[scpubgit/stemmatology.git] / lib / Text / Tradition / Parser / TEI.pm

package Text::Tradition::Parser::TEI;

use strict;
use warnings;
use Text::Tradition::Parser::Util qw( collate_variants );
use XML::LibXML;
use XML::LibXML::XPathContext;

=head1 NAME

Text::Tradition::Parser::TEI

=head1 DESCRIPTION

Parser module for Text::Tradition, given a TEI parallel-segmentation
file that describes a text and its variants.

=head1 METHODS

=over

=item B<parse>

parse( $tei_string );

Takes an initialized tradition and a string containing the TEI;
creates the appropriate nodes and edges on the graph, as well as
the appropriate witness objects.

=cut

my $text = {}; # Hash of arrays, one per eventual witness we find.
my $substitutions = {}; # Keep track of merged readings
my $app_anchors = {};   # Track apparatus references
my $app_ac = {};        # Save a.c. readings
my $app_count;          # Keep track of how many apps we have

# Create the package variables for tag names.

# Would really like to do this with varname variables, but apparently this
# is considered a bad idea.  The long way round then.
my( $LISTWIT, $WITNESS, $TEXT, $W, $SEG, $APP, $RDG, $LEM ) 
    = ( 'listWit', 'witness', 'text', 'w', 'seg', 'app', 'rdg', 'lem' );
sub make_tagnames {
    my( $ns ) = @_;
    if( $ns ) {
        $LISTWIT = "$ns:$LISTWIT";
        $WITNESS = "$ns:$WITNESS";
        $TEXT = "$ns:$TEXT";
        $W = "$ns:$W";
        $SEG = "$ns:$SEG";
        $APP = "$ns:$APP";
        $RDG = "$ns:$RDG";
        $LEM = "$ns:$LEM";
    }
}

# Parse the TEI file.
sub parse {
    my( $tradition, $xml_str ) = @_;
    
    # First, parse the XML.
    my $parser = XML::LibXML->new();
    my $doc = $parser->parse_string( $xml_str );
    my $tei = $doc->documentElement();
    my $xpc = XML::LibXML::XPathContext->new( $tei );
    my $ns;
    if( $tei->namespaceURI ) {
        $ns = 'tei';
        $xpc->registerNs( $ns, $tei->namespaceURI );
    }
    make_tagnames( $ns );

    # Then get the witnesses and create the witness objects.
    foreach my $wit_el ( $xpc->findnodes( "//$LISTWIT/$WITNESS" ) ) {
        my $sig = $wit_el->getAttribute( 'xml:id' );
        my $source = $wit_el->toString();
        $tradition->add_witness( sigil => $sig, source => $source );
    }
    map { $text->{$_->sigil} = [] } $tradition->witnesses;

    # Look for all word/seg node IDs and note their pre-existence.
    my @attrs = $xpc->findnodes( "//$W|$SEG/attribute::xml:id" );
    save_preexisting_nodeids( @attrs );

    # Count up how many apps we have.
    my @apps = $xpc->findnodes( "//$APP" );
    $app_count = scalar( @apps );

    # Now go through the children of the text element and pull out the
    # actual text.
    foreach my $xml_el ( $xpc->findnodes( "//$TEXT" ) ) {
        foreach my $xn ( $xml_el->childNodes ) {
            _get_readings( $tradition, $xn );
        }
    }
    # Our $text global now has lists of readings, one per witness.
    # Join them up.
    my $c = $tradition->collation;
    foreach my $sig ( keys %$text ) {
        next if $sig eq 'base';  # Skip base text readings with no witnesses.
        # Determine the list of readings for 
        my $sequence = $text->{$sig};
        my @real_sequence = ( $c->start );
        push( @$sequence, $c->end );
        my $source = $c->start;
        foreach( _clean_sequence( $sig, $sequence ) ) {
            my $rdg = _return_rdg( $_ );
            push( @real_sequence, $rdg );
            $c->add_path( $source, $rdg, $sig );
            $source = $rdg;
        }
        $tradition->witness( $sig )->path( \@real_sequence );
        # See if we need to make an a.c. version of the witness.
        if( exists $app_ac->{$sig} ) {
            my @uncorrected;
            push( @uncorrected, @real_sequence );
            foreach my $app ( keys %{$app_ac->{$sig}} ) {
                my $start = _return_rdg( $app_anchors->{$app}->{$sig}->{'start'} ); 
                my $end = _return_rdg( $app_anchors->{$app}->{$sig}->{'end'} );
                my @new = map { _return_rdg( $_ ) } @{$app_ac->{$sig}->{$app}};
                _replace_sequence( \@uncorrected, $start, $end, @new );
            }
            my $source = $c->start;
            foreach my $rdg ( @uncorrected ) {
                my $has_base = grep { $_->label eq $sig } $source->edges_to( $rdg );
                if( $rdg ne $c->start && !$has_base ) {
                    print STDERR sprintf( "Adding path %s from %s -> %s\n",
                        $sig.$c->ac_label, $source->name, $rdg->name );
                    $c->add_path( $source, $rdg, $sig.$c->ac_label );
                }
                $source = $rdg;
            }
            print STDERR "Adding a.c. version for witness $sig\n";
            $tradition->witness( $sig )->uncorrected_path( \@uncorrected );
        }
    }
    # Delete readings that are no longer part of the graph.
    # TODO think this is useless actually
    foreach ( keys %$substitutions ) {
        $tradition->collation->del_reading( $tradition->collation->reading( $_ ) );
    }
    $tradition->collation->calculate_ranks();
    
