it != formatstringmap.end()) {
    formatstring = (*it).second;
    return true;
  }
  it = formatstringmap.find(key2);
  if (it != formatstringmap.end()) {
    formatstring = (*it).second;
    return true;
  }
  return false;
}


text_t remove_namespace(const text_t &meta_name) {
  text_t::const_iterator end = meta_name.end();
  text_t::const_iterator it = findchar(meta_name.begin(), end, '.');
  if (it != end) {
    return substr(it+1, end);
  }

  return meta_name;

}
// returns a date of form _format:date_(year, month, day)
// input is date of type yyyy-?mm-?dd
// at least the year must be present in date
text_t format_date (const text_t &date) {

  if (date.size() < 4) return "";

  text_t::const_iterator datebegin = date.begin();

  text_t year = substr (datebegin, datebegin+4);
  int chars_seen_so_far = 4;

  if (date[chars_seen_so_far] == '-') ++chars_seen_so_far ;
  if (date.size() < chars_seen_so_far+2) return "_format:date_("+year+")";
 
  text_t month = substr (datebegin+chars_seen_so_far, datebegin+chars_seen_so_far+2);
  int imonth = month.getint();
  if (imonth <= 0 || imonth > 12) return "_format:date_("+year+")";
 
  chars_seen_so_far += 2;
  if (date[chars_seen_so_far] == '-') ++chars_seen_so_far ;
  
  if (date.size() < chars_seen_so_far+2) return "_format:date_("+year+","+month+")";

  text_t day = substr (datebegin+chars_seen_so_far, datebegin+chars_seen_so_far+2);
  if (day[0] == '0') day = substr (day.begin()+1, day.end());
  int iday = day.getint();
  if (iday <= 0 || iday > 31) return "_format:date_("+year+","+month+")";
    
  return "_format:date_("+year+","+month+","+day+")";
}  

// converts an iso639 language code to its English equivalent
// should we be checking that the macro exists??
text_t iso639 (const text_t &langcode) {
  if (langcode.empty()) return "";
  return "_iso639:iso639"+langcode+"_";
}


text_t get_href (const text_t &link) {

  text_t href;

  text_t::const_iterator here = findchar(link.begin(), link.end(), '"');
  text_t::const_iterator end = link.end();
  if (here == end) return g_EmptyText;
  
  ++here;
  while (here != end) {
    if (*here == '"') break;
    href.push_back(*here);
    ++here;
  }

  return href;
}

//this function gets the information associated with the relation
//metadata for the document associated with 'docinfo'. This relation
//metadata consists of a line of pairs containing 'collection, document OID'
//(this is the OID of the document related to the current document, and
//the collection the related document belongs to). For each of these pairs
//the title metadata is obtained and then an html link between the title
//of the related doc and the document's position (the document will be
//found in "<a href=\"_httpdocument_&c=collection&cl=search&d=OID">
//(where collection is the related documents collection, and OID is the
//related documents OID).  A list of these html links are made for as many
//related documents as there are. This list is then returned. If there are
//no related documents available for the current document then the string
//'.. no related documents .. ' is returned.
text_t get_related_docs(const text_t& collection, recptproto* collectproto,
            ResultDocInfo_t &docinfo, ostream& logout){
  
  text_tset metadata;

  //insert the metadata we wish to collect
  metadata.insert("dc.Relation"); 
  metadata.insert("Title");  
  metadata.insert("Subject"); //for emails, where title data doesn't apply
 
  FilterResponse_t response;
  text_t relation = ""; //string for displaying relation metadata
  text_t relationTitle = ""; //the related documents Title (or subject)
  text_t relationOID = ""; //the related documents OID  

  //get the information associated with the metadata for current doc
  if (get_info (docinfo.OID, collection, "", metadata, 
        false, collectproto, response, logout)) {
    
    //if the relation metadata exists, store for displaying
    if(!response.docInfo[0].metadata["dc.Relation"].values.empty()){
      relationOID += response.docInfo[0].metadata["dc.Relation"].values[0];

      //split relation data into pairs of collectionname,ID number
      text_tarray relationpairs;
      splitchar (relationOID.begin(), relationOID.end(), ' ', relationpairs);
      
      text_tarray::const_iterator currDoc = relationpairs.begin();  
      text_tarray::const_iterator lastDoc = relationpairs.end();

      //iterate through the pairs to split and display
      while(currDoc != lastDoc){
    
    //split pairs into collectionname and ID 
    text_tarray relationdata;
    splitchar ((*currDoc).begin(), (*currDoc).end(), ',', relationdata);
    
    //get first element in the array (collection)
    text_tarray::const_iterator doc_data = relationdata.begin();
    text_t document_collection = *doc_data;
    ++doc_data; //increment to get next item in array (oid)
    text_t document_OID = *doc_data;
    
    //create html link to related document
    relation += "<a href=\"_httpdocument_&c=" + document_collection;
    relation += "&cl=search&d=" + document_OID;
       
    //get the information associated with the metadata for related doc
    if (get_info (document_OID, document_collection, "", metadata,
              false, collectproto, response, logout)) {
      
      //if title metadata doesn't exist, collect subject metadata
      //if that doesn't exist, just call it 'related document'
      if (!response.docInfo[0].metadata["Title"].values[0].empty())
        relationTitle = response.docInfo[0].metadata["Title"].values[0];
      else if (!response.docInfo[0].metadata["Subject"].values.empty())
        relationTitle = response.docInfo[0].metadata["Subject"].values[0];
      else relationTitle =  "RELATED DOCUMENT";
      
    }
    
    //link the related document's title to its page
    relation += "\">" + relationTitle + "</a>"; 
    relation += "  (" + document_collection + ")<br>";
    
