Changeset 1287 for trunk/gsdl/perllib/docsave.pm

Timestamp:

2000-07-14T12:24:20+12:00 (24 years ago)

Author:

sjboddie

Message:

Implemented a -sortmeta option for import.pl to sort archives.inf file
(generated at end of import process) alphabetically by the given
metadata element. This may be useful for some collections as boolean
queries currently return matches in build (fairly random) order. Changing
the order of archives.inf changes the order that documents are built.
This option has a couple of important limitations:

1. Can't be used in conjunction with the groupsize option as it would then only change the build order of groups of documents which doesn't seem very useful.
2. Is of limited use when building indexes at a section level as the build order is only sorted by document, not by section.

File:

• trunk/gsdl/perllib/docsave.pm (modified) (5 diffs)

trunk/gsdl/perllib/docsave.pm

@@ -56 +56 @@
     $self->{'archive_dir'} = "$ENV{'GSDLHOME'}/collect/$self->{'collection'}/archives";
 
+    $self->{'sortmeta'} = undef;
+
     return bless $self, $class;
 }
@@ -66 +68 @@
 }
 
+sub set_sortmeta {
+    my $self = shift (@_);
+    my ($sortmeta) = @_;
+
+    $self->{'sortmeta'} = $sortmeta;
+}
+
 sub process {
     my $self = shift (@_);
     my ($doc_obj) = @_;
 
-    my $archive_dir = $self->{'archive_dir'};
+    if ($self->{'groupsize'} > 1) {
+    $self->group_process ($doc_obj);
+
+    } else {
+    # groupsize is 1 (i.e. one document per GML file) so sortmeta 
+    # may be used
+
+    my $OID = $doc_obj->get_OID();
+    $OID = "NULL" unless defined $OID;
+    
+    # get document's directory
+    my $doc_dir = $self->get_doc_dir ($OID);
+    
+    # copy all the associated files, add this information as metadata
+    # to the document
+    $self->process_assoc_files ($doc_obj, $doc_dir);
+
+    my $doc_file 
+        = &util::filename_cat ($self->{'archive_dir'}, $doc_dir, "doc.gml");
+    my $short_doc_file = &util::filename_cat ($doc_dir, "doc.gml");
+        
+    if (!open (OUTDOC, ">$doc_file")) {
+        print STDERR "docsave::process could not write to file $doc_file\n";
+        return;
+    }
+
+    # save this document
+    $doc_obj->output_section('docsave::OUTDOC', $doc_obj->get_top_section());
+    close OUTDOC;
+
+    if ($self->{'gzip'}) {
+        my $doc_file = $self->{'gs_filename'};
+        `gzip $doc_file`;
+        $doc_file .= ".gz";
+        $short_doc_file .= ".gz";
+        if (!-e $doc_file) {
+        print STDERR "error while gzipping: $doc_file doesn't exist\n";
+        return 0;
+        }
+    }
+
+    # do the sortmeta thing
+    my ($metadata);
+    if (defined ($self->{'sortmeta'})) {
+        $metadata = $doc_obj->get_metadata_element ($doc_obj->get_top_section(), $self->{'sortmeta'});
+    }
+
+    # store reference in the archive_info
+    $self->{'archive_info'}->add_info($OID, $short_doc_file, $metadata);
+    }
+}
+
+sub group_process {
+    my $self = shift (@_);
+    my ($doc_obj) = @_;
+
     my $OID = $doc_obj->get_OID();
     $OID = "NULL" unless defined $OID;
@@ -79 +143 @@
 
