Other packages in the module: Bio::DB::GFF

Toolbar

Summary

Bio::DB::GFF -- Storage and retrieval of sequence annotation data

Package variables

Included modules

Inherit

Bio::DasI Bio::Root::Root

Synopsis

  use Bio::DB::GFF;

  # Open the sequence database
  my $db      = Bio::DB::GFF->new( -adaptor => 'dbi::mysqlopt',
                                   -dsn     => 'dbi:mysql:elegans',
				   -fasta   => '/usr/local/fasta_files'
				 );

  # fetch a 1 megabase segment of sequence starting at landmark "ZK909"
  my $segment = $db->segment('ZK909', 1 => 1000000);

  # pull out all transcript features
  my @transcripts = $segment->features('transcript');

  # for each transcript, total the length of the introns
  my %totals;
  for my $t (@transcripts) {
    my @introns = $t->Intron;
    $totals{$t->name} += $_->length foreach @introns;
  }

  # Sort the exons of the first transcript by position
  my @exons = sort {$a->start <=> $b->start} $transcripts[0]->Exon;

  # Get a region 1000 bp upstream of first exon
  my $upstream = $exons[0]->segment(-1000,0);

  # get its DNA
  my $dna = $upstream->dna;

  # and get all curated polymorphisms inside it
  @polymorphisms = $upstream->contained_features('polymorphism:curated');

  # get all feature types in the database
  my @types = $db->types;

  # count all feature types in the segment
  my %type_counts = $segment->types(-enumerate=>1);

  # get an iterator on all curated features of type 'exon' or 'intron'
  my $iterator = $db->get_seq_stream(-type     => ['exon:curated','intron:curated']);

  while (my $s = $iterator->next_seq) {
      print $s,"\n";
  }

  # find all transcripts annotated as having function 'kinase'
  my $iterator = $db->get_seq_stream(-type=>'transcript',
			             -attributes=>{Function=>'kinase'});
  while (my $s = $iterator->next_seq) {
      print $s,"\n";
  }

Description

Bio::DB::GFF provides fast indexed access to a sequence annotation
database. It supports multiple database types (ACeDB, relational),
and multiple schemas through a system of adaptors and aggregators.
The following operations are supported by this module:

  - retrieving a segment of sequence based on the ID of a landmark
  - retrieving the DNA from that segment
  - finding all annotations that overlap with the segment
  - finding all annotations that are completely contained within the
    segment
  - retrieving all annotations of a particular type, either within a
    segment, or globally
  - conversion from absolute to relative coordinates and back again,
    using any arbitrary landmark for the relative coordinates
  - using a sequence segment to create new segments based on relative 
    offsets

The data model used by Bio::DB::GFF is compatible with the GFF flat
file format (http://www.sanger.ac.uk/localsw/GFF). The module can
load a set of GFF files into the database, and serves objects that
have methods corresponding to GFF fields.
The objects returned by Bio::DB::GFF are compatible with the
SeqFeatureI interface, allowing their use by the Bio::Graphics and
Bio::DAS modules. The bioperl distribution includes several scripts that make it easier
to work with Bio::DB::GFF databases. They are located in the scripts
directory under a subdirectory named Bio::DB::GFF:
    This script will load a Bio::DB::GFF database from a flat GFF file of
sequence annotations. Only the relational database version of
Bio::DB::GFF is supported. It can be used to create the database from
scratch, as well as to incrementally load new data.
    This script takes a --fasta argument to load raw DNA into the database
as well. However, GFF databases do not require access to the raw DNA
for most of their functionality.
    load_gff.pl also has a --upgrade option, which will perform a
non-destructive upgrade of older schemas to newer ones.
    This script will populate a Bio::DB::GFF database from a flat GFF file
of sequence annotations. Only the MySQL database version of
Bio::DB::GFF is supported. It uses the "LOAD DATA INFILE" query in
order to accelerate loading considerably; however, it can only be used
for the initial load, and not for updates.
    This script takes a --fasta argument to load raw DNA into the database
as well. However, GFF databases do not require access to the raw DNA
for most of their functionality.
    This script is as fast as bp_bulk_load_gff.pl but uses Unix pipe
tricks to allow for incremental updates. It only supports the MySQL
database version of Bio::DB::GFF and is guaranteed not to work on
non-Unix platforms.
    Arguments are the same as bp_load_gff.pl
    This script will convert the GFF-like format used by the Berkeley
Drosophila Sequencing project into a format suitable for use with this
module.
    This script will convert the tab-delimited feature files used by the
Saccharomyces Genome Database into a format suitable for use with this
module.
The GFF format is a flat tab-delimited file, each line of which
corresponds to an annotation, or feature. Each line has nine columns
and looks like this:

