Changeset 16790 for gsdl/trunk/bin/script/explode_metadata_database.pl

Timestamp:

2008-08-14T16:39:09+12:00 (16 years ago)

Author:

davidb

Message:

Support for exploding oai records. Added at the request of John Rose.

File:

• gsdl/trunk/bin/script/explode_metadata_database.pl (modified) (8 diffs)

gsdl/trunk/bin/script/explode_metadata_database.pl

@@ -16 +16 @@
 use parse2;
 use FileHandle;
+
+use File::Spec;
+use File::Basename;
 
 my $unicode_list =
@@ -95 +98 @@
         'args' => $arguments };
 
-        
+
+
 sub main
 {
@@ -131 +135 @@
     }
 
+    
     # There should one arg left after parsing (the filename)
     # Or the user may have specified -h, in which case we output the usage
@@ -177 +182 @@
     }
 
-    my $text = "";
-    # Use the plugin's read_file function to avoid duplicating code
-    $plugobj->read_file($filename, $input_encoding, undef, \$text);
-    # is there any text in the file??
-    die "\n" unless length($text);
-
     # Create a directory to store the document files...
-    my ($documents_directory_base) = ($filename =~ /(.*)\.[^\.]+$/);
-
-    # Split the text into records, using the plugin's split_exp
+    my ($exploded_base_dir) = ($filename =~ /(.*)\.[^\.]+$/);
+
+    my $orig_base_dir = &File::Basename::dirname($filename);
+
+
     my $split_exp = $plugobj->{'split_exp'};
-    my @metadata_records = split(/$split_exp/, $text);
-    print STDERR "Number of records: " . scalar(@metadata_records) . "\n";
-
-    # Write the metadata from each record to the metadata.xml file
-    my $record_number = 1;
-    my $documents_directory;
-    foreach my $record_text (@metadata_records) {
-    # Check if we need to start a new directory for these records
-    if (($record_number % $records_per_folder) == 1) {
-        $documents_directory = $documents_directory_base;
-        if (scalar(@metadata_records) > $records_per_folder) {
-        $documents_directory .= "." . sprintf("%8.8d", $record_number);
-        }
-        if (-d $documents_directory) {
-        die "Error: document directory $documents_directory already exists (bailing).\n";
-        }
-        &util::mk_dir($documents_directory);
-
-        my $documents_metadata_xml_file = &util::filename_cat($documents_directory, "metadata.xml");
-        if (-e $documents_metadata_xml_file) {
-        die "Error: documents metadata.xml file $documents_metadata_xml_file already exists (bailing).\n";
-        }
-
-        # Start the metadata.xml file
-        open(METADATA_XML_FILE, ">$documents_metadata_xml_file");
-        print METADATA_XML_FILE
-        "<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"no\"?>\n" .
-        "<!DOCTYPE DirectoryMetadata SYSTEM \"http://greenstone.org/dtd/DirectoryMetadata/1.0/DirectoryMetadata.dtd\">\n" .
-        "<DirectoryMetadata>\n";
-    }
-
-    # Use the plugin's process function to avoid duplicating code
+    if (defined $split_exp) {
+    # Read in file, and then split and process individual records
+
+    my $text = "";
+    # Use the plugin's read_file function to avoid duplicating code
+    $plugobj->read_file($filename, $input_encoding, undef, \$text);
+    # is there any text in the file??
+    die "\n" unless length($text);
+
+    # Split the text into records, using the plugin's split_exp
+
+    my @metadata_records = split(/$split_exp/, $text);
+    print STDERR "Number of records: " . scalar(@metadata_records) . "\n";
+    
+    # Write the metadata from each record to the metadata.xml file
+    my $record_number = 1;
+    foreach my $record_text (@metadata_records) {
+        
+        # Check if we need to start a new directory for these records
+        my $documents_directory;
+        check_need_new_directory($exploded_base_dir,$record_number,$records_per_folder,
+                     \@metadata_records,\$documents_directory);
+        
+        # Use the plugin's process function to avoid duplicating code
+        my $doc_obj = new doc($filename, "nonindexed_doc");
+        $plugobj->process(\$record_text, undef, undef, $filename, undef, $doc_obj, 0);
+        
+        
+        # Try to get a doc to attach the metadata to
+        # If no match found, create a dummy .nul file
+        attach_metadata_or_nul_doc($document_field, $doc_obj, $record_number,
+                       $documents_directory, $orig_base_dir,
+                       $document_prefix, $document_suffix, $metadata_set, $verbosity);
+                
+        
+        check_close_directory($record_number,$records_per_folder,\@metadata_records);
+        
+        $record_number = $record_number + 1;
+    }
+    }
+    else {
+    # Call metadata_read to sets up associated metadata 
+
+    my $pluginfo = undef;
+    my $metadata = {};
+
+    my $processor = undef;
+    my $maxdocs = undef;
+    my $gli = undef;
+
+    my $extrametakeys = {};
+    my $extrametadata = {};
+
+
+    $plugobj->metadata_read($pluginfo, "", $filename, $metadata, 
+                $extrametakeys, $extrametadata, $processor, $maxdocs, $gli);
+
+
+    my $documents_directory = need_new_directory($exploded_base_dir);
+
+    # Attach metadata to object
+    # => use the plugin's extra_metadata function to avoid duplicating code
     my $doc_obj = new doc($filename, "nonindexed_doc");
-    $plugobj->process(\$record_text, undef, undef, $filename, undef, $doc_obj, 0);
-    # Get all the metadata assigned to this record
-    my $record_metadata = $doc_obj->get_all_metadata($doc_obj->get_top_section());
-    my $document_file;
-    
-    # try to get a doc to attach the metadata to
-    if (defined $document_field) {
-        foreach my $pair (@$record_metadata) {
-        my ($field, $value) = (@$pair);
-        $value =~ s/\\\\/\\/g;
-
-        # Does this metadata element specify a document to obtain?
-        if ($field eq $document_field) {
-            my $document_file_full = $document_prefix . $value . $document_suffix;
-            $document_file = &obtain_document($document_file_full, $documents_directory, $verbosity);
-            &write_metadata_xml_file_entry(METADATA_XML_FILE, $document_file, $record_metadata, $metadata_set);
-        }
-        }
-    }
-    # Create a dummy .nul file if we haven't obtained any documents for this record
-    if (not defined $document_file) {
-        $document_file = sprintf("%8.8d", $record_number) . ".nul";
-        open(DUMMY_FILE, ">$documents_directory/$document_file");
-        close(DUMMY_FILE);
-        &write_metadata_xml_file_entry(METADATA_XML_FILE, $document_file, $record_metadata, $metadata_set);
-    }
-
-    if (($record_number % $records_per_folder) == 0 || $record_number == scalar(@metadata_records)) {
-        # Finish and close the metadata.xml file
-        print METADATA_XML_FILE "\n</DirectoryMetadata>\n";
-        close(METADATA_XML_FILE);
-    }
-    $record_number = $record_number + 1;
-    }
+
+    $plugobj->extra_metadata ($doc_obj, $doc_obj->get_top_section(), $metadata);
+
+    # Try to get a doc to attach the metadata to
+    # If no match found, create a dummy .nul file
+    attach_metadata_or_make_nul_doc($document_field, $doc_obj, undef, 
+                    $documents_directory, $orig_base_dir,
+                    $document_prefix, $document_suffix, $metadata_set, $verbosity);
+
+
+    close_directory();
+    }
+
 