    ++currDoc;
      }
    }
    
  } 

  if(relation.empty()) //no relation data for documnet
    relation = ".. no related documents .. ";

  return relation;
}



static void get_parent_options (text_t &instring, metadata_t &metaoption) {

  assert (instring.size() > 7);
  if (instring.size() <= 7) return;

  text_t meta, com, op;
  bool inbraces = false;
  bool inquotes = false;
  bool foundcolon = false;
  text_t::const_iterator here = instring.begin()+6;
  text_t::const_iterator end = instring.end();
  while (here != end) {
    if (foundcolon) meta.push_back (*here);
    else if (*here == '(') inbraces = true;
    else if (*here == ')') inbraces = false;
    else if (*here == '\'' && !inquotes) inquotes = true;
    else if (*here == '\'' && inquotes) inquotes = false;
    else if (*here == ':' && !inbraces) foundcolon = true;
    else if (inquotes) op.push_back (*here);
    else com.push_back (*here);
    ++here;
  }

  instring = meta;
  if (com.empty())
    metaoption.mqualifier.parent = pImmediate;
  else if (com == "Top")
    metaoption.mqualifier.parent = pTop;
  else if (com == "All") {
    metaoption.mqualifier.parent = pAll;
    metaoption.parentoptions = op;
  }
}


static void get_sibling_options (text_t &instring, metadata_t &metaoption) {

  assert (instring.size() > 8);
  if (instring.size() <= 8) return;
  text_t meta, com, op;
  bool inbraces = false;
  bool inquotes = false;
  bool foundcolon = false;
  text_t::const_iterator here = instring.begin()+7;
  text_t::const_iterator end = instring.end();
  while (here != end) {
    if (foundcolon) meta.push_back (*here);
    else if (*here == '(') inbraces = true;
    else if (*here == ')') inbraces = false;
    else if (*here == '\'' && !inquotes) inquotes = true;
    else if (*here == '\'' && inquotes) inquotes = false;
    else if (*here == ':' && !inbraces) foundcolon = true;    
    else if (inquotes) op.push_back (*here);
    else com.push_back (*here);
    ++here;
  }

  instring = meta;
  metaoption.siblingoptions.clear();

  if (com.empty()) {
    metaoption.mqualifier.sibling = sAll;
    metaoption.siblingoptions = " ";
  }
  else if (com == "first") {
    metaoption.mqualifier.sibling = sNum;
    metaoption.siblingoptions = "0";
  }
  else if (com == "last") {
    metaoption.mqualifier.sibling = sNum;
    metaoption.siblingoptions = "-2"; // == last
  }
  else if (com.getint()>0) {
    metaoption.mqualifier.sibling = sNum;
    int pos = com.getint()-1;
    metaoption.siblingoptions +=pos;
  }
  else {
    metaoption.mqualifier.sibling = sAll;
    metaoption.siblingoptions = op;
  }
}

static void get_child_options (text_t &instring, metadata_t &metaoption) {

  assert (instring.size() > 6);
  if (instring.size() <= 6) return;
  text_t meta, com, op;
  bool inbraces = false;
  bool inquotes = false;
  bool foundcolon = false;
  text_t::const_iterator here = instring.begin()+5;
  text_t::const_iterator end = instring.end();
  while (here != end) {
    if (foundcolon) meta.push_back (*here);
    else if (*here == '(') inbraces = true;
    else if (*here == ')') inbraces = false;
    else if (*here == '\'' && !inquotes) inquotes = true;
    else if (*here == '\'' && inquotes) inquotes = false;
    else if (*here == ':' && !inbraces) foundcolon = true;
    else if (inquotes) op.push_back (*here);
    else com.push_back (*here);
    ++here;
  }

  instring = meta;
  if (com.empty()) {
    metaoption.mqualifier.child = cAll;
    metaoption.childoptions = " ";
  }
  else if (com == "first") {
    metaoption.mqualifier.child = cNum;
    metaoption.childoptions = ".fc";
  }
  else if (com == "last") {
    metaoption.mqualifier.child = cNum;
    metaoption.childoptions = ".lc";
  }
  else if (com.getint()>0) {
    metaoption.mqualifier.child = cNum;
    metaoption.childoptions = "."+com;
  }
  else {
    metaoption.mqualifier.child = cAll;
    metaoption.childoptions = op;
  }
}



static void parse_meta (text_t &meta, metadata_t &metaoption,
            text_tset &metadata, bool &getParents) {

  if (meta.size() > 8 && (substr(meta.begin(), meta.begin()+8) == "cgisafe:")) {
    metaoption.metacommand |= mCgiSafe;
    meta = substr (meta.begin()+8, meta.end());
  }
  if (meta.size() > 7 && (substr(meta.begin(), meta.begin()+7) == "format:")) {    
    metaoption.metacommand |= mSpecial;
    meta = substr (meta.begin()+7, meta.end());
  }

  bool had_parent_or_child = true;
  bool prev_was_parent = false;
  bool prev_was_child  = false;

  while (had_parent_or_child) {
    if (meta.size() > 7 
    && (substr (meta.begin(), meta.begin()+6) == "parent")) {

      // clear out sibling and child (cmd and options)
      metaoption.metacommand &= ~(mChild|mSibling);
      metaoption.childoptions.clear();
      metaoption.siblingoptions.clear();

      getParents = true;
      metaoption.metacommand |= mParent;
      get_parent_options (meta, metaoption);

      if (prev_was_parent) {
    metaoption.pre_tree_traverse += ".pr";
      }
      else if (prev_was_child) {
    metaoption.pre_tree_traverse += ".fc";
      }

      prev_was_parent = true;
      prev_was_child  = false;
    } 
    else if (meta.size() > 6 
         && (substr (meta.begin(), meta.begin()+5) == "child")) {

      // clear out sibling and parent (cmd and options)
      metaoption.metacommand &= ~(mParent|mSibling);
      metaoption.parentoptions.clear();
      metaoption.siblingoptions.clear();

      metaoption.metacommand |= mChild;
      get_child_options (meta, metaoption);
      metadata.insert("contains");

      if (prev_was_parent) {
    metaoption.pre_tree_traverse += ".pr";
      }
      else if (prev_was_child) {
    metaoption.pre_tree_traverse += ".fc";
      }

      prev_was_child  = true;
      prev_was_parent = false;
    } 
    else {
      prev_was_child  = false;
      prev_was_parent = false;
      had_parent_or_child = false;
    }
  }

