Changeset 30022 for main/trunk/greenstone2/perllib

Timestamp:

2015-07-14T21:09:32+12:00 (9 years ago)

Author:

ak19

Message:

Finally committing Dr Bainbridge's suggested fix (tested) to handle the updated Tudor collection when downloaded from the web. The HTML files link to themselves with the <link rel> tag, in which case the HTML files were blocked from being processed and 0 docs were processed. Now HTMLPlugin tests whether any of the linked associated files found are to the document itself, and in such cases, they are not blocked.

Location:

main/trunk/greenstone2/perllib/plugins

Files:

• DirectoryPlugin.pm (modified) (1 diff)
• HTMLPlugin.pm (modified) (1 diff)

main/trunk/greenstone2/perllib/plugins/DirectoryPlugin.pm

@@ -555 +555 @@
         next;
     }
-    print STDERR "** Dir Plugin processing $raw_full_filename\n";
+    print STDERR "** DirectoryPlugin processing $raw_full_filename\n";
     # Follow Windows shortcuts
     if ($raw_subfile =~ m/(?i)\.lnk$/ && (($ENV{'GSDLOS'} =~ m/^windows$/i) && ($^O ne "cygwin"))) {

main/trunk/greenstone2/perllib/plugins/HTMLPlugin.pm

@@ -283 +283 @@
         print $outhandle " ->$unicode_url_original_filename\n";
 
-        # Allow for possibility of raw byte version and Unicode versions of file
-        &util::block_filename($block_hash,$unicode_url_original_filename);
+        # make sure not to block the file itself, as happens when an html file links to itself
+        # e.g. if the current file is mary-boleyn/index.html and contains <link rel="canonical" href="index.html" />
+        my $unicode_html_fname = "";
+        $self->decode_text($html_fname,$content_encoding,$language,\$unicode_html_fname);       
+        if($unicode_url_original_filename ne $unicode_html_fname) {
+            # Allow for possibility of raw byte version and Unicode versions of file
+            &util::block_filename($block_hash,$unicode_url_original_filename);
+        }
     }
 
     # $url_original_filename = &util::upgrade_if_dos_filename($url_original_filename);
-    &util::block_filename($block_hash,$url_original_filename);
+    &util::block_filename($block_hash,$url_original_filename) if $url_original_filename ne $html_fname;
+
+            # but only add the linked file to the blocklist if the current html file does not link to itself
         
     }