    # Now that we have ranks, see if we have distinct nodes with identical
    # text and identical rank that can be merged.
    $tradition->collation->flatten_ranks();
}

sub _clean_sequence {
    my( $wit, $sequence ) = @_;
    my @clean_sequence;
    foreach my $rdg ( @$sequence ) {
        if( $rdg =~ /^PH-(.*)$/ ) {
            # It is a placeholder.  Keep it only if we need it.
            my $app_id = $1;
            if( exists $app_ac->{$wit} &&
                exists $app_ac->{$wit}->{$app_id} ) {
                print STDERR "Retaining empty placeholder for $app_id\n";
                push( @clean_sequence, $rdg );
            }
        } else {
            push( @clean_sequence, $rdg );
        }
    }
    return @clean_sequence;
}

sub _replace_sequence {
    my( $arr, $start, $end, @new ) = @_;
    my( $start_idx, $end_idx );
    foreach my $i ( 0 .. $#{$arr} ) {
        $start_idx = $i if( $arr->[$i]->name eq $start );
        if( $arr->[$i]->name eq $end ) {
            $end_idx = $i;
            last;
        }
    }
    unless( $start_idx && $end_idx ) {
        warn "Could not find start and end";
        return;
    }
    my $length = $end_idx - $start_idx + 1;
    splice( @$arr, $start_idx, $length, @new );
}

sub _return_rdg {
    my( $rdg ) = @_;
    # If we were passed a reading name, return the name.  If we were
    # passed a reading object, return the object.
    my $wantobj = ref( $rdg ) eq 'Text::Tradition::Collation::Reading';
    my $real = $rdg;
    if( exists $substitutions->{ $wantobj ? $rdg->name : $rdg } ) {
        $real = $substitutions->{ $wantobj ? $rdg->name : $rdg };
        $real = $real->name unless $wantobj;
    }
    return $real;
}

## Recursive helper function to help us navigate through nested XML,
## picking out the text.  $tradition is the tradition, needed for
## making readings; $xn is the XML node currently being looked at,
## $in_var is a flag to say that we are inside a variant, $ac is a
## flag to say that we are inside an ante-correctionem reading, and
## @cur_wits is the list of witnesses to which this XML node applies.
## Returns the list of readings, if any, created on the run.

{
    my %active_wits;
    my $current_app;
    my $seen_apps;

    sub _get_readings {
        my( $tradition, $xn, $in_var, $ac, @cur_wits ) = @_;
        @cur_wits = grep { $active_wits{$_} } keys %active_wits unless $in_var;

