From 2bc4e97f72b39cc0093d3e295f97fcb4fa91106d Mon Sep 17 00:00:00 2001 From: Dan Scott Date: Thu, 10 Nov 2011 01:05:40 -0500 Subject: [PATCH] Create a search_normalize() variant of naco_normalize() In the quest to support searching text that contains leading articles joined by an apostrophe - for example, "l'histoire" - such that a searcher can enter either "l'histoire" or "histoire" and get results - add a variant of naco_normalize() that does not strip the apostrophe entirely, but rather replaces it with a space such that the root word can be indexed appropriately. This implementation refactors the OpenILS::Utils::Normalize code to make the differences between search_normalize() and naco_normalize() as clear as possible, but duplicates code significantly in the in-db version of the code. Someday maybe the database can rely on OpenILS::Utils::Normalize instead of inline functions :) Signed-off-by: Dan Scott --- .../Application/Storage/Driver/Pg/QueryParser.pm | 18 +++--- .../src/perlmods/lib/OpenILS/Utils/Normalize.pm | 36 ++++++++++- Open-ILS/src/perlmods/t/14-OpenILS-Utils.t | 8 ++- Open-ILS/src/sql/Pg/002.functions.config.sql | 75 ++++++++++++++++++++++ Open-ILS/src/sql/Pg/950.data.seed-values.sql | 9 ++- 5 files changed, 132 insertions(+), 14 deletions(-) diff --git a/Open-ILS/src/perlmods/lib/OpenILS/Application/Storage/Driver/Pg/QueryParser.pm b/Open-ILS/src/perlmods/lib/OpenILS/Application/Storage/Driver/Pg/QueryParser.pm index 41ed64c45e..8affc65090 100644 --- a/Open-ILS/src/perlmods/lib/OpenILS/Application/Storage/Driver/Pg/QueryParser.pm +++ b/Open-ILS/src/perlmods/lib/OpenILS/Application/Storage/Driver/Pg/QueryParser.pm @@ -359,7 +359,7 @@ sub TEST_SETUP { __PACKAGE__->add_relevance_bump( title => translated => full_match => 20 ); __PACKAGE__->add_search_field_id_map( title => proper => 6 => 1 ); - __PACKAGE__->add_query_normalizer( title => proper => 'naco_normalize' ); + __PACKAGE__->add_query_normalizer( title => proper => 'search_normalize' ); __PACKAGE__->add_relevance_bump( title => proper => first_word => 1.5 ); __PACKAGE__->add_relevance_bump( title => proper => full_match => 20 ); __PACKAGE__->add_relevance_bump( title => proper => word_order => 10 ); @@ -373,7 +373,7 @@ sub TEST_SETUP { __PACKAGE__->add_search_field_id_map( author => personal => 8 => 1 ); __PACKAGE__->add_relevance_bump( author => personal => first_word => 1.5 ); __PACKAGE__->add_relevance_bump( author => personal => full_match => 20 ); - __PACKAGE__->add_query_normalizer( author => personal => 'naco_normalize' ); + __PACKAGE__->add_query_normalizer( author => personal => 'search_normalize' ); __PACKAGE__->add_query_normalizer( author => personal => 'split_date_range' ); __PACKAGE__->add_facet_field_id_map( subject => topic => 14 => 1 ); @@ -401,8 +401,8 @@ sub TEST_SETUP { __PACKAGE__->add_search_class_alias( series => 'se' ); __PACKAGE__->add_search_class_alias( keyword => 'dc.identifier' ); - __PACKAGE__->add_query_normalizer( author => corporate => 'naco_normalize' ); - __PACKAGE__->add_query_normalizer( keyword => keyword => 'naco_normalize' ); + __PACKAGE__->add_query_normalizer( author => corporate => 'search_normalize' ); + __PACKAGE__->add_query_normalizer( keyword => keyword => 'search_normalize' ); __PACKAGE__->add_search_field_alias( subject => name => 'bib.subjectName' ); @@ -693,13 +693,13 @@ sub rel_bump { return '' if (!