  // parent/child can have sibling tacked on end also
  if (meta.size() > 8 && (substr (meta.begin(), meta.begin()+7) == "sibling")) {
    metaoption.metacommand |= mSibling;
    get_sibling_options (meta, metaoption);
  } 
  
  // check for ex. which may occur in format statements
  if (meta.size()>3 && (substr(meta.begin(), meta.begin()+3) == "ex.")) {
    meta = substr (meta.begin()+3, meta.end());
  }
  metadata.insert (meta);
  metaoption.metaname = meta;
}

static void parse_coll_meta(text_t &meta, metadata_t &metaoption) {
  if (meta == "collection") {
    // no qualifiers
    metaoption.metaname = g_EmptyText;
    return;
  }
  meta = substr (meta.begin()+11, meta.end());
  metaoption.metaname = meta;
  
}

static void parse_meta (text_t &meta, format_t *formatlistptr,
            text_tset &metadata, bool &getParents) {
  
  if (meta == "link")
    formatlistptr->command = comLink;
  else if (meta == "/link")
    formatlistptr->command = comEndLink;

  else if (meta == "href")
    formatlistptr->command = comHref;

  else if (meta == "num")
    formatlistptr->command = comNum;

  else if (meta == "icon")
    formatlistptr->command = comIcon;

  else if (meta == "Text")
    formatlistptr->command = comDoc;
  
  else if (meta == "RelatedDocuments")
   formatlistptr->command = comRel;

  else if (meta == "highlight")
    formatlistptr->command = comHighlight;

  else if (meta == "/highlight")
    formatlistptr->command = comEndHighlight;

  else if (meta == "Summary")
    formatlistptr->command = comSummary;

  else if (meta == "DocImage")
    formatlistptr->command = comImage; 

  else if (meta == "DocTOC")
    formatlistptr->command = comTOC; 

  else if (meta == "DocumentButtonDetach")
     formatlistptr->command = comDocumentButtonDetach;
  
  else if (meta == "DocumentButtonHighlight")
     formatlistptr->command = comDocumentButtonHighlight;
  
  else if (meta == "DocumentButtonExpandContents")
    formatlistptr->command = comDocumentButtonExpandContents; 

  else if (meta == "DocumentButtonExpandText")
     formatlistptr->command = comDocumentButtonExpandText; 

  else if (meta == "DocOID")
     formatlistptr->command = comOID; 
  else if (meta == "DocTopOID")
    formatlistptr->command = comTopOID;
  else if (meta == "DocRank")
    formatlistptr->command = comRank;
  else if (meta == "DocTermsFreqTotal")
    formatlistptr->command = comDocTermsFreqTotal;
  else if (meta.size() >= 10 && (substr(meta.begin(), meta.begin()+10) == "collection")) {
    formatlistptr->command = comCollection;
    parse_coll_meta(meta, formatlistptr->meta);
  }
  else {
    formatlistptr->command = comMeta;
    parse_meta (meta, formatlistptr->meta, metadata, getParents);
  }
}


static bool parse_string (const text_t &formatstring, format_t *formatlistptr, 
              text_tset &metadata, bool &getParents) {

  text_t text;
  text_t::const_iterator here = formatstring.begin();
  text_t::const_iterator end = formatstring.end();

  while (here != end) {

    if (*here == '\\') {
      ++here;
      if (here != end) text.push_back (*here);

    } else if (*here == '{') {
      if (!text.empty()) {
    formatlistptr->command = comText;
    formatlistptr->text = text;
    formatlistptr->nextptr = new format_t();
    formatlistptr = formatlistptr->nextptr;
    
    text.clear();
      }
      if (parse_action (++here, end, formatlistptr, metadata, getParents)) {

    formatlistptr->nextptr = new format_t();
    formatlistptr = formatlistptr->nextptr;
    if (here == end) break;
      }
    } else if (*here == '[') {
      if (!text.empty()) {
    formatlistptr->command = comText;
    formatlistptr->text = text;
    formatlistptr->nextptr = new format_t();
    formatlistptr = formatlistptr->nextptr;

    text.clear();
      }
      text_t meta;
      ++here;
      while (*here != ']') {
    if (here == end) return false;
    meta.push_back (*here);
    ++here;
      }
      parse_meta (meta, formatlistptr, metadata, getParents);
      formatlistptr->nextptr = new format_t();
      formatlistptr = formatlistptr->nextptr;

    } else 
      text.push_back (*here);

    if (here != end) ++here;
  }
  if (!text.empty()) {
    formatlistptr->command = comText;
    formatlistptr->text = text;
    formatlistptr->nextptr = new format_t();
    formatlistptr = formatlistptr->nextptr;

  }
  return true;
}


static bool parse_action (text_t::const_iterator &here, const text_t::const_iterator &end, 
              format_t *formatlistptr, text_tset &metadata, bool &getParents) {

  text_t::const_iterator it = findchar (here, end, '}');
  if (it == end) return false;

  text_t com = substr (here, it);
  here = findchar (it, end, '{');
  if (here == end) return false;
  else ++here;

  if (com == "If" || com == "if" || com == "IF") formatlistptr->command = comIf;
  else if (com == "Or" || com == "or" || com == "OR") formatlistptr->command = comOr;
  else return false;

  int commacount = 0;
  text_t text;
  while (here != end) {

    if (*here == '\\') {
      ++here;
      if (here != end) text.push_back(*here);
      
    } 
  
    else if (*here == ',' || *here == '}' || *here == '{') {

      if (formatlistptr->command == comOr) {
    // the {Or}{this, or this, or this, or this} statement
    format_t *or_ptr;
    