        my @new_readings;
        if( $xn->nodeType == XML_TEXT_NODE ) {
            # Some words, thus make some readings.
            my $str = $xn->data;
            return unless $str =~ /\S/; # skip whitespace-only text nodes
            #print STDERR "Handling text node " . $str . "\n";
            # Check that all the witnesses we have are active.
            foreach my $c ( @cur_wits ) {
                warn "$c is not among active wits" unless $active_wits{$c};
            }
            $str =~ s/^\s+//;
            my $final = $str =~ s/\s+$//;
            foreach my $w ( split( /\s+/, $str ) ) {
                # For now, skip punctuation.
                next if $w !~ /[[:alnum:]]/;
                my $rdg = make_reading( $tradition->collation, $w );
                push( @new_readings, $rdg );
                unless( $in_var ) {
                    $rdg->make_common;
                }
                foreach ( @cur_wits ) {
                    warn "Empty wit!" unless $_;
                    warn "Empty reading!" unless $rdg;
                    push( @{$text->{$_}}, $rdg ) unless $ac;
                }
            }
        } elsif( $xn->nodeName eq 'w' ) {
            # Everything in this tag is one word.  Also save any original XML ID.
            #print STDERR "Handling word " . $xn->toString . "\n";
            # Check that all the witnesses we have are active.
            foreach my $c ( @cur_wits ) {
                warn "$c is not among active wits" unless $active_wits{$c};
            }
            my $xml_id = $xn->getAttribute( 'xml:id' );
            my $rdg = make_reading( $tradition->collation, $xn->textContent, $xml_id );
            push( @new_readings, $rdg );
            unless( $in_var ) {
                $rdg->make_common;
            }
            foreach( @cur_wits ) {
                warn "Empty wit!" unless $_;
                warn "Empty reading!" unless $rdg;
                push( @{$text->{$_}}, $rdg ) unless $ac;
            }
        } elsif ( $xn->nodeName eq 'app' ) {
            $seen_apps++;
            $current_app = $xn->getAttribute( 'xml:id' );
            # print STDERR "Handling app $current_app\n";
            # Keep the reading sets in this app.
            my @sets;
            # Recurse through all children (i.e. rdgs) for sets of words.
            foreach ( $xn->childNodes ) {
                my @rdg_set = _get_readings( $tradition, $_, $in_var, $ac, @cur_wits );
                push( @sets, \@rdg_set ) if @rdg_set;
            }
            # Now collate these sets if we have more than one.
            my $subs = collate_variants( $tradition->collation, @sets ) if @sets > 1;
            map { $substitutions->{$_} = $subs->{$_} } keys %$subs;
            # TODO Look through substitutions to see if we can make anything common now.
            # Return the entire set of unique readings.
            my %unique;
            foreach my $s ( @sets ) {
                map { $unique{$_->name} = $_ } @$s;
            }
            push( @new_readings, values( %unique ) );
            # Exit the current app.
            $current_app = '';
        } elsif ( $xn->nodeName eq 'lem' || $xn->nodeName eq 'rdg' ) {
            # Alter the current witnesses and recurse.
            #print STDERR "Handling reading for " . $xn->getAttribute( 'wit' ) . "\n";
            # TODO handle p.c. and s.l. designations too
            $ac = $xn->getAttribute( 'type' ) && $xn->getAttribute( 'type' ) eq 'a.c.';
            my @rdg_wits = get_sigla( $xn );
            @rdg_wits = ( 'base' ) unless @rdg_wits;  # Allow for editorially-supplied readings
            my @words;
            foreach ( $xn->childNodes ) {
                my @rdg_set = _get_readings( $tradition, $_, 1, $ac, @rdg_wits );
                push( @words, @rdg_set ) if @rdg_set;
            }
            # If we have more than one word in a reading, it should become a segment.
            # $tradition->collation->add_segment( @words ) if @words > 1;
            
            if( $ac ) {
                # Add the reading set to the a.c. readings.
                foreach ( @rdg_wits ) {
                    $app_ac->{$_}->{$current_app} = \@words;
                }
            } else {
                # Add the reading set to the app anchors for each witness
                # or put in placeholders for empty p.c. readings
                foreach ( @rdg_wits ) {
                    my $start = @words ? $words[0]->name : "PH-$current_app";
                    my $end = @words ? $words[-1]->name : "PH-$current_app";
                    $app_anchors->{$current_app}->{$_}->{'start'} = $start;
                    $app_anchors->{$current_app}->{$_}->{'end'} = $end;
                    push( @{$text->{$_}}, $start ) unless @words;
                }
            }
            push( @new_readings, @words );
        } elsif( $xn->nodeName eq 'witStart' ) {
            # Add the relevant wit(s) to the active list.
            #print STDERR "Handling witStart\n";
            map { $active_wits{$_} = 1 } @cur_wits;
            # Record a lacuna in all non-active witnesses if this is
            # the first app. Get the full list from $text.
            if( $seen_apps == 1 ) {
                my $i = 0;
                foreach my $sig ( keys %$text ) {
                    next if $active_wits{$sig};
                    my $l = $tradition->collation->add_lacuna( $current_app . "_$i" );
                    $i++;
                    push( @{$text->{$sig}}, $l );
                }
            }
        } elsif( $xn->nodeName eq 'witEnd' ) {
            # Take the relevant wit(s) out of the list.
            #print STDERR "Handling witEnd\n";
            map { $active_wits{$_} = undef } @cur_wits;
            # Record a lacuna, unless this is the last app.
            unless( $seen_apps == $app_count ) {
                foreach my $i ( 0 .. $#cur_wits ) {
                    my $w = $cur_wits[$i];
                    my $l = $tradition->collation->add_lacuna( $current_app . "_$i" );
                    push( @{$text->{$w}}, $l );
                }
            }
        } elsif( $xn->nodeName eq 'witDetail' ) {
            # Ignore these for now.
            return;
        } else {
            # Recurse as if this tag weren't there.
            #print STDERR "Recursing on tag " . $xn->nodeName . "\n";
            foreach( $xn->childNodes ) {
                push( @new_readings, _get_readings( $tradition, $_, $in_var, $ac, @cur_wits ) );
            }
        }
        return @new_readings;
    }

