- fixed and improved reppid findings

- use compiled xpaths
- use precompiled xpathcontext
- use nodeValue() instead triggering findvalue()
- handle purged IEs
- hande all variants of localreps (LOCAL and LZA and LZA_INTERN)

perl/exit_strategy.pl

@@ -233,28 +233,11 @@ sub check_if_db_conform ($$) {
   return;
 }
 
-
-###############################################################################
-#
-# /mets:mets/mets:dmdSec[1]/mets:mdWrap[1]/mets:xmlData[1]/dc:record[1]/dc:title[1]
-# /mets:mets/mets:amdSec[1]/mets:techMD[1]/mets:mdWrap[1]/mets:xmlData[1]/dnx[1]/section[1]/record[1]/key[2]
-# mit ID=Label und Wert = LOCAL
-# dort die ID von techMD (Referenz für Files)
-#
-# Files via /mets:mets/mets:fileSec[1]/mets:fileGrp[1]/mets:file[1]/mets:FLocat[1]
-#
-###############################################################################
-sub parse_iexml ($$) {
-  my $filename = shift;
-  my $recovery_flag = shift;
-    # create object
-    #
-    #my $xp = XML::XPath->new (filename => $filename);
-  if ($recovery_flag) {
-    $recoverflag=2; # avoid warnings, see XML::LibXML::Parser POD about 'recovery'
-  }	
-  my $dom = XML::LibXML->load_xml (location => $filename, recover => $recovery_flag, no_blanks=>1, compact=>1);
-  my $xp = XML::LibXML::XPathContext->new($dom);
+{
+  my $xp;
+  sub get_xpath_context {
+    if (defined $xp) { return $xp};
+    $xp = XML::LibXML::XPathContext->new();
     $xp->registerNs("dnx", "http://www.exlibrisgroup.com/dps/dnx");
     $xp->registerNs("sru", "http://www.loc.gov/zing/srw/");
     $xp->registerNs("xsi", "http://www.w3.org/2001/XMLSchema-instance");
@@ -270,20 +253,62 @@ sub parse_iexml ($$) {
     $xp->registerNs("mix", "http://www.loc.gov/standards/mix/");
     $xp->registerNs("xlink", "http://www.w3.org/1999/xlink");
     $xp->registerNs("xlin", "http://www.w3.org/1999/xlink");
+    return $xp;
+  }
+}
 
