X-Git-Url: http://git.equinoxoli.org/?p=migration-tools.git;a=blobdiff_plain;f=unicorn%2Funicorn_patrons_to_tsv.pl;h=636c21a3c75dd36d02bd7c3c9caf0abfd611309a;hp=6960d4998551a64390b5d7e3ed1dda861ef53071;hb=5f358632626dc390ef79acbe304c6cceebe6fc3f;hpb=a8016c4258f1e4ca810a00070fe81ee30daf493d diff --git a/unicorn/unicorn_patrons_to_tsv.pl b/unicorn/unicorn_patrons_to_tsv.pl index 6960d49..636c21a 100755 --- a/unicorn/unicorn_patrons_to_tsv.pl +++ b/unicorn/unicorn_patrons_to_tsv.pl @@ -1,14 +1,51 @@ #!/usr/bin/perl -w +# Copyright 2009-2012, Equinox Software, Inc. +# +# This program is free software; you can redistribute it and/or +# modify it under the terms of the GNU General Public License +# as published by the Free Software Foundation; either version 2 +# of the License, or (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program; if not, write to the Free Software +# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. + # Converts a Unicorn users.data file to a tab-separated file. # 2009-08-10 Ben Ostrowsky -my @users; +#use Data::Dumper; +#$Data::Dumper::Sortkeys = true; + +my @records; my $serial = -1; my $line = 0; my $section = ''; my $field = ''; my %unique_fields; +my @output_fields = qw( +l_user_id l_user_altid l_user_pin l_user_profile l_user_status l_user_library l_user_priv_granted +l_user_priv_expires l_user_mailingaddr l_birthdate l_prefix_name l_last_name l_first_name l_middle_name +l_suffix_name l_note l_note1 l_patron l_comment l_staff l_webcatpref l_user_category1 l_user_category2 +l_user_category3 l_user_category4 l_dept l_guardian l_license l_ssn l_misc l_aup l_photo l_notify_via +l_user_claims_ret l_user_environment l_user_department l_ums_id l_user_last_activity l_placcard l_user_email +l_addr1_std_line1 l_addr1_std_line2 l_addr1_std_city l_addr1_std_state l_addr1_std_zip l_addr1_country l_addr1_county +l_addr1_township l_addr1_room l_addr1_company l_addr1_office l_addr1_phone l_addr1_dayphone l_addr1_homephone +l_addr1_workphone l_addr1_cellphone l_addr1_fax l_addr1_email l_addr1_location l_addr1_usefor l_addr1_care_of +l_addr1_known_bad l_addr1_ums_addrid l_addr2_std_line1 l_addr2_std_line2 l_addr2_std_city l_addr2_std_state +l_addr2_std_zip l_addr2_country l_addr2_county l_addr2_township l_addr2_room l_addr2_company l_addr2_office l_addr2_phone +l_addr2_dayphone l_addr2_homephone l_addr2_workphone l_addr2_cellphone l_addr2_fax l_addr2_email +l_addr2_location l_addr2_usefor l_addr2_care_of l_addr2_known_bad l_addr2_ums_addrid l_addr3_std_line1 +l_addr3_std_line2 l_addr3_std_city l_addr3_std_state l_addr3_std_zip l_addr3_country l_addr3_county l_addr3_township +l_addr3_room l_addr3_company l_addr3_office l_addr3_phone l_addr3_dayphone l_addr3_homephone l_addr3_workphone +l_addr3_cellphone l_addr3_fax l_addr3_email l_addr3_location l_addr3_usefor l_addr3_care_of l_addr3_known_bad +l_addr3_ums_addrid l_identific l_noempl l_profession l_program l_represent l_userid_active l_inactive_barcode1 +l_inactive_barcode2); # Load each record @@ -40,7 +77,7 @@ while (<>) { next; } - # Is this line the beginning of a block of data (typically an address or a note)? + # Is this line the end of a block of data (typically an address or a note)? if ( /^\.