Changeset 10218 for trunk/gsdl/perllib/plugins/BasPlug.pm

Timestamp:

2005-07-06T15:27:45+12:00 (19 years ago)

Author:

kjdon

Message:

Jeffrey's new parsing modifications, committed approx 6 July, 15.16

File:

• trunk/gsdl/perllib/plugins/BasPlug.pm (modified) (16 diffs)

trunk/gsdl/perllib/plugins/BasPlug.pm

@@ -53 +53 @@
 use gsprintf 'gsprintf';
 use printusage;
+#$%^
+use parse2;
+
 
 use GISBasPlug;
@@ -59 +62 @@
 
 my $unicode_list =
-    [ { 'name' => "auto",
-    'desc' => "{BasPlug.input_encoding.auto}" },
-      { 'name' => "ascii",
+    [ { 'name' => "ascii",
     'desc' => "{BasPlug.input_encoding.ascii}" },
       { 'name' => "utf8",
@@ -68 +69 @@
     'desc' => "{BasPlug.input_encoding.unicode}" } ];
 
+my $auto_unicode_list = 
+    [ { 'name' => "auto",
+    'desc' => "{BasPlug.input_encoding.auto}" } ];
+
 my $arguments =
     [ { 'name' => "process_exp",
@@ -90 +95 @@
     'desc' => "{BasPlug.input_encoding}",
     'type' => "enum",
-    'list' => $unicode_list,
+    'list' => $auto_unicode_list,
     'reqd' => "no" ,
     'deft' => "auto" } ,
@@ -107 +112 @@
     'type' => "language",
     'deft' => "en",
+    'char_length' => "2",
     'reqd' => "no" },
       { 'name' => "extract_acronyms",
@@ -141 +147 @@
     'type' => "int",
     'deft' => (localtime)[5]+1900,
+    'char_length' => "4",
+    #'range' => "2,100",
     'reqd' => "no"},
       { 'name' => "maximum_century",
     'desc' => "{BasPlug.maximum_century}",
     'type' => "string",
-    'deft' => "",
+    'deft' => "-1",
     'reqd' => "no" },
       { 'name' => "no_bibliography",
@@ -154 +162 @@
     'desc' => "{BasPlug.no_cover_image}",
     'type' => "flag",
-    'reqd' => "no" } ];
+    'reqd' => "no" },
+      { 'name' => "extract_keyphrases",
+    'desc' => "{BasPlug.extract_keyphrases}",
+    'type' => "flag",
+    'reqd' => "no",
+    'hiddengli' => "yes" },
+      { 'name' => "extract_keyphrase_options",
+    'desc' => "{BasPlug.extract_keyphrase_options}",
+    'type' => "string",
+    'reqd' => "no",
+    'hiddengli' => "yes" },
+      { 'name' => "separate_cjk",
+    'desc' => "{BasPlug.separate_cjk}",
+    'type' => "flag",
+    'reqd' => "no",
+    'hiddengli' => "yes" },
+      { 'name' => "smart_block",
+    'desc' => "{BasPlug.smart_block}",
+    'type' => "flag",
+    'reqd' => "no",
+    'hiddengli' => "yes" },
+      { 'name' => "new_extract_email",
+    'desc' => "",
+    'type' => "flag",
+    'reqd' => "no",
+    'hiddengli' => "yes" } ];
 
 my $gis_arguments = 
@@ -238 +271 @@
 {
     my $self = shift(@_);
-
     # Print the usage message for a plugin (recursively)
     my $descoffset = $self->determine_description_offset(0);
@@ -317 +349 @@
 