 Chr1  curated  CDS 365647  365963  .  +  1  Transcript "R119.7"

The 9 columns are as follows:
    This is the ID of the sequence that is used to establish the
coordinate system of the annotation. In the example above, the
reference sequence is "Chr1".
    The source of the annotation. This field describes how the annotation
was derived. In the example above, the source is "curated" to
indicate that the feature is the result of human curation. The names
and versions of software programs are often used for the source field,
as in "tRNAScan-SE/1.2".
    The annotation method. This field describes the type of the
annotation, such as "CDS". Together the method and source describe
the annotation type.
    The start of the annotation relative to the reference sequence.
    The stop of the annotation relative to the reference sequence. Start
is always less than or equal to stop.
    For annotations that are associated with a numeric score (for example,
a sequence similarity), this field describes the score. The score
units are completely unspecified, but for sequence similarities, it is
typically percent identity. Annotations that don't have a score can
use "."
    For those annotations which are strand-specific, this field is the
strand on which the annotation resides. It is "+" for the forward
strand, "-" for the reverse strand, or "." for annotations that are
not stranded.
    For annotations that are linked to proteins, this field describes the
phase of the annotation on the codons. It is a number from 0 to 2, or
"." for features that have no phase\.
    GFF provides a simple way of generating annotation hierarchies ("is
composed of" relationships) by providing a group field. The group
field contains the class and ID of an annotation which is the logical
parent of the current one. In the example given above, the group is
the Transcript named "R119.7".
    The group field is also used to store information about the target of
sequence similarity hits, and miscellaneous notes. See the next
section for a description of how to describe similarity targets.
    The format of the group fields is "Class ID" with a single space (not
a tab) separating the class from the ID. It is VERY IMPORTANT to
follow this format, or grouping will not work properly.
The sequences used to establish the coordinate system for annotations
can correspond to sequenced clones, clone fragments, contigs or
super-contigs. Thus, this module can be used throughout the lifecycle
of a sequencing project.
In addition to a group ID, the GFF format allows annotations to have a
group class. For example, in the ACeDB representation, RNA
interference experiments have a class of "RNAi" and an ID that is
unique among the RNAi experiments. Since not all databases support
this notion, the class is optional in all calls to this module, and
defaults to "Sequence" when not provided.
Double-quotes are sometimes used in GFF files around components of the
group field. Strictly, this is only necessary if the group name or
class contains whitespace. Some annotations do not need to be individually named. For example,
it is probably not useful to assign a unique name to each ALU repeat
in a vertebrate genome. Others, such as predicted genes, correspond
to named biological objects; you probably want to be able to fetch the
positions of these objects by referring to them by name.
To accomodate named annotations, the GFF format places the object
class and name in the group field. The name identifies the object,
and the class prevents similarly-named objects, for example clones and
sequences, from collding.
A named object is shown in the following excerpt from a GFF file:

 Chr1  curated transcript  939627 942410 . +  . Transcript Y95B8A.2

This object is a predicted transcript named Y95BA.2. In this case,
the group field is used to identify the class and name of the object,
even though no other annotation belongs to that group.
It now becomes possible to retrieve the region of the genome covered
by transcript Y95B8A.2 using the segment() method:

  $segment = $db->segment(-class=>'Transcript',-name=>'Y95B8A.2');

It is not necessary for the annotation's method to correspond to the
object class, although this is commonly the case.
As explained above, each annotation in a GFF file refers to a
reference sequence. It is important that each reference sequence also
be identified by a line in the GFF file. This allows the Bio::DB::GFF
module to determine the length and class of the reference sequence,
and makes it possible to do relative arithmetic.
For example, if "Chr1" is used as a reference sequence, then it should
have an entry in the GFF file similar to this one:

 Chr1 assembly chromosome 1 14972282 . + . Sequence Chr1

This indicates that the reference sequence named "Chr1" has length
14972282 bp, method "chromosome" and source "assembly". In addition,
as indicated by the group field, Chr1 has class "Sequence" and name
"Chr1".
The object class "Sequence" is used by default when the class is not
specified in the segment() call. This allows you to use a shortcut
form of the segment() method:

 $segment = $db->segment('Chr1');          # whole chromosome
 $segment = $db->segment('Chr1',1=>1000);  # first 1000 bp

For your convenience, if, during loading a GFF file, Bio::DB::GFF
encounters a line like the following:

  ##sequence-region Chr1 1 14972282

It will automatically generate the following entry:

 Chr1 reference Component 1 14972282 . + . Sequence Chr1

This is sufficient to use Chr1 as a reference point.
The ##sequence-region line is frequently found in the GFF files
distributed by annotation groups. There are two cases in which an annotation indicates the relationship
between two sequences. The first case is a similarity hit, where the
annotation indicates an alignment. The second case is a map assembly,
in which the annotation indicates that a portion of a larger sequence
is built up from one or more smaller ones.
Both cases are indicated by using the Target tag in the group
field. For example, a typical similarity hit will look like this:

 Chr1 BLASTX similarity 76953 77108 132 + 0 Target Protein:SW:ABL_DROME 493 544

The group field contains the Target tag, followed by an identifier for
the biological object referred to. The GFF format uses the notation
Class:Name for the biological object, and even though this is
stylistically inconsistent, that's the way it's done. The object
identifier is followed by two integers indicating the start and stop
of the alignment on the target sequence.
Unlike the main start and stop columns, it is possible for the target
start to be greater than the target end. The previous example
indicates that the the section of Chr1 from 76,953 to 77,108 aligns to
the protein SW:ABL_DROME starting at position 493 and extending to
position 544.
A similar notation is used for sequence assembly information as shown
in this example:

 Chr1        assembly Link   10922906 11177731 . . . Target Sequence:LINK_H06O01 1 254826
 LINK_H06O01 assembly Cosmid 32386    64122    . . . Target Sequence:F49B2       6 31742

This indicates that the region between bases 10922906 and 11177731 of
Chr1 are composed of LINK_H06O01 from bp 1 to bp 254826. The region
of LINK_H0601 between 32386 and 64122 is, in turn, composed of the
bases 5 to 31742 of cosmid F49B2. While not intended to serve as a general-purpose sequence database
(see bioperl-db for that), GFF allows you to tag features with
arbitrary attributes. Attributes appear in the Group field following
the initial class/name pair. For example:

 Chr1  cur trans  939 942 . +  . Transcript Y95B8A.2 ; Gene sma-3 ; Alias sma3

This line tags the feature named Transcript Y95B8A.2 as being "Gene"
named sma-3 and having the Alias "sma3". Features having these
attributes can be looked up using the fetch_feature_by_attribute() method.
Two attributes have special meaning: "Note" is for backward
compatibility and is used for unstructured text remarks. "Alias" is
considered as a synonym for the feature name and will be consulted
when looking up a feature by its name. This module uses a system of adaptors and aggregators in order to make
it adaptable to use with a variety of databases.
    The core of the module handles the user API, annotation coordinate
arithmetic, and other common issues. The details of fetching
information from databases is handled by an adaptor, which is
specified during Bio::DB::GFF construction. The adaptor encapsulates
database-specific information such as the schema, user authentication
and access methods.
    Currently there are two adaptors: 'dbi::mysql' and 'dbi::mysqlopt'.
The former is an interface to a simple Mysql schema. The latter is an
optimized version of dbi::mysql which uses a binning scheme to
accelerate range queries and the Bio::DB::Fasta module for rapid
retrieval of sequences. Note the double-colon between the words.
    The GFF format uses a "group" field to indicate aggregation properties
of individual features. For example, a set of exons and introns may
share a common transcript group, and multiple transcripts may share
the same gene group.
    Aggregators are small modules that use the group information to
rebuild the hierarchy. When a Bio::DB::GFF object is created, you
indicate that it use a set of one or more aggregators. Each
aggregator provides a new composite annotation type. Before the
database query is generated each aggregator is called to
"disaggregate" its annotation type into list of component types
contained in the database. After the query is generated, each
aggregator is called again in order to build composite annotations
from the returned components.
    For example, during disaggregation, the standard
"processed_transcript" aggregator generates a list of component
feature types including "UTR", "CDS", and "polyA_site". Later, it
aggregates these features into a set of annotations of type
"processed_transcript".
    During aggregation, the list of aggregators is called in reverse
order. This allows aggregators to collaborate to create multi-level
structures: the transcript aggregator assembles transcripts from
introns and exons; the gene aggregator then assembles genes from sets
of transcripts.
    Three default aggregators are provided:

      transcript   assembles transcripts from features of type
                   exon, CDS, 5'UTR, 3'UTR, TSS, and PolyA
      clone        assembles clones from Clone_left_end, Clone_right_end
                   and Sequence features.
      alignment    assembles gapped alignments from features of type
                   "similarity".

    In addition, this module provides the optional "wormbase_gene"
aggregator, which accomodates the WormBase representation of genes.
This aggregator aggregates features of method "exon", "CDS", "5'UTR",
"3'UTR", "polyA" and "TSS" into a single object. It also expects to
find a single feature of type "Sequence" that spans the entire gene.
    The existing aggregators are easily customized.
    Note that aggregation will not occur unless you specifically request
the aggregation type. For example, this call:

  @features = $segment->features('alignment');

    will generate an array of aggregated alignment features. However,
this call:

  @features = $segment->features();

    will return a list of unaggregated similarity segments.
    For more informnation, see the manual pages for
Bio::DB::GFF::Aggregator::processed_transcript, Bio::DB::GFF::Aggregator::clone,
etc.

Methods

Methods description

 Title   : _delete_features(), _delete_groups(),_delete()
 Usage   : $count = $db->_delete_features(@feature_ids)
           $count = $db->_delete_groups(@group_ids)
           $count = $db->_delete(\%delete_spec)
 Function: low-level feature/group deleter
 Returns : count of groups removed
 Args    : list of feature or group ids removed
 Status  : for implementation by subclasses

 Title   : _feature_by_id
 Usage   : $db->_feature_by_id($ids,$type,$callback)
 Function: get a feature based
 Returns : count of number of features retrieved
 Args    : arrayref to feature IDs to fetch
 Status  : abstract

 Title   : _feature_by_name
 Usage   : $db->_feature_by_name($class,$name,$location,$callback)
 Function: get a list of features by name and class
 Returns : count of number of features retrieved
 Args    : name of feature, class of feature, and a callback
 Status  : abstract

 Title   : _features
 Usage   : $db->_features($search,$options,$parent)
 Function: internal method
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : see below
 Status  : internal

  rangetype     One of "overlaps", "contains" or "contained_in".  Indicates
                the type of range query requested.
  refseq        reference sequence ID
  refclass      reference sequence class
  start	        start of range
  stop		stop of range
  types	        arrayref containing list of types in "method:source" form

  sparse	turn on optimizations for a rare feature
  automerge	if true, invoke aggregators to merge features
  iterator	if true, return an iterator

 Title   : abscoords
 Usage   : $db->abscoords($name,$class,$refseq)
 Function: finds position of a landmark in reference coordinates
 Returns : ($ref,$class,$start,$stop,$strand)
 Args    : name and class of landmark
 Status  : public

 Title   : absolute
 Usage   : $abs = $db->absolute([$abs]);
 Function: gets/sets absolute mode
 Returns : current setting of absolute mode boolean
 Args    : new setting for absolute mode boolean
 Status  : public

 Title   : add_aggregator
 Usage   : $db->add_aggregator($aggregator)
 Function: add an aggregator to the list
 Returns : nothing
 Args    : an aggregator
 Status  : public

  1) a Bio::DB::GFF::Aggregator object -- will be added
  2) a string in the form "aggregator_name{subpart1,subpart2,subpart3/main_method}"
         -- will be turned into a Bio::DB::GFF::Aggregator object (the /main_method
        part is optional).
  3) a valid Perl token -- will be turned into a Bio::DB::GFF::Aggregator
        subclass, where the token corresponds to the subclass name.