}

# Helper to extract a list of witness sigla from a reading element.
sub get_sigla {
    my( $rdg ) = @_;
    # Cope if we have been handed a NodeList.  There is only
    # one reading here.
    if( ref( $rdg ) eq 'XML::LibXML::NodeList' ) {
        $rdg = $rdg->shift;
    }

    my @wits;
    if( ref( $rdg ) eq 'XML::LibXML::Element' ) {
        my $witstr = $rdg->getAttribute( 'wit' );
        $witstr =~ s/^\s+//;
        $witstr =~ s/\s+$//;
        @wits = split( /\s+/, $witstr );
        map { $_ =~ s/^\#// } @wits;
    }
    return @wits;
}

# Helper with its counters to actually make the readings.
{
    my $word_ctr = 0;
    my %used_nodeids;

    sub save_preexisting_nodeids {
        foreach( @_ ) {
            $used_nodeids{$_->getValue()} = 1;
        }
    }

    sub make_reading {
        my( $graph, $word, $xml_id ) = @_;
        if( $xml_id ) {
            if( exists $used_nodeids{$xml_id} ) {
                if( $used_nodeids{$xml_id} != 1 ) {
                    warn "Already used assigned XML ID somewhere else!";
                    $xml_id = undef;
                }
            } else {
                warn "Undetected pre-existing XML ID";
            }
        }
        if( !$xml_id ) {
            until( $xml_id ) {
                my $try_id = 'w'.$word_ctr++;
                next if exists $used_nodeids{$try_id};
                $xml_id = $try_id;
            }
        }
        my $rdg = $graph->add_reading( $xml_id );
        $rdg->text( $word );
        $used_nodeids{$xml_id} = $rdg;
        return $rdg;
    }
}