    // find the next unused orptr
    if (formatlistptr->orptr == NULL) {
      formatlistptr->orptr = new format_t();
      or_ptr = formatlistptr->orptr;
    } else {
      or_ptr = formatlistptr->orptr;
      while (or_ptr->nextptr != NULL)
        or_ptr = or_ptr->nextptr;
      or_ptr->nextptr = new format_t();
      or_ptr = or_ptr->nextptr;
    }

    if (!text.empty())
      {
        if (!parse_string(text, or_ptr, metadata, getParents)) { return false; }
      }

    if (*here == '{')
      {
        // Supports: {Or}{[Booktitle],[Title],{If}{[XXXX],aaa,bbb}}
        // but not : {Or}{[Booktitle],[Title]{If}{[XXXX],aaa,bbb}}
        // The latter can always be re-written:
        // {Or}{[Booktitle],{If}{[Title],[Title]{If}{[XXXX],aaa,bbb}}}
        
        if (!text.empty()) // already used up allocated format_t
          {
        // => allocate new one for detected action
        or_ptr->nextptr = new format_t();
        or_ptr = or_ptr->nextptr;
          }
        if (!parse_action(++here, end, or_ptr, metadata, getParents))
          { 
        return false;
          }
      }
    else
      {
        if (*here == '}') break;
      }
    text.clear();

      } 

      // Parse an {If}{decide,do,else} statement
      else {
    
    // Read the decision component.  
    if (commacount == 0) {
      // Decsion can be a metadata element, or a piece of text.
      // Originally Stefan's code, updated 25/10/2000 by Gordon.

      text_t::const_iterator beginbracket = text.begin();
      text_t::const_iterator endbracket = (text.end() - 1);

      // Decision is based on a metadata element
      if ((*beginbracket == '[') && (*endbracket == ']')) {
        // Ignore the surrounding square brackets
        text_t meta = substr (beginbracket+1, endbracket);
        parse_meta (meta, formatlistptr->decision.meta, metadata, getParents);
        ++commacount;
        text.clear();
      } 

      // Decision is a piece of text (probably a macro like _cgiargmode_).
      else {

        // hunt for any metadata in string, which might be uses in
        // to test a condition, e.g. [Format] eq 'PDF'
        format_t* dummyformat = new format_t();
        // update which metadata fields needed 
        // (not interested in updatng formatlistptr)
        parse_string (text, dummyformat, metadata, getParents);
        delete dummyformat;

        formatlistptr->decision.command = dText;
        formatlistptr->decision.text = text;
        ++commacount;
        text.clear();
      }
    } 

    // Read the "then" and "else" components of the {If} statement.
    else {
      format_t** nextlistptr = NULL;
      if (commacount == 1) {
        nextlistptr = &formatlistptr->ifptr;
      } else if (commacount == 2 ) {
        nextlistptr = &formatlistptr->elseptr;
      } else {
        return false;
      }

      if (!text.empty()) {
        if (*nextlistptr == NULL) {
          *nextlistptr = new format_t();
        } else {

          // skip to the end of any format_t statements already added
          while ((*nextlistptr)->nextptr != NULL)
          {
        nextlistptr = &(*nextlistptr)->nextptr;
          }

          (*nextlistptr)->nextptr = new format_t();
          nextlistptr = &(*nextlistptr)->nextptr;
        }

        if (!parse_string (text, *nextlistptr, metadata, getParents)) 
          { 
        return false; 
          }
        text.clear();
      }
      
      if (*here == '{')
        {
          if (*nextlistptr == NULL) {
        *nextlistptr = new format_t();
          } else {
        // skip to the end of any format_t statements already added
        while ((*nextlistptr)->nextptr != NULL)
          {
            nextlistptr = &(*nextlistptr)->nextptr;
          }

        (*nextlistptr)->nextptr = new format_t();
        nextlistptr = &(*nextlistptr)->nextptr;
          }

          if (!parse_action(++here, end, *nextlistptr, metadata, getParents))
        { 
          return false;
        }
        }
      else 
        { 
          if (*here == '}') break;
          ++commacount; 
        }
    }
      }
      
    } else text.push_back(*here);
    
    if (here != end) ++here;
  }

  return true;
}


bool parse_formatstring (const text_t &formatstring, format_t *formatlistptr, 
             text_tset &metadata, bool &getParents) {

  formatlistptr->clear();
  getParents = false;

  return (parse_string (formatstring, formatlistptr, metadata, getParents));
}

// position -1 for all, -2 for the last, 0 for the first, or x for a particular piece
// metainfo has all the values for the metadata, meta has the request metadata name and options, position tells which values to get
static text_t get_formatted_meta_text(MetadataInfo_t &metainfo, const metadata_t &meta, int position, bool no_cgisafe = false) 
{
  text_t no_ns_metaname = remove_namespace(meta.metaname);
  text_t tmp;
  bool first = true;
  
  const int start_i=0;
  const int end_i = metainfo.values.size()-1;
  
  if (position == -1) { // all
    for (int i=start_i; i<=end_i; ++i) {
      if (!first) tmp += meta.siblingoptions;
      if (meta.metacommand & mSpecial) {
    // special formatting
    if (no_ns_metaname == "Date") tmp += format_date (metainfo.values[i]);
    else if (no_ns_metaname == "Language") tmp += iso639(metainfo.values[i]);
    else tmp += "_format:"+meta.metaname+"_("+metainfo.values[i]+")";
      }
      else tmp += metainfo.values[i];
      first = false;
      
    }
  } else {
    if (position == -2) { // end
      position = end_i;
    } else if (position < start_i || position > end_i) {
      return "";
    }
    if (meta.metacommand & mSpecial) {
      // special formatting
      if (no_ns_metaname == "Date") tmp += format_date (metainfo.values[position]);
      else if (no_ns_metaname == "Language") tmp += iso639(metainfo.values[position]);
      else tmp += "_format:"+meta.metaname+"_("+metainfo.values[position]+")";
    } 
    else tmp += metainfo.values[position];
  }
  if (meta.metacommand & mCgiSafe && !no_cgisafe) return cgi_safe_unicode (tmp);
  else return tmp;
}

static text_t get_parent_meta (ResultDocInfo_t &docinfo, const metadata_t &meta, int siblings_values) 
{
  