 Title   : aggregators
 Usage   : $db->aggregators([@new_aggregators]);
 Function: retrieve list of aggregators
 Returns : list of aggregators
 Args    : a list of aggregators to set (optional)
 Status  : public

 Title   : all_seqfeatures
 Usage   : @features = $db->all_seqfeatures(@args)
 Function: fetch all the features in the database
 Returns : an array of features, or an iterator
 Args    : See below
 Status  : public

 Title   : attributes
 Usage   : @attributes = $db->attributes($id,$name)
 Function: get the "attributres" on a particular feature
 Returns : an array of string
 Args    : feature ID
 Status  : public

  %attributes = $db->attributes($id);

  @notes = $feature->attributes('Note');

 Title   : automerge
 Usage   : $db->automerge( [$new automerge] );
 Function: get or set automerge value
 Returns : current value (boolean)
 Args    : an optional argument to set the automerge value
 Status  : Public

 Title   : classes
 Usage   : $db->classes
 Function: return list of landmark classes in database
 Returns : a list of classes
 Args    : none
 Status  : public

 Title   : clear_aggregators
 Usage   : $db->clear_aggregators
 Function: clears list of aggregators
 Returns : nothing
 Args    : none
 Status  : public

 Title   : contained_features
 Usage   : $db->contained_features(@args)
 Function: get features that are contained within the indicated range
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : see overlapping_features()
 Status  : public

 Title   : contained_in
 Usage   : @features = $s->contained_in(@args)
 Function: get features that contain this segment
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : see features()
 Status  : Public

 Title   : debug
 Usage   : $db->debug( [$flag] );
 Function: read or set debug flag
 Returns : current value of debug flag
 Args    : new debug flag (optional)
 Status  : Public

 Title   : default_aggregators
 Usage   : $db->default_aggregators;
 Function: retrieve list of aggregators
 Returns : array reference containing list of aggregator names
 Args    : none
 Status  : protected

 Title   : default_meta_values
 Usage   : %values = $db->default_meta_values
 Function: empty the database
 Returns : a list of tag=>value pairs
 Args    : none
 Status  : protected

 Title   : delete
 Usage   : $db->delete(@args)
 Function: delete features
 Returns : count of features deleted -- if available
 Args    : numerous, see below
 Status  : public

 -name         ID of the landmark sequence.

 -ref          ID of the landmark sequence (synonym for -name).

 -class        Database object class for the landmark sequence.
               "Sequence" assumed if not specified.  This is
               irrelevant for databases which do not recognize
               object classes.

 -start        Start of the segment relative to landmark.  Positions
               follow standard 1-based sequence rules.  If not specified,
               defaults to the beginning of the landmark.

 -end          Stop of the segment relative to the landmark.  If not specified,
               defaults to the end of the landmark.

 -offset       Zero-based addressing

 -length       Length of region

 -type,-types  Either a single scalar type to be deleted, or an
               reference to an array of types.

 -force        Force operation to be performed even if it would delete
               entire feature table.

 -range_type   Control the range type of the deletion.  One of "overlaps" (default)
               "contains" or "contained_in"

  $db->delete(-type=>['intron','repeat:repeatMasker']);  # remove all introns & repeats
  $db->delete(-name=>'chr3',-start=>1,-end=>1000);       # remove annotations on chr3 from 1 to 1000
  $db->delete(-name=>'chr3',-type=>'exon');              # remove all exons on chr3

  $db->delete("chr3",1=>1000);
  $db->delete("chr3");

 Title   : delete_features
 Usage   : $db->delete_features(@ids_or_features)
 Function: delete one or more features
 Returns : count of features deleted
 Args    : list of features or feature ids
 Status  : public

 Title   : delete_groups
 Usage   : $db->delete_groups(@ids_or_features)
 Function: delete one or more feature groups
 Returns : count of features deleted
 Args    : list of features or feature group ids
 Status  : public