(.*?)_END.$/ ) { if ("$1." ne $section) { print STDERR "Error in record $serial, line $line (input line $.): got an end-of-$1 but I thought I was in a $section block!\n"; @@ -65,9 +102,9 @@ while (<>) { $unique_fields{$field} = 1; # Now we can actually store this line of data! - $users[$serial]{$field} = $3; + $records[$serial]{$field} = $3; - # print STDERR "Data extracted: \$users[$serial]{'$field'} = '$3'\n"; + # print STDERR "Data extracted: \$records[$serial]{'$field'} = '$3'\n"; next; } @@ -75,35 +112,127 @@ while (<>) { # This is the continuation of the previous line. else { chomp($_); - $users[$serial]{$field} .= ' ' . $_; - # print STDERR "Appended data to previous field. \$users[$serial]{'$field'} is now '" . $users[$serial]{$field} . "'.\n"; + $records[$serial]{$field} .= ' ' . $_; + # print STDERR "Appended data to previous field. \$records[$serial]{'$field'} is now '" . $records[$serial]{$field} . "'.\n"; } } -print STDERR "Loaded " . scalar(@users) . " users.\n"; +print STDERR "Loaded " . scalar(@records) . " records.\n"; -# Print a header line -print "SERIAL\t"; -@sorted_fields = sort keys %unique_fields; -foreach $i (@sorted_fields) { - print "$i\t"; -} -print "\n"; +# Process the records: +for (my $u = 0; $u < @records; $u++) { + +#print STDERR "Before processing:\n\n"; +#print STDERR Dumper($records[$u]); + + # Some fields can be mapped straightforwardly: + foreach $f (qw( user_id user_alt_id user_pin user_profile user_status user_library user_priv_granted user_priv_expires user_mailingaddr user_claims_ret user_environment user_department user_last_activity user_category1 user_category2 user_category3 user_category4 )) { + $records[$u]{uc($f)} = '' unless defined $records[$u]{uc($f)}; + $records[$u]{'l_' . $f} = $records[$u]{uc($f)}; + } + + # Addresses are a bit different: + foreach $a (qw( addr1 addr2 addr3 )) { + foreach $f (qw( std_line1 std_line2 std_city std_state std_zip country county township room company office phone dayphone homephone workphone cellphone fax email location usefor care_of known_bad ums_addrid )) { + $records[$u]{uc('USER_' . $a . '.' . $f)} = '' unless defined $records[$u]{uc('USER_' . $a . '.' . $f)}; + $records[$u]{'l_' . $a . '_' . $f} = $records[$u]{uc('USER_' . $a . '.' . $f)}; + } + $records[$u]{'l_' . $a . '_std_line1'} = $records[$u]{'USER_' . uc($a) . '.STREET'}; + $records[$u]{'l_' . $a . '_std_line2'} = $records[$u]{'USER_' . uc($a) . '.LINE2'}; + if ((defined $records[$u]{'USER_' . uc($a) . '.CITY/STATE'}) && ($records[$u]{'USER_' . uc($a) . '.CITY/STATE'} =~ m/^(.*),?(\s+)(.