1;
Commit	Line	Data
f6066bac	1	package Text::Tradition::Parser::TEI;
	2
	3	use strict;
	4	use warnings;
910a0a6d	5	use Text::Tradition::Parser::Util qw( collate_variants );
f6066bac	6	use XML::LibXML;
	7	use XML::LibXML::XPathContext;
	8
	9	=head1 NAME
	10
	11	Text::Tradition::Parser::TEI
	12
	13	=head1 DESCRIPTION
	14
	15	Parser module for Text::Tradition, given a TEI parallel-segmentation
	16	file that describes a text and its variants.
	17
	18	=head1 METHODS
	19
	20	=over
	21
	22	=item B<parse>
	23
	24	parse( $tei_string );
	25
	26	Takes an initialized tradition and a string containing the TEI;
	27	creates the appropriate nodes and edges on the graph, as well as
	28	the appropriate witness objects.
	29
	30	=cut
	31
910a0a6d	32	my $text = {}; # Hash of arrays, one per eventual witness we find.
910a0a6d	33	my $substitutions = {}; # Keep track of merged readings
	34	my $app_anchors = {}; # Track apparatus references
	35	my $app_ac = {}; # Save a.c. readings
eca16057	36	my $app_count; # Keep track of how many apps we have
910a0a6d	37
	38	# Create the package variables for tag names.
	39
	40	# Would really like to do this with varname variables, but apparently this
	41	# is considered a bad idea. The long way round then.
	42	my( $LISTWIT, $WITNESS, $TEXT, $W, $SEG, $APP, $RDG, $LEM )
	43	= ( 'listWit', 'witness', 'text', 'w', 'seg', 'app', 'rdg', 'lem' );
	44	sub make_tagnames {
	45	my( $ns ) = @_;
	46	if( $ns ) {
	47	$LISTWIT = "$ns:$LISTWIT";
	48	$WITNESS = "$ns:$WITNESS";
	49	$TEXT = "$ns:$TEXT";
	50	$W = "$ns:$W";
	51	$SEG = "$ns:$SEG";
	52	$APP = "$ns:$APP";
	53	$RDG = "$ns:$RDG";
	54	$LEM = "$ns:$LEM";
	55	}
	56	}
	57
	58	# Parse the TEI file.
f6066bac	59	sub parse {
	60	my( $tradition, $xml_str ) = @_;
	61
	62	# First, parse the XML.
	63	my $parser = XML::LibXML->new();
	64	my $doc = $parser->parse_string( $xml_str );
	65	my $tei = $doc->documentElement();
f2b9605f	66	my $xpc = XML::LibXML::XPathContext->new( $tei );
910a0a6d	67	my $ns;
	68	if( $tei->namespaceURI ) {
	69	$ns = 'tei';
	70	$xpc->registerNs( $ns, $tei->namespaceURI );
	71	}
	72	make_tagnames( $ns );
	73
f6066bac	74	# Then get the witnesses and create the witness objects.
910a0a6d	75	foreach my $wit_el ( $xpc->findnodes( "//$LISTWIT/$WITNESS" ) ) {
	76	my $sig = $wit_el->getAttribute( 'xml:id' );
	77	my $source = $wit_el->toString();
	78	$tradition->add_witness( sigil => $sig, source => $source );
f6066bac	79	}
910a0a6d	80	map { $text->{$_->sigil} = [] } $tradition->witnesses;
eca16057	81
910a0a6d	82	# Look for all word/seg node IDs and note their pre-existence.
	83	my @attrs = $xpc->findnodes( "//$W\|$SEG/attribute::xml:id" );
	84	save_preexisting_nodeids( @attrs );
	85
eca16057	86	# Count up how many apps we have.
	87	my @apps = $xpc->findnodes( "//$APP" );
	88	$app_count = scalar( @apps );
	89
910a0a6d	90	# Now go through the children of the text element and pull out the
	91	# actual text.
	92	foreach my $xml_el ( $xpc->findnodes( "//$TEXT" ) ) {
	93	foreach my $xn ( $xml_el->childNodes ) {
	94	_get_readings( $tradition, $xn );
	95	}
	96	}
	97	# Our $text global now has lists of readings, one per witness.
	98	# Join them up.
	99	my $c = $tradition->collation;
	100	foreach my $sig ( keys %$text ) {
	101	next if $sig eq 'base'; # Skip base text readings with no witnesses.
	102	# Determine the list of readings for
	103	my $sequence = $text->{$sig};
	104	my @real_sequence = ( $c->start );
	105	push( @$sequence, $c->end );
	106	my $source = $c->start;
	107	foreach( _clean_sequence( $sig, $sequence ) ) {
	108	my $rdg = _return_rdg( $_ );
	109	push( @real_sequence, $rdg );
	110	$c->add_path( $source, $rdg, $sig );
	111	$source = $rdg;
	112	}