  MetadataInfo_t *parent = docinfo.metadata[meta.metaname].parent;
  switch (meta.mqualifier.parent) {
  case pNone:
    return "Nothing!!";
    break;

  case pImmediate:
    if (parent != NULL) {
      return get_formatted_meta_text(*parent, meta, siblings_values);
    }
    break;

  case pTop:
    if (parent != NULL) {
      while (parent->parent != NULL) parent = parent->parent;
      return get_formatted_meta_text(*parent, meta, siblings_values);
    }
    break;

  case pAll:
    MetadataInfo_t *parent = docinfo.metadata[meta.metaname].parent;
    if (parent != NULL) {
      text_tarray tmparray;
      while (parent != NULL) {
    tmparray.push_back (get_formatted_meta_text(*parent, meta, siblings_values, true)); // set no_cgisafe to true, as we'll do it once we have all the metadata
    parent = parent->parent;
      }
      // now join them up - use teh parent separator
      bool first = true;
      text_t tmp;
      text_tarray::reverse_iterator here = tmparray.rbegin();
      text_tarray::reverse_iterator end = tmparray.rend();
      while (here != end) {
    if (!first) tmp += meta.parentoptions;
    tmp += *here;
    first = false;
    ++here;
      }
      if (meta.metacommand & mCgiSafe) return cgi_safe_unicode (tmp);
      else return tmp;
    }
  }
  return "";

}

static text_t get_child_meta (const text_t& collection, 
                  recptproto* collectproto,
                  ResultDocInfo_t &docinfo, displayclass &disp,
                  const metadata_t &meta, text_tmap &options,
                  ostream& logout, int siblings_values) 
{
  if (docinfo.metadata["contains"].values[0].size()==0) return ""; // no children
  
  const text_t& pre_tree_trav = meta.pre_tree_traverse;
  const text_t& child_metaname = meta.metaname;
  const text_t& child_field = meta.childoptions;
  text_tset child_metadata;
  child_metadata.insert(child_metaname); 

  FilterResponse_t child_response;
  if (meta.mqualifier.child == cNum) {
    // just one child
    //get the information associated with the metadata for child doc
    if (!get_info (docinfo.OID+pre_tree_trav+child_field, collection, "", 
           child_metadata, false, collectproto, child_response, 
           logout)) return ""; // invalid child number

      if (child_response.docInfo.empty()) return false; // no info for the child
  
      ResultDocInfo_t& child_docinfo = child_response.docInfo[0];
      MetadataInfo_t& metaname_rec = child_docinfo.metadata[child_metaname];
  
      text_t child_metavalue = get_formatted_meta_text(metaname_rec,meta,siblings_values);
      return expand_metadata(child_metavalue,collection,collectproto,
             child_docinfo,disp,options,logout);
  }
  
    
  if (meta.mqualifier.child != cAll) return false; // invalid qualifier


  if (!pre_tree_trav.empty()) {
    // need to get relevant "contains" metadata for new (e.g. pre tree trav) node
    FilterResponse_t trav_response;

    text_tset trav_metadata;
    trav_metadata.insert("contains"); 

    if (!get_info (docinfo.OID+pre_tree_trav, collection, "", 
           trav_metadata, false, collectproto, trav_response, 
           logout)) return ""; // invalid pre_tree_trav

    if (trav_response.docInfo.empty()) return false; // no info for the pre_tree_trav OID
  
    ResultDocInfo_t& trav_docinfo = trav_response.docInfo[0];
    // use this for rest of routine
    docinfo = trav_docinfo;
  }
  
  // we need to get all children
  text_t result = "";
  text_tarray children;
  text_t contains = docinfo.metadata["contains"].values[0];
  splitchar (contains.begin(), contains.end(), ';', children);
  text_tarray::const_iterator here = children.begin();
  text_tarray::const_iterator end = children.end();
  bool first = true;
  while (here !=end) {
    text_t oid = *here;
    here++;
    if (*(oid.begin()) == '"') translate_parent (oid, docinfo.OID);
    
    //get the information associated with the metadata for child doc
    if (!get_info (oid, collection, "", child_metadata, 
           false, collectproto, child_response, logout) ||
    child_response.docInfo.empty()) {
      first = false;
      continue;
    }
    
    
    ResultDocInfo_t& child_docinfo = child_response.docInfo[0];
    MetadataInfo_t& metaname_rec = child_docinfo.metadata[child_metaname];
    
    text_t child_metavalue = get_formatted_meta_text(metaname_rec,meta,siblings_values);
    if (!first) result += child_field;
    first = false;
    // need to do this here cos otherwise we are in the wrong document
    result +=  expand_metadata(child_metavalue,collection,collectproto,
                   child_docinfo,disp,options,logout);
  }
  return result;
    
}

static text_t get_meta (const text_t& collection, recptproto* collectproto,
            ResultDocInfo_t &docinfo, displayclass &disp,
            const metadata_t &meta, text_tmap &options,
            ostream& logout) {
  
  // make sure we have the requested metadata
  MetadataInfo_tmap::iterator it = docinfo.metadata.find (meta.metaname);
  if (it == docinfo.metadata.end()) return "";

  int siblings_values = 0; // default is no siblings, just the first metadata available
  if (meta.metacommand & mSibling) {
    if (meta.mqualifier.sibling == sAll) {
      siblings_values = -1; //all
    } else if (meta.mqualifier.sibling == sNum) {
      siblings_values = meta.siblingoptions.getint();
    }
  }
  if (meta.metacommand & mParent) {
    return get_parent_meta(docinfo,meta,siblings_values);
  }

  else if (meta.metacommand & mChild) {
    return get_child_meta(collection,collectproto,docinfo,disp,meta,
                options,logout, siblings_values);
  }
  else if (meta.metacommand & mSibling) { // only siblings
    MetadataInfo_t& metaname_rec = docinfo.metadata[meta.metaname];
    return get_formatted_meta_text(docinfo.metadata[meta.metaname],meta, siblings_values);
  }
  else {
    