+###############################################################################
+#
+# /mets:mets/mets:dmdSec[1]/mets:mdWrap[1]/mets:xmlData[1]/dc:record[1]/dc:title[1]
+# /mets:mets/mets:amdSec[1]/mets:techMD[1]/mets:mdWrap[1]/mets:xmlData[1]/dnx[1]/section[1]/record[1]/key[2]
+# mit ID=Label und Wert = LOCAL
+# dort die ID von techMD (Referenz für Files)
+#
+# Files via /mets:mets/mets:fileSec[1]/mets:fileGrp[1]/mets:file[1]/mets:FLocat[1]
+#
+###############################################################################
+{
+  my $compiled_xpath_titles = XML::LibXML::XPathExpression->new('/mets:mets/mets:dmdSec/mets:mdWrap[1]/mets:xmlData[1]/dc:record/dc:title[1]');
+  my $compiled_xpath_dcrecords = XML::LibXML::XPathExpression->new('/mets:mets/mets:dmdSec/mets:mdWrap/mets:xmlData/dc:record/*');
+  my $compiled_xpath_amdsecs = XML::LibXML::XPathExpression->new('/mets:mets/mets:amdSec[starts-with(@ID, \'REP\')]');
+  my $str_local_reps = 'mets:techMD/mets:mdWrap/mets:xmlData/*[local-name()=\'dnx\']/*[local-name()=\'section\']/*[local-name()=\'record\']/*[local-name()=\'key\' and @id=\'label\']';
+  my $compiled_xpath_localreps = XML::LibXML::XPathExpression->new( $str_local_reps);
+  my $compiled_xpath_localreps2 = XML::LibXML::XPathExpression->new('/mets:mets/mets:amdSec/mets:techMD/mets:mdWrap/mets:xmlData/dnx/section/record/key[@id=\'label\']=\'LOCAL\'');
+  my $compiled_xpath_filegrps = XML::LibXML::XPathExpression->new('/mets:mets/mets:fileSec/mets:fileGrp');
+  my $compiled_xpath_flocat = XML::LibXML::XPathExpression->new('mets:file/mets:FLocat');
+  my $compiled_xpath_id = XML::LibXML::XPathExpression->new('@ID');
+  my $compiled_xpath_admid = XML::LibXML::XPathExpression->new('@ADMID');
+  my $compiled_xpath_xlinhref = XML::LibXML::XPathExpression->new('@xlin:href');
+  my $compiled_xpath_dot = XML::LibXML::XPathExpression->new('.');
+  sub parse_iexml($$) {
+    my $filename = shift;
+    my $recovery_flag = shift;
+    if ($recovery_flag) {
+      $recovery_flag = 2; # avoid warnings, see XML::LibXML::Parser POD about 'recovery'
+    }
+    my $dom = XML::LibXML->load_xml(
+        location  => $filename,
+        recover   => $recovery_flag,
+        no_blanks => 1,
+        compact   => 1,
+        no_network    => 1,
+    );
+    my $xp = get_xpath_context();
+    $xp->setContextNode($dom);
     ############################################
     # get title
-  my $compiled_xpath_titles = '/mets:mets/mets:dmdSec/mets:mdWrap[1]/mets:xmlData[1]/dc:record/dc:title[1]';
     my $title = $xp->findvalue($compiled_xpath_titles);
     check_if_db_conform($title, $filename);
     ############################################
     # get dc-records
     my @dcrecords;
-  my $compiled_xpath_dcrecords='/mets:mets/mets:dmdSec/mets:mdWrap/mets:xmlData/dc:record/*';
+
     my $dcnodes = $xp->find($compiled_xpath_dcrecords);
     foreach my $dcnode ($dcnodes->get_nodelist) {
+      #my $ref = ref $dcnode; use Data::Printer; p( $ref);
       my $key = $dcnode->getName();
-    my $value = $dcnode->findvalue(".");
+      #my $value = $dcnode->findvalue($compiled_xpath_dot);
+      my $value = $dcnode->nodeValue;
       if (defined $value) {
         $value =~ s/\n/ /g;
         $value =~ s/'/\\'/g;
@@ -296,41 +321,46 @@ sub parse_iexml ($$) {
     }
     ############################################
     # get right representation ID (has a dnx-section with <key id=label>LOCAL</key>)
-  my $compiled_xpath_amdsecs = '/mets:mets/mets:amdSec';
-  my $compiled_xpath_localreps = 'mets:techMD/mets:mdWrap/mets:xmlData/dnx/section/record/key[@id=\'label\']';
-  my $compiled_xpath_filegrps = '/mets:mets/mets:fileSec/mets:fileGrp';
-  my $compiled_xpath_flocat = 'mets:file/mets:FLocat';
-
-  my $repids = $xp->find($compiled_xpath_amdsecs);
+    my $repids = $xp->find($compiled_xpath_amdsecs); #/mets:mets/mets:amdSec
     my $repid;
-  # FIXME: if only one represenation exists (Qucosa), select this. If there
-  # are more than one, use them with label LOCAL
     my @repnodes = $repids->get_nodelist;
-
-  $repid = $repnodes[0]->findvalue('@ID' );
+    if (scalar @repnodes == 0) {
+      say STDERR "No reppid found in file $filename, is IE purged?";
+      my %tmp;
+      $tmp{"filename"}=$filename;
+      $tmp{"purged"}=1;
+      return \%tmp;
+    } elsif (scalar @repnodes == 1) {
+      $repid = $repnodes[0]->findvalue($compiled_xpath_id);
+    } else { #multiple representations found
+      # choose reppid with LZA, LZA_INTERN or LOCAL (for very old IEs)
       foreach my $node (@repnodes) {
-    my $id = $node->findvalue('@ID' );
+        my $id = $node->findvalue($compiled_xpath_id);
         check_if_db_conform($id, $filename);
-    #/mets:mets/mets:amdSec[1]/mets:techMD[1]/mets:mdWrap[1]/mets:xmlData[1]/dnx[1]/section[1]/record[1]/key[1]
-    #
-    
-    if ($node->findvalue($compiled_xpath_localreps) eq 'LOCAL') {
+        my $localreps = $node->findvalue($compiled_xpath_localreps);
+        if (
+            ($localreps eq 'LOCAL')
+                or ($localreps eq 'LZA')
+                or ($localreps eq 'LZA_INTERN')
+        ) {
           $repid = $id;
+          last;
         }
         #print XML::XPath::XMLParser::as_string($node), "\n\n";
       }
+    }
+    if (!defined $repid) {
+      say "No repid found in file $filename";
+    }
     ############################################
     # get all files of LOCAL representation
     my @files;
     my $filegrpnodes = $xp->find($compiled_xpath_filegrps);
     foreach my $filegrpnode ($filegrpnodes->get_nodelist) {
-    #die XML::XPath::XMLParser::as_string($filegrpnode), "\n\n";
-    #die Dumper($filegrpnode);
-    if ($filegrpnode->findvalue('@ADMID') eq $repid) {
-      #die Dumper($filegrpnode);
+      if ($filegrpnode->findvalue($compiled_xpath_admid) eq $repid) {
         my $filesnodes = $filegrpnode->find($compiled_xpath_flocat);
         foreach my $filesnode ($filesnodes->get_nodelist) {
-        my $value = $filesnode->findvalue('@xlin:href');
+          my $value = $filesnode->findvalue($compiled_xpath_xlinhref);
           check_if_db_conform($value, $filename);
           push @files, sprintf("%s", $value);
         }
@@ -344,6 +374,7 @@ sub parse_iexml ($$) {
     $ret{"dcrecords"} = \@dcrecords;
     return \%ret;
   }
+}
 
 ###############################################################################
 # because ExLibris Rosetta produces filenames of following format:
@@ -398,6 +429,7 @@ sub searching_ie_files ($$) {
   my $tmp_ies_unsorted_file = shift;
   my $cnt_unsorted_files = 0;
   my $first_two_levels_of_dirs = 0;
+  ###
   my $wanted_twolevel_dircount = sub {
     my $relpath = $File::Find::name;
     $relpath =~ s{^\Q$dir\E/?}{};
@@ -407,9 +439,11 @@ sub searching_ie_files ($$) {
         and $File::Find::prune = 1;
     if (-d $_) { $first_two_levels_of_dirs++;}
   };
+  ###
   find( $wanted_twolevel_dircount, $dir);
   my $progressbar=Time::Progress->new(min => 0, max => $first_two_levels_of_dirs, smoothing => 1);
   my $dircount = 0;
+  ###
   my $wanted_process_sip = sub  {
     if (-f && m/V(\d+)-IE\d+\.xml$/) {
       my $version = $1;
@@ -430,6 +464,7 @@ sub searching_ie_files ($$) {
     }
     return;
   };
+  ###
   find($wanted_process_sip, $dir);
   say "";
   return $cnt_unsorted_files;