	113	$tradition->witness( $sig )->path( \@real_sequence );
	114	# See if we need to make an a.c. version of the witness.
	115	if( exists $app_ac->{$sig} ) {
	116	my @uncorrected;
	117	push( @uncorrected, @real_sequence );
	118	foreach my $app ( keys %{$app_ac->{$sig}} ) {
	119	my $start = _return_rdg( $app_anchors->{$app}->{$sig}->{'start'} );
	120	my $end = _return_rdg( $app_anchors->{$app}->{$sig}->{'end'} );
	121	my @new = map { _return_rdg( $_ ) } @{$app_ac->{$sig}->{$app}};
	122	_replace_sequence( \@uncorrected, $start, $end, @new );
	123	}
	124	my $source = $c->start;
	125	foreach my $rdg ( @uncorrected ) {
	126	my $has_base = grep { $_->label eq $sig } $source->edges_to( $rdg );
	127	if( $rdg ne $c->start && !$has_base ) {
	128	print STDERR sprintf( "Adding path %s from %s -> %s\n",
	129	$sig.$c->ac_label, $source->name, $rdg->name );
	130	$c->add_path( $source, $rdg, $sig.$c->ac_label );
	131	}
	132	$source = $rdg;
	133	}
78fab1cf	134	print STDERR "Adding a.c. version for witness $sig\n";
910a0a6d	135	$tradition->witness( $sig )->uncorrected_path( \@uncorrected );
	136	}
	137	}
	138	# Delete readings that are no longer part of the graph.
	139	# TODO think this is useless actually
	140	foreach ( keys %$substitutions ) {
	141	$tradition->collation->del_reading( $tradition->collation->reading( $_ ) );
	142	}
c9bf3dbf	143	$tradition->collation->calculate_ranks();
3bc0cd18	144
	145	# Now that we have ranks, see if we have distinct nodes with identical
	146	# text and identical rank that can be merged.
0e476982	147	$tradition->collation->flatten_ranks();
910a0a6d	148	}
	149
	150	sub _clean_sequence {
	151	my( $wit, $sequence ) = @_;
	152	my @clean_sequence;
	153	foreach my $rdg ( @$sequence ) {
	154	if( $rdg =~ /^PH-(.*)$/ ) {
	155	# It is a placeholder. Keep it only if we need it.
	156	my $app_id = $1;
78fab1cf	157	if( exists $app_ac->{$wit} &&
78fab1cf	158	exists $app_ac->{$wit}->{$app_id} ) {
910a0a6d	159	print STDERR "Retaining empty placeholder for $app_id\n";
	160	push( @clean_sequence, $rdg );
	161	}
	162	} else {
	163	push( @clean_sequence, $rdg );
	164	}
f6066bac	165	}
910a0a6d	166	return @clean_sequence;
910a0a6d	167	}
f6066bac	168
910a0a6d	169	sub _replace_sequence {
	170	my( $arr, $start, $end, @new ) = @_;
	171	my( $start_idx, $end_idx );
	172	foreach my $i ( 0 .. $#{$arr} ) {
	173	$start_idx = $i if( $arr->[$i]->name eq $start );
	174	if( $arr->[$i]->name eq $end ) {
	175	$end_idx = $i;
	176	last;
	177	}
	178	}
	179	unless( $start_idx && $end_idx ) {
	180	warn "Could not find start and end";
	181	return;
f2b9605f	182	}
910a0a6d	183	my $length = $end_idx - $start_idx + 1;
	184	splice( @$arr, $start_idx, $length, @new );
	185	}
f6066bac	186
910a0a6d	187	sub _return_rdg {
	188	my( $rdg ) = @_;
	189	# If we were passed a reading name, return the name. If we were
	190	# passed a reading object, return the object.
	191	my $wantobj = ref( $rdg ) eq 'Text::Tradition::Collation::Reading';
	192	my $real = $rdg;
	193	if( exists $substitutions->{ $wantobj ? $rdg->name : $rdg } ) {
	194	$real = $substitutions->{ $wantobj ? $rdg->name : $rdg };
	195	$real = $real->name unless $wantobj;
	196	}
	197	return $real;
f6066bac	198	}
f6066bac	199
910a0a6d	200	## Recursive helper function to help us navigate through nested XML,
	201	## picking out the text. $tradition is the tradition, needed for
	202	## making readings; $xn is the XML node currently being looked at,
	203	## $in_var is a flag to say that we are inside a variant, $ac is a
	204	## flag to say that we are inside an ante-correctionem reading, and
	205	## @cur_wits is the list of witnesses to which this XML node applies.
	206	## Returns the list of readings, if any, created on the run.
