source: main/trunk/greenstone2/bin/script/indexes/buildkpiS.pl@ 24192

#! /user/bin/perl 

#usage: perl buildkpiS.pl [-R] [collection] [collection] etc
#
#-r or -R will remove previous index files so that you may build new ones
#
#The program performs the following tasks:
#-gathers the specified collections on the command line OR
#-gathers the directories of all the collections in the collect directory, this is all
# the directories apart from CVS, modelcol, . and .. which are not collections.
#-It then retrieves the archive.inf file from the archive directory of each collection
# to obtain the unique file ID and filepath of every document in the collection
#-Then parse through each doc.gml stored in filepath to gather information
#-From each file collect stems of kea keyphrases.
#-Determine the number of keyphrase stems for each document 
#-Search for the keyphrase stems in the numbered phrase index. If a phrase is not there then 
# the program will write the keyphrase stem to the phrase index . 
#-Search for the keyphrase stem in the keyphrase to document index. If the stem is there, 
# it will increment the number of documents that the keyphrase stem appears in and replace that
# then append the hash ID to the list of documents in the entry. If the stem is not there 
# then the program will write the stem to the phrase index.
#-Then write document ID, no of phrase, phrase number from index followed by number of times 
# phrase appears into the document_keyphrase index

$gsdlhome = $ENV{'GSDLHOME'};
$collection;

require "getopts.pl";
&Getopts('R'); #process option arguments

#if option R remove all previous indexes
if($opt_R == 1){ #remove indexes
    print STDERR "\nremoving $gsdlhome/bin/script/indexes/keyphrase_index.txt\n";
    print STDERR "removing $gsdlhome/bin/script/indexes/keyphrase_document.txt\n";
    print STDERR "removing $gsdlhome/bin/script/indexes/document_keyphrase.txt\n";
    system("rm $gsdlhome/bin/script/indexes/keyphrase_document.txt");   
    system("rm $gsdlhome/bin/script/indexes/document_keyphrase.txt"); 
    system("rm $gsdlhome/bin/script/indexes/keyphrase_index.txt");
}

#collections may be specified in the command line
#otherwise, all collections will be used to build
#the indexes.  
if(@ARGV){

    @directories = @ARGV;

} else { #open collect directory and get a list of all collections
    opendir(DIR, "$gsdlhome/collect");
    @directories = grep(!/(^\.|(CVS)|(modelcol))/, readdir(DIR));
    closedir(DIR);
}

#for each collection specified to build indexes for
foreach $collection (@directories){

    my @filelist;

    #archives.inf contains a list of unique hash ID's of each file and file paths
    open(INFO, "$gsdlhome/collect/$collection/archives/archives.inf")
    or die "$gsdlhome/collect/$collection/archives/archives.inf could not be opened.";

    while(<INFO>){ #get each line of text from archives.inf (OID \t filepath)
    chomp;
    push(@filelist, $_);
    }

    foreach $file (@filelist){ #add each document to the indexes
    build_index($file, $collection);
    }
}

#This function opens the file in the filepath sent as an argument.  From this it obtains 
#the kea and/or stem data, and then searches for these phrases in the file, counting and storing 
#how many times each phrase appears. The data is then sent to function keyphrase_document 
#with arguments hash ID, kea phrases and stem phrases to build the keyphrase_document index.
#The function which builds the document_keyphrase index is then passed the hash ID, the kea 
#phrases and/or the stemmed phrases and the array/s which hold the number of times each phrase
#appears in the document so that the data it has collected can be written to document_ keyphrase
#index.

sub build_index {

    my $args = shift(@_);
    my $collection = shift(@_);
    my ($ID, $filepath) = split(/\t/, $args);
    my $stemsS = "";
    my @stem_phrase_counts = 0;
    my $text = "";

    print STDERR "\nID: $ID\n";
    print STDERR "filepath: $filepath\n";

    #open file to extract keyphrase information
    open(FILE, "$gsdlhome/collect/$collection/archives/$filepath")
    or die "$gsdlhome/collect/$collection/archives/$filepath could not be opened.";