    // straightforward metadata request (nothing fancy)

    text_t classifier_metaname = docinfo.classifier_metadata_type;
    int metaname_index 
      = (classifier_metaname == meta.metaname) ? docinfo.classifier_metadata_offset : 0;
    return get_formatted_meta_text(docinfo.metadata[meta.metaname], meta, metaname_index);
  }
  
  return "";
}

static text_t get_or (const text_t& collection, recptproto* collectproto,
              ResultDocInfo_t &docinfo, displayclass &disp,
              format_t *orptr, text_tmap &options,
              ostream& logout) {

  text_t tmp;
  while (orptr != NULL) {

    tmp = format_string (collection,collectproto,docinfo, disp, orptr, 
             options, logout);
    if (!tmp.empty()) return tmp;

    orptr = orptr->nextptr;
  }
  return "";
}

static bool char_is_whitespace(const char c) 
{
  return ((c == ' ') || (c == '\t') || (c == '\n') || (c == '\r'));

}

static int scan_over_whitespace(const text_t& outstring, const int start_pos)
{
  int pos = start_pos;
  while (pos<outstring.size()) {
    if (!char_is_whitespace(outstring[pos])) {
      break;
    }
    ++pos;
  }

  return pos;
}

static int rscan_over_whitespace(const text_t& outstring, const int start_pos)
{
  int pos = start_pos;
  while (pos>=0) {
    if (!char_is_whitespace(outstring[pos])) {
      break;
    }
    --pos;
  }

  return pos;
}

static int rscan_for_whitespace(const text_t& outstring, const int start_pos)
{
  int pos = start_pos;
  while (pos>=0) {
    if (char_is_whitespace(outstring[pos])) {
      break;
    }
    --pos;
  }

  return pos;
}


static int rscan_for(const text_t& outstring, const int start_pos, 
             const char find_c)
{
  int pos = start_pos;
  while (pos>=0) {
    char c = outstring[pos];
    if (outstring[pos] == find_c) {
      break;
    }
    --pos;
  }

  return pos;
}

text_t extract_substr(const text_t& outstring, const int start_pos,
              const int end_pos)
{
  text_t extracted_str;
  extracted_str.clear();

  for (int pos=start_pos; pos<=end_pos; ++pos) {
    extracted_str.push_back(outstring[pos]);
  }

  return extracted_str;
}


static text_t expand_potential_metadata(const text_t& collection, 
                    recptproto* collectproto,
                    ResultDocInfo_t &docinfo, 
                    displayclass &disp,
                    const text_t& intext,
                    text_tmap &options,
                    ostream& logout)
{
  text_t outtext;

  // decide if dealing with metadata or text

  text_t::const_iterator beginbracket = intext.begin();
  text_t::const_iterator endbracket = (intext.end() - 1);

  // Decision is based on a metadata element
  if ((*beginbracket == '[') && (*endbracket == ']')) {
    // Ignore the surrounding square brackets
    text_t meta_text = substr (beginbracket+1, endbracket);

    if (meta_text == "Text") {
      outtext = format_text(collection, collectproto, docinfo, disp, options, logout);
    } else {

      text_tset metadata;
      bool getParents =false;
      metadata_t meta;
      
      parse_meta (meta_text, meta, metadata, getParents);   
      outtext 
    = get_meta (collection,collectproto,docinfo,disp,meta,options,logout);
    }

  }
  else {
    outtext = intext;
  }

  return outtext;
}




static bool uses_expression(const text_t& collection, recptproto* collectproto,
                ResultDocInfo_t &docinfo,
                displayclass &disp,
                const text_t& outstring, text_t& lhs_expr,
                text_t& op_expr, text_t& rhs_expr,
                text_tmap &options,
                ostream& logout)
{
  // Note: the string may not be of the form: str1 op str2, however
  // to deterine this we have to process it on the assumption it is,
  // and if at any point an 'erroneous' value is encountered, return
  // false and let something else have a go at evaluating it

  // Starting at the end of the string and working backwards ..

  const int outstring_len = outstring.size();

  // skip over white space
  int rhs_end = rscan_over_whitespace(outstring,outstring_len-1);

  if (rhs_end<=0) { 
    // no meaningful text or (rhs_end==0) no room for operator
    return false;
  }

  // check for ' or " and then scan over token
  const char potential_quote = outstring[rhs_end];
  int rhs_start=rhs_end;
  bool quoted = false;

  if ((potential_quote == '\'') || (potential_quote == '\"')) {
    --rhs_end;
    rhs_start = rscan_for(outstring,rhs_end-1,potential_quote) +1;
    quoted = true;
  }
  else {
    rhs_start = rscan_for_whitespace(outstring,rhs_end-1) +1;
  }

  if ((rhs_end-rhs_start)<0) {
    // no meaningful rhs expression
    return false;
  }

  // form rhs_expr
  rhs_expr = extract_substr(outstring,rhs_start,rhs_end);

  // skip over white space
  const int to_whitespace = (quoted) ? 2 : 1;

  int op_end = rscan_over_whitespace(outstring,rhs_start-to_whitespace);
  int op_start = rscan_for_whitespace(outstring,op_end-1)+1;

  if ((op_end<0) && (op_start<0)) {
    // no meaningful expression operator
    return false;
  }

  if (op_end-op_start<0) {
    // no meaningful expression operator
    return false;
  }

  op_expr = extract_substr(outstring,op_start,op_end);