	207
	208	{
3bc0cd18	209	my %active_wits;
910a0a6d	210	my $current_app;
eca16057	211	my $seen_apps;
910a0a6d	212
	213	sub _get_readings {
	214	my( $tradition, $xn, $in_var, $ac, @cur_wits ) = @_;
3bc0cd18	215	@cur_wits = grep { $active_wits{$_} } keys %active_wits unless $in_var;
910a0a6d	216
	217	my @new_readings;
	218	if( $xn->nodeType == XML_TEXT_NODE ) {
	219	# Some words, thus make some readings.
	220	my $str = $xn->data;
	221	return unless $str =~ /\S/; # skip whitespace-only text nodes
	222	#print STDERR "Handling text node " . $str . "\n";
	223	# Check that all the witnesses we have are active.
	224	foreach my $c ( @cur_wits ) {
3bc0cd18	225	warn "$c is not among active wits" unless $active_wits{$c};
910a0a6d	226	}
	227	$str =~ s/^\s+//;
	228	my $final = $str =~ s/\s+$//;
	229	foreach my $w ( split( /\s+/, $str ) ) {
	230	# For now, skip punctuation.
	231	next if $w !~ /[[:alnum:]]/;
	232	my $rdg = make_reading( $tradition->collation, $w );
	233	push( @new_readings, $rdg );
	234	unless( $in_var ) {
910a0a6d	235	$rdg->make_common;
	236	}
	237	foreach ( @cur_wits ) {
	238	warn "Empty wit!" unless $_;
	239	warn "Empty reading!" unless $rdg;
	240	push( @{$text->{$_}}, $rdg ) unless $ac;
	241	}
	242	}
	243	} elsif( $xn->nodeName eq 'w' ) {
	244	# Everything in this tag is one word. Also save any original XML ID.
	245	#print STDERR "Handling word " . $xn->toString . "\n";
	246	# Check that all the witnesses we have are active.
	247	foreach my $c ( @cur_wits ) {
3bc0cd18	248	warn "$c is not among active wits" unless $active_wits{$c};
910a0a6d	249	}
	250	my $xml_id = $xn->getAttribute( 'xml:id' );
	251	my $rdg = make_reading( $tradition->collation, $xn->textContent, $xml_id );
	252	push( @new_readings, $rdg );
	253	unless( $in_var ) {
910a0a6d	254	$rdg->make_common;
	255	}
	256	foreach( @cur_wits ) {
	257	warn "Empty wit!" unless $_;
	258	warn "Empty reading!" unless $rdg;
	259	push( @{$text->{$_}}, $rdg ) unless $ac;
	260	}
	261	} elsif ( $xn->nodeName eq 'app' ) {
eca16057	262	$seen_apps++;
910a0a6d	263	$current_app = $xn->getAttribute( 'xml:id' );
	264	# print STDERR "Handling app $current_app\n";
	265	# Keep the reading sets in this app.
	266	my @sets;
	267	# Recurse through all children (i.e. rdgs) for sets of words.
	268	foreach ( $xn->childNodes ) {
	269	my @rdg_set = _get_readings( $tradition, $_, $in_var, $ac, @cur_wits );
	270	push( @sets, \@rdg_set ) if @rdg_set;
	271	}
	272	# Now collate these sets if we have more than one.
	273	my $subs = collate_variants( $tradition->collation, @sets ) if @sets > 1;
	274	map { $substitutions->{$_} = $subs->{$_} } keys %$subs;
	275	# TODO Look through substitutions to see if we can make anything common now.
	276	# Return the entire set of unique readings.
	277	my %unique;
	278	foreach my $s ( @sets ) {
	279	map { $unique{$_->name} = $_ } @$s;
	280	}
	281	push( @new_readings, values( %unique ) );
	282	# Exit the current app.
	283	$current_app = '';
	284	} elsif ( $xn->nodeName eq 'lem' \|\| $xn->nodeName eq 'rdg' ) {
	285	# Alter the current witnesses and recurse.
	286	#print STDERR "Handling reading for " . $xn->getAttribute( 'wit' ) . "\n";
3bc0cd18	287	# TODO handle p.c. and s.l. designations too
910a0a6d	288	$ac = $xn->getAttribute( 'type' ) && $xn->getAttribute( 'type' ) eq 'a.c.';
	289	my @rdg_wits = get_sigla( $xn );
	290	@rdg_wits = ( 'base' ) unless @rdg_wits; # Allow for editorially-supplied readings
	291	my @words;
	292	foreach ( $xn->childNodes ) {
	293	my @rdg_set = _get_readings( $tradition, $_, 1, $ac, @rdg_wits );
	294	push( @words, @rdg_set ) if @rdg_set;
	295	}
	296	# If we have more than one word in a reading, it should become a segment.
