Changeset 28251 for main/trunk/greenstone2

Timestamp:

2013-09-11T14:11:32+12:00 (11 years ago)

Author:

davidb

Message:

Change to calculation of curr_sec_num so it correctly pops off values as a series of </Section> tags are encountered

File:

• main/trunk/greenstone2/perllib/cgiactions/docextractaction.pm (modified) (13 diffs)

main/trunk/greenstone2/perllib/cgiactions/docextractaction.pm

@@ -46 +46 @@
     "extract-archives-doc" => { # where param can be ONE of: index (default), import, archives, live
         'compulsory-args' => [ "d", "json-sections" ],
-        'optional-args'   => [ "json-metadata", "newd" ],
+        'optional-args'   => [ "json-metadata", "newd", 
+                   "keep-parent-metadata", "keep-parent-content" ],
 #       'optional-args'   => [ "where" ],
         'help-string' => [
@@ -96 +97 @@
 
     if ($new_depth > $old_depth) {
-    # child subsection
+    # first child subsection
     $new_secnum = "$old_secnum.1";
     }
@@ -108 +109 @@
     else {
     # back up to parent section => lopp off tail
-### print STDERR "**### back up to parent section, $old_secnum = $old_secnum\n";
-
-    $new_secnum = $old_secnum;
-    $new_secnum =~ s/\.\d+$//;
+#   $new_secnum = $old_secnum;
+#   $new_secnum =~ s/\.\d+$//;
     }
 
@@ -125 +124 @@
     my ($tagname, $attrHash, $contextArray, $parentDataArray, $parser) = @_;
 
+    my $curr_sec_num = $parser->{'parameters'}->{'curr_section_num'} || undef;
+
     my $sec_num_hash = $parser->{'parameters'}->{'sec_num_hash'};
-    my $curr_sec_num = $parser->{'parameters'}->{'curr_section_num'} || undef;
+    my $parent_sec_num_hash = $parser->{'parameters'}->{'parent_sec_num_hash'};
+    
+    my $keep_parent_metadata = $parser->{'parameters'}->{'keep_parent_metadata'};
+    my $keep_parent_content  = $parser->{'parameters'}->{'keep_parent_content'};
 
     my $mode = $parser->{'parameters'}->{'mode'};
 
-    my $depth = $parser->{'parameters'}->{'curr_section_depth'};
-
+    my $prev_depth = $parser->{'parameters'}->{'curr_section_depth'};
     my $live_depth = scalar(@$contextArray);
 
-#   print STDERR "closing secdtion tag, mode: $mode, curr sec num = $curr_sec_num, live depth = $live_depth\n";
+    if ($live_depth < $prev_depth) {
+        # In a closing-sections poping off situation:
+        #   </Section>
+        # </Section>
+
+        # => Back up to parent section => lopp off tail
+
+        $curr_sec_num =~ s/\.\d+$//;
+        $parser->{'parameters'}->{'curr_section_depth'} = $live_depth;
+        $parser->{'parameters'}->{'curr_section_num'}   = $curr_sec_num;
+    }
+
 
     if ($live_depth == 1) {
         # root sectin tag, which must always exist
-#       print STDERR "*** root, tagname = $tagname, attrHash = $attrHash ", 
-#         " '_content' = ", join(",",@{$attrHash->{'_content'}}), "\n";
         return [$tagname => $attrHash];
     }
-    elsif (defined $sec_num_hash->{$curr_sec_num}) {
-###     print STDERR "*** got a match on $curr_sec_num, mode = $mode\n";
-
-        if ($mode eq "extract") {
-        # keep it
-        return [$tagname => $attrHash];
-        }
-        else {
-        # remove
-        return undef;
-        }
-    }
-    else {
-        # not in our list
-        if ($mode eq "extract") {
-        # remove
-        return undef;
+    elsif ($mode eq "delete") {
+        if (defined $sec_num_hash->{$curr_sec_num}) {
+        # remove it
+        return undef
         }
         else {
@@ -165 +164 @@
         }
     }
+    else {
+        # mode is extract
+
+        if (defined $sec_num_hash->{$curr_sec_num}) {
+        # keep it
+##      print STDERR "**** Asked to keep: sec num = $curr_sec_num\n";
+
+        return [$tagname => $attrHash];
+        }
+        elsif (defined $parent_sec_num_hash->{$curr_sec_num}) {
+        # want this element, but cut down to just the child <Section>
+        
+        my $section_child = undef;
+
+##      print STDERR "**** Parent match: sec num = $curr_sec_num\n";
+
+        my $filtered_elems = [];
+        
+        foreach my $elem ( @{$attrHash->{'_content'}}) {
+            if (ref $elem eq "ARRAY") {
+            my $child_tagname = $elem->[0];
+##          print STDERR "***## elem name $child_tagname\n";
+
+            
+            if ($child_tagname eq "Description") {
+                if ($keep_parent_metadata) {
+                push(@$filtered_elems,$elem);
+                }
+            }
+            elsif ($child_tagname eq "Content") {
+                if ($keep_parent_content) {
+                push(@$filtered_elems,$elem);
+                }
+            }
+            else {
+                push(@$filtered_elems,$elem);
+            }
+            }
+            else {
+            push(@$filtered_elems,$elem);
+            }
+        }
+
+        $attrHash->{'_content'} = $filtered_elems;
+
+        return [$tagname => $attrHash];
+
+        }
+        else {
+        # not in our list => remove it
+        return undef;
+        }
+    }
 }
 