  // check for operator
  if ((op_expr != "eq") && (op_expr != "ne") && (op_expr != "gt") &&
     (op_expr != "ge") && (op_expr != "lt") && (op_expr != "le") && (op_expr != "==") && (op_expr != "!=") &&  (op_expr != ">") && (op_expr != ">=") && (op_expr != "<") && (op_expr != "<=") && (op_expr != "sw") && (op_expr != "ew")) {

    // not a valid operator
    return false;
  }

  int lhs_end = rscan_over_whitespace(outstring,op_start-1);
  if (lhs_end<0) {
    // no meaningful lhs expression
    return false;
  }

  int lhs_start = scan_over_whitespace(outstring,0);

  // form lhs_expr from remainder of string
  lhs_expr = extract_substr(outstring,lhs_start,lhs_end);

  // Now we know we have a valid expression, look up any 
  // metadata terms

  rhs_expr = expand_potential_metadata(collection,collectproto,docinfo,
                       disp,rhs_expr,options,logout);
  lhs_expr = expand_potential_metadata(collection,collectproto,docinfo,
                       disp,lhs_expr,options,logout);

  return true;
}

static bool eval_expression_true(const text_t& lhs_expr,const text_t& op_expr,
                 const text_t& rhs_expr, ostream& logout)
{
  if (op_expr == "eq") return (lhs_expr == rhs_expr);
  else if (op_expr == "ne" ) return (lhs_expr != rhs_expr);
  else if (op_expr == "gt") return (lhs_expr > rhs_expr);
  else if (op_expr == "ge") return (lhs_expr >= rhs_expr);
  else if (op_expr == "lt") return (lhs_expr < rhs_expr);
  else if (op_expr == "le") return (lhs_expr <= rhs_expr);
  else if (op_expr == "==") return (lhs_expr.getint() == rhs_expr.getint());
  else if (op_expr == "!=") return (lhs_expr.getint() != rhs_expr.getint());
  else if (op_expr == ">") return (lhs_expr.getint() > rhs_expr.getint());
  else if (op_expr == ">=") return (lhs_expr.getint() >= rhs_expr.getint());
  else if (op_expr == "<") return (lhs_expr.getint() < rhs_expr.getint());
  else if (op_expr == "<=") return (lhs_expr.getint() <= rhs_expr.getint());
  else if (op_expr == "sw") return (starts_with(lhs_expr,rhs_expr));
  else if (op_expr == "ew") return (ends_with(lhs_expr,rhs_expr));
  else {
    logout << "Error: '" << op_expr << "' is not a recognised operator." << endl;
  }
  
  return false;
}


static text_t get_if (const text_t& collection, recptproto* collectproto,
              ResultDocInfo_t &docinfo, displayclass &disp, 
              const decision_t &decision,
              format_t *ifptr, format_t *elseptr,
              text_tmap &options, ostream& logout) 
{
  // If the decision component is a metadata element, then evaluate it
  // to see whether we output the "then" or the "else" clause
  if (decision.command == dMeta) {
    if (get_meta (collection,collectproto,docinfo,disp,decision.meta,options,
          logout) != "") {
      if (ifptr != NULL)
    return get_formatted_string (collection,collectproto,docinfo, disp, ifptr, 
                     options, logout);
    }
    else {
      if (elseptr != NULL)
    return get_formatted_string (collection,collectproto,docinfo, disp, elseptr,
                     options, logout);
    }
  }

  // If the decision component is text, then evaluate it (it is probably a 
  // macro like _cgiargmode_) to decide what to output.
  else if (decision.command == dText) {

    text_t outstring;
    disp.expandstring (decision.text, outstring);

    // Check for if expression in form: str1 op str2 
    // (such as [x] eq "y")
    text_t lhs_expr, op_expr, rhs_expr;
    if (uses_expression(collection,collectproto,docinfo, disp, outstring,lhs_expr,op_expr,rhs_expr, options,logout)) {
      if (eval_expression_true(lhs_expr,op_expr,rhs_expr,logout)) {
    if (ifptr != NULL) {
      return get_formatted_string (collection, collectproto, docinfo, disp, ifptr, 
                       options, logout);
    }
    else {
      return "";
    }
      } else {
    if (elseptr != NULL) {
      return get_formatted_string (collection, collectproto, docinfo, disp, elseptr,
                       options, logout);
    }
    else {
      return "";
    }
      }
    }


    // This is a tad tricky.  When we expand a string like _cgiargmode_, that is
    // a cgi argument macro that has not been set, it evaluates to itself.
    // Therefore, were have to say that a piece of text evalautes true if
    // it is non-empty and if it is a cgi argument evaulating to itself.

    if ((outstring != "") && !((outstring == decision.text) && (outstring[0] == '_'))) {
      if (ifptr != NULL)
    return get_formatted_string (collection, collectproto, docinfo, disp, ifptr, 
                     options, logout);
    } else {
      if (elseptr != NULL)
    return get_formatted_string (collection, collectproto, docinfo, disp, elseptr,
                     options, logout);
    }
  }
  
  return "";
}

bool includes_metadata(const text_t& text)
{
  text_t::const_iterator here = text.begin();
  text_t::const_iterator end = text.end();
  while (here != end) {
    if (*here == '[') return true;
    ++here;
  }

  return false;
}

static text_t expand_metadata(const text_t &metavalue, const text_t& collection,
                  recptproto* collectproto,
                  ResultDocInfo_t &docinfo,
                  displayclass &disp, text_tmap &options,
                  ostream &logout) {
      
  if (includes_metadata(metavalue)) {
    
    // text has embedded metadata in it => expand it
    FilterRequest_t request;
    FilterResponse_t response;
    
    request.getParents = false;
    
    format_t *expanded_formatlistptr = new format_t();
    parse_formatstring (metavalue, expanded_formatlistptr, 
            request.fields, request.getParents);
    