@$only_atoms); if ($bump eq 'first_word') { - return " /* first_word */ COALESCE(NULLIF( (naco_normalize(".$node->table_alias.".value) ~ ('^'||naco_normalize(".$self->QueryParser->quote_value($only_atoms->[0]->content)."))), FALSE )::INT * $multiplier, 1)"; + return " /* first_word */ COALESCE(NULLIF( (search_normalize(".$node->table_alias.".value) ~ ('^'||search_normalize(".$self->QueryParser->quote_value($only_atoms->[0]->content)."))), FALSE )::INT * $multiplier, 1)"; } elsif ($bump eq 'full_match') { - return " /* full_match */ COALESCE(NULLIF( (naco_normalize(".$node->table_alias.".value) ~ ('^'||". - join( "||' '||", map { "naco_normalize(".$self->QueryParser->quote_value($_->content).")" } @$only_atoms )."||'\$')), FALSE )::INT * $multiplier, 1)"; + return " /* full_match */ COALESCE(NULLIF( (search_normalize(".$node->table_alias.".value) ~ ('^'||". + join( "||' '||", map { "search_normalize(".$self->QueryParser->quote_value($_->content).")" } @$only_atoms )."||'\$')), FALSE )::INT * $multiplier, 1)"; } elsif ($bump eq 'word_order') { - return " /* word_order */ COALESCE(NULLIF( (naco_normalize(".$node->table_alias.".value) ~ (". - join( "||'.*'||", map { "naco_normalize(".$self->QueryParser->quote_value($_->content).")" } @$only_atoms ).")), FALSE )::INT * $multiplier, 1)"; + return " /* word_order */ COALESCE(NULLIF( (search_normalize(".$node->table_alias.".value) ~ (". + join( "||'.*'||", map { "search_normalize(".$self->QueryParser->quote_value($_->content).")" } @$only_atoms ).")), FALSE )::INT * $multiplier, 1)"; } return ''; diff --git a/Open-ILS/src/perlmods/lib/OpenILS/Utils/Normalize.pm b/Open-ILS/src/perlmods/lib/OpenILS/Utils/Normalize.pm index d71503c5e1..e3e699f813 100644 --- a/Open-ILS/src/perlmods/lib/OpenILS/Utils/Normalize.pm +++ b/Open-ILS/src/perlmods/lib/OpenILS/Utils/Normalize.pm @@ -5,10 +5,9 @@ use Unicode::Normalize; use Encode; use Exporter 'import'; -our @EXPORT_OK = qw( naco_normalize ); +our @EXPORT_OK = qw( naco_normalize search_normalize ); sub naco_normalize { - my $str = decode_utf8(shift); my $sf = shift; @@ -18,10 +17,34 @@ sub naco_normalize { # Note that unlike a strict reading of the NACO normalization rules, # output is returned as lowercase instead of uppercase for compatibility # with previous versions of the Evergreen naco_normalize routine. + $str = _normalize_substitutions($str, $sf); + + # Remove apostrophes, per NACO specs + $str =~ tr/'//d; + + $str = _normalize_codes($str, $sf); + + return $str; +} + +sub search_normalize { + my $str = decode_utf8(shift); + my $sf = shift; + + $str = _normalize_substitutions($str, $sf); + $str = _normalize_codes($str, $sf); + + return $str; +} + +sub _normalize_substitutions { + my $str = shift; + my $sf = shift; # Convert to upper-case first; even though final output will be lowercase, doing this will # ensure that the German eszett (ß) and certain ligatures (ff, fi, ffl, etc.) will be handled correctly. # If there are any bugs in Perl's implementation of upcasing, they will be passed through here. + $str = uc $str; # remove non-filing strings @@ -33,7 +56,14 @@ sub naco_normalize { $str =~ s/\x{00C6}/AE/g; $str =~ s/\x{00DE}/TH/g; $str =~ s/\x{0152}/OE/g; - $str =~ tr/\x{0110}\x{00D0}\x{00D8}\x{0141}\x{2113}\x{02BB}\x{02BC}]['/DDOLl/d; + $str =~ tr/\x{0110}\x{00D0}\x{00D8}\x{0141}\x{2113}\x{02BB}\x{02BC}][/DDOLl/d; + + return $str; +} + +sub _normalize_codes { + my $str = shift; + my $sf = shift; # transformations based on Unicode category codes $str =~ s/[\p{Cc}\p{Cf}\p{Co}\p{Cs}\p{Lm}\p{Mc}\p{Me}\p{Mn}]//g; diff --git a/Open-ILS/src/perlmods/t/14-OpenILS-Utils.t b/Open-ILS/src/perlmods/t/14-OpenILS-Utils.t index 28a7267e4f..924e2a3f83 100644 --- a/Open-ILS/src/perlmods/t/14-OpenILS-Utils.t +++ b/Open-ILS/src/perlmods/t/14-OpenILS-Utils.t @@ -1,6 +1,6 @@ #!perl -T -use Test::More tests => 20; +use Test::More tests => 22; use_ok( 'OpenILS::Utils::Configure' ); use_ok( 'OpenILS::Utils::Cronscript' ); @@ -37,3 +37,9 @@ is(@comp_holdings, 0, "Compressed holdings for an MFHD record that only has a ca my @decomp_holdings = $co_mfhd->get_decompressed_holdings($co_mfhd->field('853')); is(@decomp_holdings, 0, "Decompressed holdings for an MFHD record that only has a caption"); + +my $apostring = OpenILS::Utils::Normalize::naco_normalize("it's time"); +is($apostring, "its time", "naco_normalize: strip apostrophes"); + +my $apos = OpenILS::Utils::Normalize::search_normalize("it's time"); +is($apos, "it s time", "search_normalize: replace apostrophes with space"); diff --git a/Open-ILS/src/sql/Pg/002.functions.config.sql b/Open-ILS/src/sql/Pg/002.functions.config.sql index 48363842be..3ea61ae289 100644 --- a/Open-ILS/src/sql/Pg/002.functions.config.sql +++ b/Open-ILS/src/sql/Pg/002.functions.config.sql @@ -699,6 +699,72 @@ CREATE OR REPLACE FUNCTION public.naco_normalize( TEXT, TEXT ) RETURNS TEXT AS $ return lc $str; $func$ LANGUAGE 'plperlu' STRICT IMMUTABLE; +-- Currently, the only difference from naco_normalize is that search_normalize +-- turns apostrophes into spaces, while naco_normalize collapses them. +CREATE OR REPLACE FUNCTION public.search_normalize( TEXT, TEXT ) RETURNS TEXT AS $func$ + + use strict; + use Unicode::Normalize; + use Encode; + + my $str = decode_utf8(shift); + my $sf = shift; + + # Apply NACO normalization to input string; based on + # http://www.loc.gov/catdir/pcc/naco/SCA_PccNormalization_Final_revised.pdf + # + # Note that unlike a strict reading of the NACO normalization rules, + # output is returned as lowercase instead of uppercase for compatibility + # with previous versions of the Evergreen naco_normalize routine. + + # Convert to upper-case first; even though final output will be lowercase, doing this will + # ensure that the German eszett (ß) and certain ligatures (ff, fi, ffl, etc.) will be handled correctly. + # If there are any bugs in Perl's implementation of upcasing, they will be passed through here. + $str = uc $str; + + # remove non-filing strings + $str =~ s/\x{0098}.*?