 Title   : dna
 Usage   : $db->dna($id,$start,$stop,$class)
 Function: return the raw DNA string for a segment
 Returns : a raw DNA string
 Args    : id of the sequence, its class, start and stop positions
 Status  : public

  $db->dna($name_of_sequence);

   $db->dna(-name=>$name_of_sequence,-class=>$class_of_sequence);

 Title   : do_attributes
 Usage   : $db->do_attributes($id [,$tag]);
 Function: internal method to retrieve attributes given an id and tag
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : a feature id and a attribute tag (optional)
 Status  : protected

 Title   : do_initialize
 Usage   : $db->do_initialize([$erase])
 Function: initialize and possibly erase database
 Returns : true if successful
 Args    : optional erase flag
 Status  : protected

 Title   : do_load_gff
 Usage   : $db->do_load_gff($handle)
 Function: load a GFF input stream
 Returns : number of features loaded
 Args    : A filehandle.
 Status  : protected

 Title   : error
 Usage   : $db->error( [$new error] );
 Function: read or set error message
 Returns : error message
 Args    : an optional argument to set the error message
 Status  : Public

 Title   : fast_queries
 Usage   : $flag = $db->fast_queries([$flag])
 Function: turn on and off the "fast queries" option
 Returns : a boolean
 Args    : a boolean flag (optional)
 Status  : public

 Title   : features
 Usage   : $db->features(@args)
 Function: get all features, possibly filtered by type
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : see below
 Status  : public

  -types     List of feature types to return.  Argument is an array
	     reference containing strings of the format "method:source"

  -merge     Whether to apply aggregators to the generated features.

  -rare      Turn on optimizations suitable for a relatively rare feature type,
             where it makes more sense to filter by feature type first,
             and then by position.

  -attributes A hash reference containing attributes to match.

  -iterator  Whether to return an iterator across the features.

  -binsize   A true value will create a set of artificial features whose
             start and stop positions indicate bins of the given size, and
             whose scores are the number of features in the bin.  The
             class and method of the feature will be set to "bin",
             its source to "method:source", and its group to "bin:method:source".
             This is a handy way of generating histograms of feature density.

  -attributes => { Gene => 'abc-1',
                   Note => 'confirmed' }

 Title   : finish_load
 Usage   : $db->finish_load
 Function: called after load_gff_line()
 Returns : number of records loaded
 Args    : none
 Status  :abstract

 Title   : get_Seq_by_accession
 Usage   : $seq = $db->get_Seq_by_accession('AL12234')
 Function: Gets a Bio::Seq object by its accession
 Returns : a Bio::Seq object
 Args    : the id (as a string) of a sequence
 Throws  : "id does not exist" exception

 Title   : get_Seq_by_id
 Usage   : $seq = $db->get_Seq_by_id('ROA1_HUMAN')
 Function: Gets a Bio::Seq object by its name
 Returns : a Bio::Seq object
 Args    : the id (as a string) of a sequence
 Throws  : "id does not exist" exception

  Title   : get_Stream_by_id
  Usage   : $seq = $db->get_Stream_by_id(@ids);
  Function: Retrieves a stream of Seq objects given their ids
  Returns : a Bio::SeqIO stream object
  Args    : an array of unique ids/accession numbers, or 
            an array reference

 Title   : get_abscoords
 Usage   : $db->get_abscoords($name,$class,$refseq)
 Function: get the absolute coordinates of sequence with name & class
 Returns : ($absref,$absstart,$absstop,$absstrand)
 Args    : name and class of the landmark
 Status  : protected

  $absref      the ID of the reference sequence that contains this landmark
  $absstart    the position at which the landmark starts
  $absstop     the position at which the landmark stops
  $absstrand   the strand of the landmark, relative to the reference sequence

 Title   : get_dna
 Usage   : $db->get_dna($id,$start,$stop,$class)
 Function: get DNA for indicated segment
 Returns : the dna string
 Args    : sequence ID, start, stop and class
 Status  : protected

 Title   : get_feature_by_attribute
 Usage   : $db->get_feature_by_attribute(attribute1=>value1,attribute2=>value2)
 Function: fetch segments by combinations of attribute values
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : the class and name of the desired feature
 Status  : public