 sub new {
+    # Set Encodings to the list!!
+
+    my $e = $encodings::encodings;
+    foreach my $enc (sort {$e->{$a}->{'name'} cmp $e->{$b}->{'name'}} keys (%$e)) 
+    {
+    my $hashEncode =
+        {'name' => $enc,
+         'desc' => $e->{$enc}->{'name'}};
+
+    push(@{$unicode_list},$hashEncode);
+    }
+
+    push(@{$auto_unicode_list},@{$unicode_list});
+
+    # Start the BasPlug Constructor
     my $class = shift (@_);
-    my $plugin_name = shift (@_);
-    my $self = {};
-    $self->{'plugin_type'} = "BasPlug";
+    my ($pluginlist,$args,$hashArgOptLists) = @_;
+    push(@$pluginlist, $class);
+    my $plugin_name = (defined $pluginlist->[0]) ? $pluginlist->[0] : $class;
+
+    if(defined $arguments){ push(@{$hashArgOptLists->{"ArgList"}},@{$arguments});}
+    if(defined $options) { push(@{$hashArgOptLists->{"OptList"}},$options)};
 
     if (GISBasPlug::has_mapdata()) {
     push(@$arguments,@$gis_arguments);
     }
-
-    my $enc = "^(";
-    map {$enc .= "$_|";} keys %$encodings::encodings;
-    my $denc = $enc . "ascii|utf8|unicode)\$";
-    $enc .= "ascii|utf8|unicode|auto)\$";
-    
+    
+    my $self = {};
+    if(!parse2::parse($args,$hashArgOptLists->{"ArgList"},$self))
+    {
+    my $classTempClass = bless $self, $class;
+    &gsprintf(STDERR, "\n{BasPlug.bad_general_option}\n", $plugin_name);
+    $classTempClass->print_txt_usage("");  # Use default resource bundle
+    die "\n";
+    }
+
+    # else parsing was successful.
+
+    $self->{'plugin_type'} = $plugin_name;
     $self->{'outhandle'} = STDERR;
-    my $year = (localtime)[5]+1900;
-
     $self->{'textcat'} = new textcat();
-
     $self->{'num_processed'} = 0;
     $self->{'num_not_processed'} = 0;
@@ -341 +395 @@
     $self->{'num_archives'} = 0;
     $self->{'cover_image'} = 1; # cover image is on by default
-
-    # 14-05-02 To allow for proper inheritance of arguments - John Thompson
-    $self->{'option_list'} = [ $options ];
-    
-    my $no_cover_image = 0;
-    # general options available to all plugins
-    if (!parsargv::parse(\@_,
-             q^process_exp/.*/^, \$self->{'process_exp'},
-             q^block_exp/.*/^, \$self->{'block_exp'},
-             q^associate_ext/.*/^, \$self->{'associate_ext'},
-             q^extract_language^, \$self->{'extract_language'},
-             q^extract_acronyms^, \$self->{'extract_acronyms'}, 
-             q^extract_keyphrases^, \$self->{'kea'}, #with extra options (UNDOCUMENTED)
-             q^extract_keyphrase_options/.*/^, \$self->{'kea_options'}, #no extra options (UNDOCUMENTED)
-             qq^input_encoding/$enc/auto^, \$self->{'input_encoding'},
-             qq^default_encoding/$denc/utf8^, \$self->{'default_encoding'},
-             q^extract_email^, \$self->{'extract_email'},
-             q^extract_placenames^, \$self->{'extract_placenames'}, 
-             q^gazetteer/.*/^, \$self->{'gazetteer'},
-             q^place_list^, \$self->{'place_list'}, 
-             q^markup_acronyms^, \$self->{'markup_acronyms'},
-             q^default_language/.{2}/en^, \$self->{'default_language'},
-             q^first/.*/^, \$self->{'first'},
-             q^extract_historical_years^, \$self->{'date_extract'},
-             qq^maximum_year/\\d{4}/$year^, \$self->{'max_year'},
-             q^no_bibliography^, \$self->{'no_biblio'},
-             qq^maximum_century/-?\\d{1,2}( ?B\\.C\\.E\\.)?/-1^, \$self->{'max_century'},
-             q^no_cover_image^, \$no_cover_image,
-             q^separate_cjk^, \$self->{'separate_cjk'},
-             q^smart_block^, \$self->{'smart_block'},
-             q^smart_block_BN^, \$self->{'smart_block_BN'}, 
-             "allow_extra_options")) {
-
-    gsprintf(STDERR, "\n{BasPlug.bad_general_option}\n", $plugin_name);
-        bless $self, $class;
-    $self->print_txt_usage("");  # Use default resource bundle
-    die "\n";
-    }
-
+    $self->{'cover_image'} = 0 if ($self->{'no_cover_image'});
+    $self->{'file_blocks'} = {};
+    $self->{'option_list'} = $hashArgOptLists->{"OptList"};
+    
     my $associate_ext = $self->{'associate_ext'};
     if ((defined $associate_ext) && ($associate_ext ne "")) {
@@ -395 +414 @@
     $self->{'file_blocks'} = {};
 