@@ -171 +223 @@
 {
     my $self = shift @_;
-    my ($gsdl_cgi, $doc_xml_filename, $newdoc_xml_filename, $sec_num_hash, $mode) = @_;
+    my ($gsdl_cgi, $doc_xml_filename, $newdoc_xml_filename, 
+        $sec_num_hash, $parent_sec_num_hash, $mode) = @_;
     
     my @start_rules = ('Section' => \&dxml_start_section);
@@ -219 +272 @@
             binmode($MOUT,":utf8");
 
-            my $options = {sec_num_hash => $sec_num_hash, mode => $mode };
+            my $options = { sec_num_hash         => $sec_num_hash, 
+                    parent_sec_num_hash  => $parent_sec_num_hash,
+                    keep_parent_metadata => $self->{'keep-parent-metadata'},
+                    keep_parent_content  => $self->{'keep-parent-content'},
+                        mode => $mode };
 
             $parser->filter($xml_in, $MOUT, $options);
@@ -231 +288 @@
 {
     my $self = shift @_;
-
-    my ($json_sections_array,$mode) = @_;
+    
+    my ($json_sections_array) = @_;
 
     my $sec_num_hash = {};
@@ -241 +298 @@
     # Need to do the same here, so things can be matched up
     $sec_num_hash->{".$sn"} = 1; 
-
-### print STDERR "** storeing .$sn\n";
-
-    if ($mode eq "with-parents") {
-        my $sn_copy = $sn; # needs to be a copy, otherwise chaning version stored in json_sections
-        while ($sn_copy =~ s/\.\d+$//) {
-        $sec_num_hash->{".$sn_copy"} = 1; # See '.' comment above
-
-####        print STDERR "** ***** parent storeing .$sn_copy\n";
-        }
-    }
     }
 
@@ -258 +304 @@
 
 
+sub parent_sections_as_hash
+{
+    my $self = shift @_;
+
+    my ($json_sections_array) = @_;
+
+    my $sec_num_hash = {};
+
+    foreach my $sn ( @$json_sections_array ) {
+
+    # needs to make a copy, otherwise version stored in json_sections gets changed
+    my $sn_copy = $sn; 
+    while ($sn_copy =~ s/\.\d+$//) {
+        # our XML parser curr_sec_num puts '.' at the root
+        # Need to do the same here, so things can be matched up
+
+        $sec_num_hash->{".$sn_copy"} = 1; 
+    }
+    }
+
+    return $sec_num_hash;
+}
+
+sub parse_flag
+{
+        my $self = shift @_;
+
+    my ($arg_name) = @_;
+
+    my $flag = $self->{$arg_name} || 0;
+
+    $flag =~ s/^true/1/i;
+    $flag =~ s/^false/0/i;
+
+    return $flag;
+}
 
 sub _extract_archives_doc
@@ -278 +360 @@
     my $new_docid = $self->{'newd'} || "HASH" . localtime(time);
 
+    $self->{'keep-parent-metadata'} = $self->parse_flag("keep-parent-metadata");
+    $self->{'keep-parent-content'}  = $self->parse_flag("keep-parent-content");
+
     my $json_sections_str = $self->{'json-sections'};
     my $json_sections_array = decode_json($json_sections_str);
@@ -293 +378 @@
 #   my $doc_filename = $doc_rec->{'doc-file'}->[0]; 
 
-    my $extract_sec_num_hash = $self->sections_as_hash($json_sections_array,"with-parents");
-
-    my $extract_status = $self->remove_from_doc_xml($gsdl_cgi, $doc_filename ,$newdoc_filename, $extract_sec_num_hash, "extract");
+    my $sec_num_hash = $self->sections_as_hash($json_sections_array);
+    my $parent_sec_num_hash = $self->parent_sections_as_hash($json_sections_array);
+
+    my $extract_status = $self->remove_from_doc_xml($gsdl_cgi, $doc_filename ,$newdoc_filename, $sec_num_hash, $parent_sec_num_hash, "extract");
     
     if ($extract_status == 0) 
@@ -301 +387 @@
         my $delete_sec_num_hash = $self->sections_as_hash($json_sections_array,"no-parents");
 
-        my $delete_status = $self->remove_from_doc_xml($gsdl_cgi, $doc_filename ,$doc_filename, $delete_sec_num_hash, "delete");
+        my $delete_status = $self->remove_from_doc_xml($gsdl_cgi, $doc_filename ,$doc_filename, $sec_num_hash, undef, "delete");
 
         if ($delete_status == 0) {