     # opening a new file, or document has assoicated files => directory needed
-    if (($open_new_file) || (scalar(@{$doc_obj->get_assoc_files()})>0))
-    {
-    # get the document's directory.
-    my $doc_info = $self->{'archive_info'}->get_info($OID);
-    my $doc_dir = "";
-    if (defined $doc_info && scalar(@$doc_info) >= 1) {
-        # this OID already has an assigned directory, use the
-        # same one.
-        $doc_dir = $doc_info->[0];
-        $doc_dir =~ s/\/?doc\.gml(\.gz)?$//;
-    } else {
-        # have to get a new document directory
-        my $doc_dir_rest = $OID;
-        my $doc_dir_num = 0;
-        do {
-        $doc_dir .= "/" if $doc_dir_num > 0;
-        if ($doc_dir_rest =~ s/^(.{1,8})//) {
-            $doc_dir .= $1;
-            $doc_dir_num++;
-        }
-        } while ($doc_dir_rest ne "" && 
-             ((-d &util::filename_cat ($archive_dir, "$doc_dir.dir")) || 
-              ($self->{'archive_info'}->size() >= 1024 && $doc_dir_num < 2)));
-        $doc_dir .= ".dir";
-        
-    }
-    
-    &util::mk_all_dir ("$archive_dir/$doc_dir");
-    
+    if (($open_new_file) || (scalar(@{$doc_obj->get_assoc_files()})>0)) {
+
+    # get document's directory
+    my $doc_dir = $self->get_doc_dir ($OID);
+
     # copy all the associated files, add this information as metadata
     # to the document
-    my @assoc_files = ();
-    foreach $assoc_file (@{$doc_obj->get_assoc_files()}) {
-        my ($dir, $afile) = $assoc_file->[1] =~ /^(.*?)([^\/\\]+)$/;
-        $dir = "" unless defined $dir;
-        if (-e $assoc_file->[0]) {
-        my $filepath = &util::filename_cat($archive_dir, $doc_dir, $afile);
-        &util::hard_link ($assoc_file->[0], $filepath);
-        $doc_obj->add_metadata ($doc_obj->get_top_section(),
-                    "gsdlassocfile",
-                    "$afile:$assoc_file->[2]:$dir");
-        } else {
-        print STDERR "docsave::process couldn't copy the associated file " .
-            "$assoc_file->[0] to $afile\n";
-        }
-    }
-
-    if ($open_new_file)
-    {
+    $self->process_assoc_files ($doc_obj, $doc_dir);
+
+
+    if ($open_new_file) {
         # only if opening new file
         my $doc_file 
-        = &util::filename_cat ($archive_dir, $doc_dir, "doc.gml");
+        = &util::filename_cat ($self->{'archive_dir'}, $doc_dir, "doc.gml");
         my $short_doc_file = &util::filename_cat ($doc_dir, "doc.gml");
         
@@ -139 +165 @@
         
         if (!open (OUTDOC, ">$doc_file")) {
-        print STDERR "docsave::process could not write to file $doc_file\n";
+        print STDERR "docsave::group_process could not write to file $doc_file\n";
         return;
         }
@@ -153 +179 @@
     $self->{'gs_count'}++;
 }
+
+
+sub get_doc_dir {
+    my $self = shift (@_);
+    my ($OID) = @_;
+
+    my $doc_info = $self->{'archive_info'}->get_info($OID);
+    my $doc_dir = "";
+    if (defined $doc_info && scalar(@$doc_info) >= 1) {
+    # this OID already has an assigned directory, use the
+    # same one.
+    $doc_dir = $doc_info->[0];
+    $doc_dir =~ s/\/?doc\.gml(\.gz)?$//;
+    } else {
+    # have to get a new document directory
+    my $doc_dir_rest = $OID;
+    my $doc_dir_num = 0;
+    do {
+        $doc_dir .= "/" if $doc_dir_num > 0;
+        if ($doc_dir_rest =~ s/^(.{1,8})//) {
+        $doc_dir .= $1;
+        $doc_dir_num++;
+        }
+    } while ($doc_dir_rest ne "" && 
+         ((-d &util::filename_cat ($self->{'archive_dir'}, "$doc_dir.dir")) || 
+          ($self->{'archive_info'}->size() >= 1024 && $doc_dir_num < 2)));
+    $doc_dir .= ".dir";
+    
+    }
+    
+    &util::mk_all_dir (&util::filename_cat ($self->{'archive_dir'}, $doc_dir));
+
+    return $doc_dir;
+}
+
+
+sub process_assoc_files {
+    my $self = shift (@_);
+    my ($doc_obj, $doc_dir) = @_;
+
+    my @assoc_files = ();
+    foreach $assoc_file (@{$doc_obj->get_assoc_files()}) {
+    my ($dir, $afile) = $assoc_file->[1] =~ /^(.*?)([^\/\\]+)$/;
+    $dir = "" unless defined $dir;
+    if (-e $assoc_file->[0]) {
+        my $filepath = &util::filename_cat($self->{'archive_dir'}, $doc_dir, $afile);
+        &util::hard_link ($assoc_file->[0], $filepath);
+        $doc_obj->add_utf8_metadata ($doc_obj->get_top_section(),
+                     "gsdlassocfile",
+                     "$afile:$assoc_file->[2]:$dir");
+    } else {
+        print STDERR "docsave::process couldn't copy the associated file " .
+        "$assoc_file->[0] to $afile\n";
+    }
+    }
+}
+    
 
 sub close_file_output