    // retrieve metadata
    get_info(docinfo.OID, collection, "", request.fields, request.getParents,
         collectproto, response, logout);
    
    if (!response.docInfo.empty()) {
      
      text_t expanded_metavalue
    = get_formatted_string(collection, collectproto,
                   response.docInfo[0], disp, expanded_formatlistptr, 
                   options, logout);
      
      return expanded_metavalue;
    }
    else {
      return metavalue;
    }
  }
  else {
    
    return metavalue;
  }
}

text_t get_collection_meta(const text_t& collection, recptproto* collectproto,
               displayclass &disp, 
               text_t meta_name, ostream& logout) {
  
  ColInfoResponse_t collectinfo;
  comerror_t err;
  collectproto->get_collectinfo (collection, collectinfo,err,logout);
  text_t meta_value = "";
  text_t lang;
  disp.expandstring("_cgiargl_",lang);
  if (lang.empty()) {
    lang = "en";
  }

  if (err == noError) {
    meta_value = collectinfo.get_collectionmeta(meta_name, lang);
  }
  return meta_value;
  

}
text_t format_string (const text_t& collection, recptproto* collectproto,
              ResultDocInfo_t &docinfo, displayclass &disp, 
              format_t *formatlistptr, text_tmap &options,
              ostream& logout) {

  if (formatlistptr == NULL) return "";

  switch (formatlistptr->command) {
     case comOID:
    return docinfo.OID;
  case comTopOID:
    {
      text_t top_id;
      get_top(docinfo.OID, top_id);
      return top_id;
    }
  case comRank:
    return text_t(docinfo.ranking);
     case comText:
    return formatlistptr->text;
     case comLink:
    return options["link"];
     case comEndLink:
    if (options["link"].empty()) return "";
    else return "</a>";
     case comHref:
    return get_href(options["link"]);
     case comIcon:
    return options["icon"];
     case comNum:
    return docinfo.result_num;
     case comRel: //if [RelatedDocuments] appears in format string, collect relation data
    return get_related_docs(collection, collectproto, docinfo, logout);
     case comSummary:
    return format_summary(collection, collectproto, docinfo, disp, options, logout);
     case comMeta:
    {
       const text_t& metavalue =  get_meta (collection,collectproto, docinfo, disp,formatlistptr->meta,options,logout);
       return expand_metadata(metavalue, collection, collectproto, docinfo, disp, options, logout);
    }
     case comDoc:
    return format_text(collection, collectproto, docinfo, disp, options, logout);
    //return options["text"];
     case comImage:
    return expand_metadata(options["DocImage"], collection, collectproto, docinfo, disp, options, logout);
     case comTOC:
    return options["DocTOC"];
     case comDocumentButtonDetach:
    return options["DocumentButtonDetach"];
     case comDocumentButtonHighlight:
    return options["DocumentButtonHighlight"];
     case comDocumentButtonExpandContents:
    return options["DocumentButtonExpandContents"];
     case comDocumentButtonExpandText:
    return options["DocumentButtonExpandText"];
     case comHighlight:
    if (options["highlight"] == "1") return "<b>";
    break;
     case comEndHighlight:
    if (options["highlight"] == "1") return "</b>";
    break;
     case comIf:
    return get_if (collection, collectproto, docinfo, disp, 
               formatlistptr->decision, formatlistptr->ifptr, 
               formatlistptr->elseptr, options, logout);
     case comOr:
    return get_or (collection,collectproto, docinfo, disp, formatlistptr->orptr, 
               options, logout);
     case comDocTermsFreqTotal:
       return docinfo.num_terms_matched;
     case comCollection:
       if (formatlistptr->meta.metaname == g_EmptyText) {
     return collection;
       }
       return get_collection_meta(collection, collectproto, disp, formatlistptr->meta.metaname, logout);
    
  }
  return "";
}

text_t get_formatted_string (const text_t& collection, recptproto* collectproto,
                 ResultDocInfo_t &docinfo, displayclass &disp,
                 format_t *formatlistptr, text_tmap &options,
                 ostream& logout) {

   text_t ft;
   while (formatlistptr != NULL) 
      {
     ft += format_string (collection, collectproto, docinfo, disp, formatlistptr, 
                  options, logout);
     formatlistptr = formatlistptr->nextptr;
      }
   
   return ft;
}


// we have only preloaded the text in DocumentAction. But you may want to get the text in query. so copy what we have done with format_summary and get the text here. probably is quite expensive?
text_t format_text (const text_t& collection, recptproto* collectproto,
            ResultDocInfo_t &docinfo, displayclass &disp, 
            text_tmap &options, ostream& logout) {
  if(!options["text"].empty()) { 
    return options["text"];
  }
  // else get document text here
  DocumentRequest_t docrequest;
  DocumentResponse_t docresponse;
  comerror_t err;
  docrequest.OID = docinfo.OID;
  collectproto->get_document (collection, docrequest, docresponse, err, logout);
  return docresponse.doc;

}
  
/* FUNCTION NAME: format_summary
 * DESC: this is invoked when a [Summary] special metadata is processed.
 * RETURNS: a query-biased summary for the document */

text_t format_summary (const text_t& collection, recptproto* collectproto,
               ResultDocInfo_t &docinfo, displayclass &disp, 
               text_tmap &options, ostream& logout) {

  // GRB: added code here to ensure that the cstr (and other collections)
  //      uses the document metadata item Summary, rather than compressing
  //      the text of the document, processed via the methods in
  //      summarise.cpp
  if (docinfo.metadata.count("Summary") > 0 &&
      docinfo.metadata["Summary"].values.size() > 0) {
    return docinfo.metadata["Summary"].values[0];
  }

  text_t textToSummarise, query;
  if(options["text"].empty()) { // get document text
     DocumentRequest_t docrequest;
     DocumentResponse_t docresponse;
     comerror_t err;
     docrequest.OID = docinfo.OID;
     collectproto->get_document (collection, docrequest, docresponse, err, logout);
     textToSummarise = docresponse.doc;
  } else // in practice, this would not happen, because text is only 
         // loaded with the [Text] command
     textToSummarise = options["text"];
  disp.expandstring("_cgiargq_",query);
  return summarise(textToSummarise,query,80);
}