     # Explode means just that: the original file is deleted
     &util::rm($filename);
     $plugobj->clean_up_after_exploding();
-}
+
+}
+
+
+sub need_new_directory
+{
+    my ($exploded_base_dir) = @_;
+    
+    my $documents_directory = $exploded_base_dir;
+
+    if (-d $documents_directory) {
+    die "Error: document directory $documents_directory already exists (bailing).\n";
+    }
+    &util::mk_dir($documents_directory);
+
+    my $documents_metadata_xml_file = &util::filename_cat($documents_directory, "metadata.xml");
+    if (-e $documents_metadata_xml_file) {
+    die "Error: documents metadata.xml file $documents_metadata_xml_file already exists (bailing).\n";
+    }
+
+    # Start the metadata.xml file
+    open(METADATA_XML_FILE, ">$documents_metadata_xml_file");
+    print METADATA_XML_FILE
+    "<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"no\"?>\n" .
+    "<!DOCTYPE DirectoryMetadata SYSTEM \"http://greenstone.org/dtd/DirectoryMetadata/1.0/DirectoryMetadata.dtd\">\n" .
+    "<DirectoryMetadata>\n";
+
+    return $documents_directory;
+}
+
+sub check_need_new_directory
+{
+    my ($exploded_base_dir,$record_number, $records_per_folder,$metadata_records,
+    $documents_dir_ref) = @_;
+    
+
+    # Check if we need to start a new directory for these records
+    if (($record_number % $records_per_folder) == 1) {
+    my $documents_directory = $exploded_base_dir;
+
+    if (scalar(@$metadata_records) > $records_per_folder) {
+        $documents_directory .= "." . sprintf("%8.8d", $record_number);
+    }
+
+    $$documents_dir_ref = need_new_directory($documents_directory);
+    }
+}
+
+
+
+
+
+sub attach_metadata_or_make_nul_doc
+{
+    my ($document_field, $doc_obj, $record_number, 
+    $documents_directory, $orig_base_dir,
+    $document_prefix, $document_suffix, $metadata_set, $verbosity) = @_;
+
+    my $record_metadata = $doc_obj->get_all_metadata($doc_obj->get_top_section());
+    my $document_file;
+
+    # try to get a doc to attach the metadata to
+    if (defined $document_field) {
+    foreach my $pair (@$record_metadata) {
+        my ($field, $value) = (@$pair);
+
+        $value =~ s/\\\\/\\/g;
+        
+        # Does this metadata element specify a document to obtain?
+        if ($field eq $document_field) {
+        my $document_file_full = $document_prefix . $value . $document_suffix;
+
+        $document_file = &obtain_document($document_file_full, $documents_directory, $orig_base_dir, $verbosity);
+        &write_metadata_xml_file_entry(METADATA_XML_FILE, $document_file, $record_metadata, $metadata_set);
+        }
+    }
+    }
+    
+    # Create a dummy .nul file if we haven't obtained any documents for this record
+    if (not defined $document_file) {
+
+    if (defined ($record_number)) {
+        $document_file = sprintf("%8.8d", $record_number) . ".nul";
+    }
+    else {
+        $document_file = "doc.nul";
+    }
+
+    open(DUMMY_FILE, ">$documents_directory/$document_file");
+    close(DUMMY_FILE);
+    &write_metadata_xml_file_entry(METADATA_XML_FILE, $document_file, $record_metadata, $metadata_set);
+    }
+
+}
+
+sub close_directory
+{
+    # Finish and close the metadata.xml file
+    print METADATA_XML_FILE "\n</DirectoryMetadata>\n";
+    close(METADATA_XML_FILE);
+
+}
+
+
+sub check_close_directory
+{
+    my ($record_number,$records_per_folder,$metadata_records) = @_;
+
+    if (($record_number % $records_per_folder) == 0 || $record_number == scalar(@$metadata_records)) {
+    # Finish and close the metadata.xml file
+    close_directory();
+    }
+}
+        
 