*)$/)) { + $records[$u]{'l_' . $a . '_std_city'} = $1; + $records[$u]{'l_' . $a . '_std_state'} = $3; + } + $records[$u]{'l_' . $a . '_std_zip'} = $records[$u]{'USER_' . uc($a) . '.ZIP'}; + + } + + # Handle fields that don't exactly match (e.g. parse USER_NAME into l_last_name etc.) + + $records[$u]{'l_birthdate'} = $records[$u]{'USER_BIRTH_DATE'}; + $records[$u]{'l_note'} = $records[$u]{'USER_XINFO.NOTE'}; + $records[$u]{'l_note1'} = ''; + $records[$u]{'l_patron'} = ''; + $records[$u]{'l_comment'} = $records[$u]{'USER_XINFO.COMMENT'}; + $records[$u]{'l_staff'} = $records[$u]{'USER_XINFO.STAFF'}; + $records[$u]{'l_webcatpref'} = $records[$u]{'USER_XINFO.WEBCATPREF'}; + $records[$u]{'l_dept'} = $records[$u]{'USER_DEPARTMENT'}; + $records[$u]{'l_guardian'} = ''; + $records[$u]{'l_license'} = $records[$u]{'USER_XINFO.LICENSE'}; + $records[$u]{'l_ssn'} = $records[$u]{'USER_XINFO.SSN'}; + $records[$u]{'l_misc'} = ''; + $records[$u]{'l_aup'} = ''; + $records[$u]{'l_photo'} = ''; + $records[$u]{'l_notify_via'} = $records[$u]{'USER_XINFO.NOTIFY_VIA'}; + $records[$u]{'l_ums_id'} = ''; + $records[$u]{'l_placcard'} = ''; + $records[$u]{'l_user_email'} = ''; + $records[$u]{'l_identific'} = ''; + $records[$u]{'l_noempl'} = ''; + $records[$u]{'l_profession'} = ''; + $records[$u]{'l_program'} = ''; + $records[$u]{'l_represent'} = ''; + $records[$u]{'l_userid_active'} = ''; + $records[$u]{'l_inactive_barcode1'} = $records[$u]{'USER_XINFO.PREV_ID'}; + $records[$u]{'l_inactive_barcode2'} = $records[$u]{'USER_XINFO.PREV_ID2'}; +# $records[$u]{'l_user_category1'} = $records[$u]{'USER_ROUTING_FLAG'}; +# $records[$u]{'l_user_category2'} = $records[$u]{'USER_WEB_AUTH'}; + + # We can parse the name like so: + + # Copy the name to a temp value + $temp_name = $records[$u]{'USER_NAME'}; + + # If there's no comma, don't try to parse the name + unless ($temp_name =~ m/,/) { + $records[$u]{'l_last_name'} = $temp_name; + next; + } + # Strip off a prefix, if there is one + foreach $prefix (qw( Ms\. Mrs\. Mr\. Dr\. )) { + if ($temp_name =~ /$prefix /i) { + $records[$u]{'l_prefix_name'} = $prefix; + $temp_name =~ s/$prefix //i; + } + } + + # Strip off a suffix, if there is one + foreach $suffix (qw( Jr\. Jr Sr\. Sr III II IV )) { + if ($temp_name =~ / $suffix/i) { + $records[$u]{'l_suffix_name'} = $suffix; + $temp_name =~ s/ $suffix//i; + } + } + + # Strip off the family name (before the comma) + # Of what remains, whatever is before the first space is the first name and the rest is the middle name + $temp_name =~ m/^([^,]*)\s*,.*$/; + $records[$u]{'l_last_name'} = $1; + $temp_name =~ m/^[^,]*\s*,\s*([^,\s]*)\s*(.*)$/; + $records[$u]{'l_first_name'} = $1; + $records[$u]{'l_middle_name'} = $2; + +#print STDERR "\n\nAfter processing:\n\n"; +#print STDERR Dumper($records[$u]); + + +} # Print the results -for (my $u = 0; $u < @users; $u++) { - print "$u\t"; - foreach $f (@sorted_fields) { - if (defined $users[$u]{$f}) { - print $users[$u]{$f}; +print join("\t", @output_fields) . "\n"; + +for (my $u = 0; $u < @records; $u++) { + foreach $f (@output_fields) { + if (defined $records[$u]{$f}) { + print $records[$u]{$f}; } print "\t"; } print "\n"; } -print STDERR "Wrote " . scalar(@users) . " users.\n"; +print STDERR "Wrote " . scalar(@records) . " records.\n"; # uh-bdee-uh-bdee-uh-bdee-uh- THAT'S ALL, FOLKS