Changeset 12834

Timestamp:

2006-09-22T16:03:36+12:00 (18 years ago)

Author:

kjdon

Message:

these convertto plugins were all setting extract_language=1 to their secondary plugins. we don't want this - only pass to secondary plugin if user has asked for it. textcat can be very slow, so don't want to run it unless we have to

Location:

trunk/gsdl/perllib/plugins

Files:

• ExcelPlug.pm (modified) (1 diff)
• PDFPlug.pm (modified) (1 diff)
• PPTPlug.pm (modified) (2 diffs)
• PSPlug.pm (modified) (1 diff)
• RTFPlug.pm (modified) (1 diff)
• WordPlug.pm (modified) (4 diffs)

trunk/gsdl/perllib/plugins/ExcelPlug.pm

@@ -74 +74 @@
     #$self->{'extract_language'} = 1;
     push(@$html_options, "-input_encoding", "utf8");
-    push(@$html_options,"-extract_language");
+    push(@$html_options,"-extract_language") if $self->{'extract_language'};
     $self = bless $self, $class;
     

trunk/gsdl/perllib/plugins/PDFPlug.pm

@@ -149 +149 @@
    
     if ($self->{'input_encoding'} eq "auto") {
-    # pdftohtml will always produce html files encoded as utf-8
-    # => restrict primary PDFPlug and secondary HTML plugin to use 
-    # utf8 and extract language.
     $self->{'input_encoding'} = "utf8";
-    $self->{'extract_language'} = 1;
-
-    push(@$html_options,"-extract_language");
-    }
+    }
+
     # if pdftohtml is always producing utf8, then htmlplug always needs this option
     push(@$html_options,"-input_encoding", "utf8");
-
+    push(@$html_options,"-extract_language") if $self->{'extract_language'};
     # Instruct HTMLPlug (when eventually accessed through read_into_doc_obj) 
     # to extract these metadata fields from the HEAD META fields

trunk/gsdl/perllib/plugins/PPTPlug.pm

@@ -112 +112 @@
     if ($self->{'input_encoding'} eq "auto") {
     $self->{'input_encoding'} = "utf8";
-    $self->{'extract_language'} = 1;
     if (defined $secondary_plugin_options->{'HTMLPlug'}){
         push(@$html_options,"-input_encoding", "utf8");
-        push(@$html_options,"-extract_language");
+        push(@$html_options,"-extract_language") if $self->{'extract_language'};
 
         # Instruct HTMLPlug (when eventually accessed through read_into_doc_obj) 
@@ -123 +122 @@
     if (defined $secondary_plugin_options->{'PagedImgPlug'}){
         push(@$pageimg_options,"-input_encoding", "utf8");
-        push(@$pageimg_options,"-extract_language");
+        push(@$pageimg_options,"-extract_language") if $self->{'extract_language'};
     }
     }

trunk/gsdl/perllib/plugins/PSPlug.pm

@@ -124 +124 @@
     #$self->{'extract_language'} = 1;
     push(@$text_options, "-input_encoding", "utf8");
-    push(@$text_options,"-extract_language");
+    push(@$text_options,"-extract_language") if $self->{'extract_language'};
     push(@$text_options , "-title_sub", '^(Page\s+\d+)?(\s*1\s+)?');
 

trunk/gsdl/perllib/plugins/RTFPlug.pm

@@ -82 +82 @@
     #$self->{'extract_language'} = 1;
     push(@$text_options, "-input_encoding", "utf8");
-    push(@$text_options,"-extract_language");
-    if ($self->{'description_tags'} == 1) {
-    push(@$html_options, "-description_tags");
-    }
-    
+    push(@$text_options,"-extract_language") if $self->{'extract_language'};
+    push(@$html_options, "-description_tags") if $self->{'description_tags'};
+    push(@$html_options,"-extract_language") if $self->{'extract_language'};
 
     $self = bless $self, $class;

trunk/gsdl/perllib/plugins/WordPlug.pm

@@ -115 +115 @@
     $self->{'convert_options'} = "-windows_scripting" if $self->{'windows_scripting'};
 
+    # we always save as utf-8
+    if ($self->{'input_encoding'} eq "auto") {
+    $self->{'input_encoding'} = "utf8";
+    }
+
     my $secondary_plugin_options = $self->{'secondary_plugin_options'};
     if (defined $self->{'windows_scripting'}) { 
@@ -120 +125 @@
         $secondary_plugin_options->{'StructuredHTMLPlug'} = [];
         my $structhtml_options = $secondary_plugin_options->{'StructuredHTMLPlug'};
-        if ($self->{'input_encoding'} eq "auto") {
-        $self->{'input_encoding'} = "utf8";
-        $self->{'extract_language'} = 1;
-        #push(@$structhtml_options,"-input_encoding", "utf8");
-        push(@$structhtml_options,"-extract_language");
-        }
-                
+        
         # Instruct HTMLPlug (when eventually accessed through read_into_doc_obj)
         # to extract these metadata fields from the HEAD META fields
@@ -132 +131 @@
         push (@$structhtml_options , "-title_sub", '^(Page\s+\d+)?(\s*1\s+)?');
         push (@$structhtml_options, "-description_tags") if $self->{'windows_scripting'};
+        push(@$structhtml_options,"-extract_language") if $self->{'extract_language'};
         push (@$structhtml_options, "-delete_toc") if $self->{'delete_toc'};
         push (@$structhtml_options, "-toc_header", $self->{'toc_header'}) if $self->{'toc_header'};
@@ -153 +153 @@
     # wvWare will always produce html files encoded as utf-8, so make sure the secondary HTMLPlug knows this
     push(@$html_options,"-input_encoding", "utf8");
-
-    if ($self->{'input_encoding'} eq "auto") {
-    $self->{'input_encoding'} = "utf8";
-    $self->{'extract_language'} = 1;
-    push(@$html_options,"-extract_language");
-    }
-    if ($self->{'description_tags'} == 1) {
-    push(@$html_options, "-description_tags");
-    }
+    push(@$html_options,"-extract_language") if $self->{'extract_language'};
+    push(@$html_options, "-description_tags") if $self->{'description_tags'};
+
     # Instruct HTMLPlug (when eventually accessed through read_into_doc_obj)
     # to extract these metadata fields from the HEAD META fields