 
@@ -322 +445 @@
 sub obtain_document
 {
-    my $document_file_full = shift(@_);
-    my $documents_directory = shift(@_);
-    my $verbosity = shift(@_);
+    my ($document_file_full,$documents_directory,$orig_base_dir,$verbosity) = @_;
     
     print STDERR "Obtaining document file $document_file_full...\n" if ($verbosity > 1);
@@ -332 +453 @@
 
     # Document specified is on the web
-    if ($document_file_full =~ /^http:/ || $document_file_full =~ /^ftp:/) {
+    if ($document_file_full =~ /^https?:/ || $document_file_full =~ /^ftp:/) {
     $document_file_full =~ /([^\/]+)$/;
     $document_file_name = $1;
@@ -351 +472 @@
     else {
     my $dir_sep = &util::get_os_dirsep();
-    $document_file_full =~ /(.+$dir_sep)?(.*)$/;
+
+    $document_file_full =~ m/(.+$dir_sep)?(.*)$/;
     $document_file_name = $2;
+
+
+    my $is_absolute = File::Spec->file_name_is_absolute($document_file_full);
+    print STDERR "doc file full = $document_file_full\n";
+
+    if (!$is_absolute) {
+        $document_file_full 
+        = &util::filename_cat($orig_base_dir,$document_file_full);
+    }
+
     $local_document_file = &util::filename_cat($documents_directory, $document_file_name);
 
-    # Only bother trying to copy the file if it contained some path information
-    if ($document_file_full ne $document_file_name) {
-        &util::cp($document_file_full, $documents_directory);
-
-        # Check the document was obtained successfully
-        if (!-e $local_document_file) {
-        print STDERR "WARNING: Could not obtain document file $document_file_full\n";
+    &util::cp($document_file_full, $documents_directory);
+
+    # Check the document was obtained successfully
+    if (!-e $local_document_file) {
+        print STDERR "WARNING: Could not obtain document file $document_file_full\n";
+    }
+    else {
+        if ($document_file_full =~ m/^$orig_base_dir.*/) {
+        # file local to metadata record
+        # => copy has been made successfully, so remove original
+        &util::rm($document_file_full);
         }
     }
@@ -377 +513 @@
 
 &main(@ARGV);
+