-    $self->{'cover_image'} = 0 if ($no_cover_image);
-
     if ($self->{'extract_placenames'}) {
 
     my $outhandle = $self->{'outhandle'};
-
+    
     my $places_ref 
         = GISBasPlug::loadGISDatabase($outhandle,$self->{'gazetteer'});
-
+    
     if (!defined $places_ref) {
         print $outhandle "Warning: Error loading mapdata gazetteer \"$self->{'gazetteer'}\"\n";
@@ -414 +431 @@
     }   
     return bless $self, $class;
+    
 }
 
@@ -1121 +1139 @@
     @email = sort @email;
     
-    my @email2 = ();
+#    if($self->{"new_extract_email"} == 0)
+#    {
+#    my @email2 = ();
+#    foreach my $address (@email) 
+#   {
+#   if (!(join(" ",@email2) =~ m/(^| )$address( |$)/ )) 
+#       {
+#       push @email2, $address;
+#       $doc_obj->add_utf8_metadata ($thissection, "emailAddress", $address);
+#       # print $outhandle "  extracting $address\n" 
+#       &gsprintf($outhandle, "  {BasPlug.extracting} $address\n")
+#           if ($self->{'verbosity'} > 3);
+#       }
+#   }
+#    }
+#    else
+#    {
+    my $hashExistMail = {};
     foreach my $address (@email) {
-    if (!(join(" ",@email2) =~ m/$address/ )) {
-        push @email2, $address;
+    if (!(defined $hashExistMail->{$address}))
+    {
+        $hashExistMail->{$address} = 1;
         $doc_obj->add_utf8_metadata ($thissection, "emailAddress", $address);
         gsprintf($outhandle, "  {BasPlug.extracting} $address\n")
@@ -1157 +1193 @@
     } 
 
-    # adding kea keyphrases
-    if ($self->{'kea'}) {
+    #adding kea keyphrases
+    if ($self->{'extract_keyphrases'}) {  
     
     my $thissection = $doc_obj->get_top_section();
@@ -1171 +1207 @@
     } 
        
-    if ($self->{'kea_options'}) {
-        #if kea options flag is set, call Kea with specified options 
-        $list = &Kea::extract_KeyPhrases ($text, $self->{'kea_options'}); 
-    } else { 
-        #otherwise call Kea with no options
+    if($self->{'extract_keyphrase_options'}) { #if kea options flag is set, call Kea with specified options 
+        $list = &Kea::extract_KeyPhrases ($text, $self->{'extract_keyphrase_options'}); 
+    } else { #otherwise call Kea with no options
         $list = &Kea::extract_KeyPhrases ($text); 
     }
+     
     if ($list){
         # if a list of kea keyphrases was returned (ie not empty)
@@ -1227 +1262 @@
     }
 
-    if($self->{'date_extract'}) {
+    if($self->{'extract_historical_years'}) {
     my $thissection = $doc_obj->get_top_section();
     while (defined $thissection) {
-        
+
         my $text = $doc_obj->get_text($thissection);
         &DateExtract::get_date_metadata($text, $doc_obj, 
                         $thissection, 
-                        $self->{'no_biblio'}, 
-                        $self->{'max_year'}, 
-                        $self->{'max_century'});
+                        $self->{'no_bibliography'}, 
+                        $self->{'maximum_year'}, 
+                        $self->{'maximum_century'});
         $thissection = $doc_obj->get_next_section ($thissection);
     }