    #patterns to search for so that we can extract the kea information
    my $stem_search = "stems=\"([^\"]*)\"";

    while(<FILE>){ #get kea and stem data and store
    chomp;
    $stemsS = $1 if (/$stem_search/);
    }

    close(FILE);
    
    print STDERR "stems: $stemsS\n";

    my @stems = split(", ", $stemsS);

    if(@stems){ #if the data exists
    
    #open the filepath to the current document
    open(FILE, "$gsdlhome/collect/$collection/archives/$filepath")
        or die "$gsdlhome/collect/$collection/archives/$filepath could not be opened.";
    
    while(<FILE>){ #get the text
        chomp;
        $text .= $_;
    }
 
    #chop out all things in angled brackets
    $text =~ s/(<[^>]*>)//g;

    #initilise counts
    for($i=0; $i<=$#stems; $i++){
        $stem_phrase_counts[$i] = 0;
    }
    
    #using regular expressions generated from stem-reg
    #count how many of each phrase appear in the document
    
    $text_copy = $text;
    for($i=0; $i<=$#stems; $i++){ #search for text with stem phrases
        my $stem = $stems[$i];
        $reg = &stem_reg(split(/\s+/, $stem));
        while($text_copy =~ s/$reg//i){
        $stem_phrase_counts[$i]++; #count the number of stem phrases
        } 
        $text_copy = $text;
    }

    
    #write data to keyphrase_document index
    &keyphrase_document($ID, $stemsS); 
    
    #write data to document_keyphrase index
    $stem_counts = join(", ", @stem_phrase_counts);
    &document_keyphrase($ID, $stemsS, $stem_counts);
    
    } else { 
    print STDERR "No stem data was found in file $filepath\n";
    }

}

#returns a regular expression designed to 
#search for stems in text
#eg 'agri cari'
#    agri followed by 0 or more non-whitespace characters
#         followed by one or more whitespace OR 0 or 1 non-whitespace characters
#    cari followed by 0 or more non-whitespace characters
#modified from original by Stephen Lundy 

sub stem_reg {
   
    $regexp = "";

    $l = @_;

    if ($l > 0) {
        $s = shift;
        $regexp = "$s\\S*";

        if ($l-1 > 0) {
            foreach $s (@_) {
                $regexp .= "(\\s+|\\S?)$s\\S*";
            }
        }
    }

    return $regexp;
}



#This function is passed as arguments a list of kea phrase stems. Its purpose is to 
#check in the keyphrase index file for each phrase and determine whether or not an entry has
#been made for that phrase and an index number assigned to it. If there has not been an entry
#made then an index number is assigned to the phrase and it is written to the file.  This 
#function is called by document_keyphrase and keyphrase_document.  Each line in the file has
#this form:
#-phrase index number:phrase
#This function then returns a table of pairs of the phrases that were sent as arguments to it
#{phrase => phrase index number}.
 
sub keyphrase_index_search {

    my $phrases = shift(@_);
    my @phrases = split(", ", $phrases);
    my %table;
    my $index = 1;
    my $create_new_index = 0;

    print STDERR "searching keyphrase index...\n";

    #initilise table of phrases and index numbers
    foreach $phrase (@phrases){
    $table{"$phrase"} = "0";
    }

    #open keyphrase index for appending data and for reading
    open(INDEX_OUT, ">>$gsdlhome/bin/script/indexes/keyphrase_index.txt");
    open(INDEX_IN, "$gsdlhome/bin/script/indexes/keyphrase_index.txt")
    or $create_new_index = 1;

    if($create_new_index == 0){
    #if the index already exists read in the phrases
    while(<INDEX_IN>){
        chomp;
        foreach $phrase (@phrases){
        if(/(\d+):$phrase/){
            $index = $1;
            $table{"$phrase"} = "$index";
        }
        }
        $index++; #new starting index (one + the last index)
    }

    close(INDEX_IN);

    }

    #add new phrases to the phrase index
    foreach $phrase (keys %table){
    if($table{"$phrase"} eq "0"){
        print INDEX_OUT "$index:$phrase\n";
        $table{"$phrase"} = "$index"; 
        $index++;
    }
    }

    close(INDEX_OUT);
    return %table;
}

#This function is passed as arguments file hash ID and a list of kea phrases and/or stems
#that exist for that particular file.  Its purpose is to write to the keyphrase_document
#index a line for the document it has been sent:
#-phrase index number:number of documents it appears in|ID
sub keyphrase_document{

    my ($ID, $stems) = @_;
    my $text = ""; 
    my @textlist;
    my $create_new_index = 0;

    print STDERR "writing to keyphrase_document.txt...\n";