\x{009C}//g; + + $str = NFKD($str); + + # additional substitutions - 3.6. + $str =~ s/\x{00C6}/AE/g; + $str =~ s/\x{00DE}/TH/g; + $str =~ s/\x{0152}/OE/g; + $str =~ tr/\x{0110}\x{00D0}\x{00D8}\x{0141}\x{2113}\x{02BB}\x{02BC}][/DDOLl/d; + + # transformations based on Unicode category codes + $str =~ s/[\p{Cc}\p{Cf}\p{Co}\p{Cs}\p{Lm}\p{Mc}\p{Me}\p{Mn}]//g; + + if ($sf && $sf =~ /^a/o) { + my $commapos = index($str, ','); + if ($commapos > -1) { + if ($commapos != length($str) - 1) { + $str =~ s/,/\x07/; # preserve first comma + } + } + } + + # since we've stripped out the control characters, we can now + # use a few as placeholders temporarily + $str =~ tr/+&@\x{266D}\x{266F}#/\x01\x02\x03\x04\x05\x06/; + $str =~ s/[\p{Pc}\p{Pd}\p{Pe}\p{Pf}\p{Pi}\p{Po}\p{Ps}\p{Sk}\p{Sm}\p{So}\p{Zl}\p{Zp}\p{Zs}]/ /g; + $str =~ tr/\x01\x02\x03\x04\x05\x06\x07/+&@\x{266D}\x{266F}#,/; + + # decimal digits + $str =~ tr/\x{0660}-\x{0669}\x{06F0}-\x{06F9}\x{07C0}-\x{07C9}\x{0966}-\x{096F}\x{09E6}-\x{09EF}\x{0A66}-\x{0A6F}\x{0AE6}-\x{0AEF}\x{0B66}-\x{0B6F}\x{0BE6}-\x{0BEF}\x{0C66}-\x{0C6F}\x{0CE6}-\x{0CEF}\x{0D66}-\x{0D6F}\x{0E50}-\x{0E59}\x{0ED0}-\x{0ED9}\x{0F20}-\x{0F29}\x{1040}-\x{1049}\x{1090}-\x{1099}\x{17E0}-\x{17E9}\x{1810}-\x{1819}\x{1946}-\x{194F}\x{19D0}-\x{19D9}\x{1A80}-\x{1A89}\x{1A90}-\x{1A99}\x{1B50}-\x{1B59}\x{1BB0}-\x{1BB9}\x{1C40}-\x{1C49}\x{1C50}-\x{1C59}\x{A620}-\x{A629}\x{A8D0}-\x{A8D9}\x{A900}-\x{A909}\x{A9D0}-\x{A9D9}\x{AA50}-\x{AA59}\x{ABF0}-\x{ABF9}\x{FF10}-\x{FF19}/0-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-90-9/; + + # intentionally skipping step 8 of the NACO algorithm; if the string + # gets normalized away, that's fine. + + # leading and trailing spaces + $str =~ s/\s+/ /g; + $str =~ s/^\s+//; + $str =~ s/\s+$//g; + + return lc $str; +$func$ LANGUAGE 'plperlu' STRICT IMMUTABLE; + CREATE OR REPLACE FUNCTION public.naco_normalize_keep_comma( TEXT ) RETURNS TEXT AS $func$ SELECT public.naco_normalize($1,'a'); $func$ LANGUAGE SQL STRICT IMMUTABLE; @@ -707,5 +773,14 @@ CREATE OR REPLACE FUNCTION public.naco_normalize( TEXT ) RETURNS TEXT AS $func$ SELECT public.naco_normalize($1,''); $func$ LANGUAGE 'sql' STRICT IMMUTABLE; +CREATE OR REPLACE FUNCTION public.search_normalize_keep_comma( TEXT ) RETURNS TEXT AS $func$ + SELECT public.search_normalize($1,'a'); +$func$ LANGUAGE SQL STRICT IMMUTABLE; + +CREATE OR REPLACE FUNCTION public.search_normalize( TEXT ) RETURNS TEXT AS $func$ + SELECT public.search_normalize($1,''); +$func$ LANGUAGE 'sql' STRICT IMMUTABLE; + + COMMIT; diff --git a/Open-ILS/src/sql/Pg/950.data.seed-values.sql b/Open-ILS/src/sql/Pg/950.data.seed-values.sql index e032835fb5..7f0942428b 100644 --- a/Open-ILS/src/sql/Pg/950.data.seed-values.sql +++ b/Open-ILS/src/sql/Pg/950.data.seed-values.sql @@ -7170,6 +7170,13 @@ INSERT INTO config.index_normalizer (name, description, func, param_count) VALUE 1 ); +INSERT INTO config.index_normalizer (name, description, func, param_count) VALUES ( + 'Search Normalize', + 'Apply search normalization rules to the extracted text. A less extreme version of NACO normalization.', + 'search_normalize', + 0 +); + -- make use of the index normalizers INSERT INTO config.metabib_field_index_norm_map (field,norm) @@ -7177,7 +7184,7 @@ INSERT INTO config.metabib_field_index_norm_map (field,norm) i.id FROM config.metabib_field m, config.index_normalizer i - WHERE i.func IN ('naco_normalize','split_date_range') + WHERE i.func IN ('search_normalize','split_date_range') AND m.id NOT IN (18, 19); INSERT INTO config.metabib_field_index_norm_map (field,norm,pos) -- 2.11.0