	297	# $tradition->collation->add_segment( @words ) if @words > 1;
	298
	299	if( $ac ) {
	300	# Add the reading set to the a.c. readings.
	301	foreach ( @rdg_wits ) {
	302	$app_ac->{$_}->{$current_app} = \@words;
	303	}
	304	} else {
	305	# Add the reading set to the app anchors for each witness
	306	# or put in placeholders for empty p.c. readings
	307	foreach ( @rdg_wits ) {
	308	my $start = @words ? $words[0]->name : "PH-$current_app";
	309	my $end = @words ? $words[-1]->name : "PH-$current_app";
	310	$app_anchors->{$current_app}->{$_}->{'start'} = $start;
	311	$app_anchors->{$current_app}->{$_}->{'end'} = $end;
	312	push( @{$text->{$_}}, $start ) unless @words;
	313	}
	314	}
	315	push( @new_readings, @words );
	316	} elsif( $xn->nodeName eq 'witStart' ) {
	317	# Add the relevant wit(s) to the active list.
	318	#print STDERR "Handling witStart\n";
3bc0cd18	319	map { $active_wits{$_} = 1 } @cur_wits;
	320	# Record a lacuna in all non-active witnesses if this is
	321	# the first app. Get the full list from $text.
	322	if( $seen_apps == 1 ) {
	323	my $i = 0;
	324	foreach my $sig ( keys %$text ) {
	325	next if $active_wits{$sig};
	326	my $l = $tradition->collation->add_lacuna( $current_app . "_$i" );
	327	$i++;
	328	push( @{$text->{$sig}}, $l );
	329	}
	330	}
910a0a6d	331	} elsif( $xn->nodeName eq 'witEnd' ) {
	332	# Take the relevant wit(s) out of the list.
	333	#print STDERR "Handling witEnd\n";
3bc0cd18	334	map { $active_wits{$_} = undef } @cur_wits;
eca16057	335	# Record a lacuna, unless this is the last app.
	336	unless( $seen_apps == $app_count ) {
	337	foreach my $i ( 0 .. $#cur_wits ) {
	338	my $w = $cur_wits[$i];
	339	my $l = $tradition->collation->add_lacuna( $current_app . "_$i" );
	340	push( @{$text->{$w}}, $l );
	341	}
	342	}
910a0a6d	343	} elsif( $xn->nodeName eq 'witDetail' ) {
	344	# Ignore these for now.
	345	return;
	346	} else {
	347	# Recurse as if this tag weren't there.
	348	#print STDERR "Recursing on tag " . $xn->nodeName . "\n";
	349	foreach( $xn->childNodes ) {
	350	push( @new_readings, _get_readings( $tradition, $_, $in_var, $ac, @cur_wits ) );
	351	}
	352	}
	353	return @new_readings;
	354	}
	355
	356	}
	357
	358	# Helper to extract a list of witness sigla from a reading element.
f6066bac	359	sub get_sigla {
	360	my( $rdg ) = @_;
	361	# Cope if we have been handed a NodeList. There is only
	362	# one reading here.
	363	if( ref( $rdg ) eq 'XML::LibXML::NodeList' ) {
910a0a6d	364	$rdg = $rdg->shift;
f6066bac	365	}
	366
	367	my @wits;
	368	if( ref( $rdg ) eq 'XML::LibXML::Element' ) {
910a0a6d	369	my $witstr = $rdg->getAttribute( 'wit' );
	370	$witstr =~ s/^\s+//;
	371	$witstr =~ s/\s+$//;
	372	@wits = split( /\s+/, $witstr );
	373	map { $_ =~ s/^\#// } @wits;
f6066bac	374	}
	375	return @wits;
	376	}
	377
910a0a6d	378	# Helper with its counters to actually make the readings.
f2b9605f	379	{
	380	my $word_ctr = 0;
	381	my %used_nodeids;
	382
910a0a6d	383	sub save_preexisting_nodeids {
	384	foreach( @_ ) {
	385	$used_nodeids{$_->getValue()} = 1;
	386	}
	387	}
	388
f2b9605f	389	sub make_reading {
910a0a6d	390	my( $graph, $word, $xml_id ) = @_;
	391	if( $xml_id ) {
	392	if( exists $used_nodeids{$xml_id} ) {
	393	if( $used_nodeids{$xml_id} != 1 ) {
	394	warn "Already used assigned XML ID somewhere else!";
	395	$xml_id = undef;
	396	}
	397	} else {
	398	warn "Undetected pre-existing XML ID";
	399	}
	400	}
	401	if( !$xml_id ) {
	402	until( $xml_id ) {
	403	my $try_id = 'w'.$word_ctr++;
	404	next if exists $used_nodeids{$try_id};
	405	$xml_id = $try_id;
	406	}
	407	}
	408	my $rdg = $graph->add_reading( $xml_id );
	409	$rdg->text( $word );
	410	$used_nodeids{$xml_id} = $rdg;
	411	return $rdg;
f2b9605f	412	}
	413	}
	414
f6066bac	415	1;