    #get table of phrases and phrase indexes 
    my %table = keyphrase_index_search($stems); 
    
  
    #open index for reading
    open(INDEX_IN, "$gsdlhome/bin/script/indexes/keyphrase_document.txt")
    or $create_new_index = 1;

    #read in document if file exists
    if($create_new_index == 0){

    while(<INDEX_IN>){
        $text .= $_;
    }

    close(INDEX_IN);

    #split text into lines
    @textlist = split(/\n/, $text);

    } 

    #open index for output
    open(INDEX_OUT, ">$gsdlhome/bin/script/indexes/keyphrase_document.txt");

    if($create_new_index == 0){ #amend existing index
    
    foreach $line (@textlist){
        foreach $phrase (keys %table){
        if($line =~ /(\d+):(\d+)(.*)/){ #all lines of this form
            $index = $1;
            if($table{"$phrase"} eq "$index") { #if phrase exists in index
            $ids = $3; #get all doc IDs for that keyphrase
            if($ids !~ /$ID/){ #if doc ID not already included
                $num_docs = $2;
                $num_docs++; #increment number of docs
                $line = "$index:$num_docs$3|$ID"; #line to append to index
                $table{"$phrase"} = "0"; 
            }
            }
        }
        }
        print INDEX_OUT "$line\n";
    }
    }

    #add new phrases to the index
    foreach $phrase (keys %table){ #write 'phrase index:1:file ID
    if($table{"$phrase"} ne "0"){
        my $line = "$table{$phrase}:1:$ID";
        print INDEX_OUT "$line\n"; 
    }
    }

    close(INDEX_OUT);

}

#This function is passed as arguments file hash ID and a list of kea phrase stems
#that exist for that particular file and a list of the number of times each stem
#phrase appear in that document.  Its purpose is to write to the document_keyphrase
#index a line for the document it has been sent:
#-file ID:number of phrases and/or stems appear in the document
#      |pairs of 'phrase index,number of times the phrase appears in the document' 
sub document_keyphrase {

    my ($ID, $stemsS, $stem_c) = @_;
    my $text = "";
    my @textlist;
    my %phrases;
    my $create_new_index = 0;

    print STDERR "writing to document_keyphrase.txt...\n";

    #split phrase counts into arrays
    my @stem_counts = split(", ", $stem_c);
  
    #get table of phrases and phrase indexes
    my %table = keyphrase_index_search($stemsS);

    #split phrases into arrays
    my @stems = split(", ", $stemsS);

    #build new phrases dictionary
    for($i=0; $i<=$#stems; $i++){
    my $phrase = $table{"$stems[$i]"};
    $phrases{"$phrase"} = "$stem_counts[$i]";
    }   
    my @num = keys %phrases;
    my $phrasenum = $#num + 1; #number of phrases in doc

    #open index for reading
    open(INDEX_IN, "$gsdlhome/bin/script/indexes/document_keyphrase.txt")
    or $create_new_index = 1;

     
    if($create_new_index == 0){ #index doesn't need to be created

    while(<INDEX_IN>){
        $text .= $_;
    }

    close(INDEX_IN);
    
    #split text into lines
    @textlist = split(/\n/, $text);

    }  


    #must write this line to the file
    #'document ID:num of phrases|phrase index, number of times phrases appears
    my $newline = "$ID:$phrasenum";
    foreach $phrase (keys %phrases){
    $newline .= "|$phrase,$phrases{$phrase}";
    }

    #open index for output
    open(INDEX_OUT, ">$gsdlhome/bin/script/indexes/document_keyphrase.txt");

    if($create_new_index == 1){ #create a new index

    print INDEX_OUT "$newline\n";

    } else {

    #if ID is already in the file write line overtop incase
    #someone has modified the file.  Otherwise add the line
    #to the end of the file
    my $found = 0;

    foreach $line (@textlist){
        if($line =~ /([^:]+):(.*)/){ #all lines should follow this pattern
        $id = $1;
        if($ID eq $id){ #id is already in the file
            print INDEX_OUT "$newline\n"; #print line overtop
            $found = 1;
        } else {
            print INDEX_OUT "$line\n"; #print old line out
        }
        }
    }

    print INDEX_OUT "$newline\n" if ($found == 0); #append new line to end of file 

    }

    close(INDEX_OUT);

}