 Title   : get_feature_by_gid
 Usage   : $db->get_feature_by_gid($id)
 Function: fetch segments by feature ID
 Returns : a Bio::DB::GFF::Feature object
 Args    : the feature ID
 Status  : public

 Title   : get_feature_by_id
 Usage   : $db->get_feature_by_id($id)
 Function: fetch segments by feature ID
 Returns : a Bio::DB::GFF::Feature object
 Args    : the feature ID
 Status  : public

 Title   : get_feature_by_name
 Usage   : $db->get_feature_by_name($class => $name)
 Function: fetch features by their name
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : the class and name of the desired feature
 Status  : public

  @f = $db->get_feature_by_name(-class => $class,-name=>$name,
                                -ref   => $sequence_name,
                                -start => $start,
                                -end   => $end);

 Title   : get_feature_by_target
 Usage   : $db->get_feature_by_target($class => $name)
 Function: fetch features by their similarity target
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : the class and name of the desired feature
 Status  : public

 Title   : get_features
 Usage   : $db->get_features($search,$options,$callback)
 Function: get list of features for a region
 Returns : count of number of features retrieved
 Args    : see below
 Status  : protected

   rangetype One of "overlaps", "contains" or "contained_in".  Indicates
              the type of range query requested.

   refseq    ID of the landmark that establishes the absolute 
              coordinate system.

   refclass  Class of this landmark.  Can be ignored by implementations
              that don't recognize such distinctions.

   start     Start of the range, inclusive.

   stop      Stop of the range, inclusive.

   types     Array reference containing the list of annotation types
              to fetch from the database.  Each annotation type is an
              array reference consisting of [source,method].

   sort_by_group
             A flag.  If true, means that the returned features should be
             sorted by the group that they're in.

   sparse    A flag.  If true, means that the expected density of the 
             features is such that it will be more efficient to search
             by type rather than by range.  If it is taking a long
             time to fetch features, give this a try.

   binsize   A true value will create a set of artificial features whose
             start and stop positions indicate bins of the given size, and
             whose scores are the number of features in the bin.  The
             class of the feature will be set to "bin", and its name to
             "method:source".  This is a handy way of generating histograms
             of feature density.

  $refseq      The reference sequence
  $start       feature start
  $stop        feature stop
  $source      feature source
  $method      feature method
  $score       feature score
  $strand      feature strand
  $phase       feature phase
  $groupclass  group class (may be undef)
  $groupname   group ID (may be undef)
  $tstart      target start for similarity hits (may be undef)
  $tstop       target stop for similarity hits (may be undef)
  $feature_id  A unique feature ID (may be undef)

 Title   : get_features_iterator
 Usage   : $db->get_features_iterator($search,$options,$callback)
 Function: get an iterator on a features query
 Returns : a Bio::SeqIO object
 Args    : as per get_features()
 Status  : Public

 Title   : get_seq_stream
 Usage   : my $seqio = $self->get_seq_sream(@args)
 Function: Performs a query and returns an iterator over it
 Returns : a Bio::SeqIO stream capable of producing sequence
 Args    : As in features()
 Status  : public

  $stream = $db->get_seq_stream('exon');
  while (my $exon = $stream->next_seq) {
     print $exon,"\n";
  }

 Title   : get_types
 Usage   : $db->get_types($absref,$class,$start,$stop,$count)
 Function: get list of all feature types on the indicated segment
 Returns : list or hash of Bio::DB::GFF::Typename objects
 Args    : see below
 Status  : protected

  $absref      the ID of the reference sequence
  $class       the class of the reference sequence
  $start       the position to start counting
  $stop        the position to end counting
  $count       a boolean indicating whether to count the number
	       of occurrences of each feature type

 Title   : initialize
 Usage   : $db->initialize(-erase=>$erase,-option1=>value1,-option2=>value2);
 Function: initialize a GFF database
 Returns : true if initialization successful
 Args    : a set of named parameters
 Status  : Public

  -erase     A boolean value.  If true the database will be wiped clean if it
             already contains data.

 Title   : load_fasta
 Usage   : $db->load_fasta($file|$directory|$filehandle);
 Function: load FASTA data into database
 Returns : count of records loaded
 Args    : a directory, a file, a list of files, 
           or a filehandle
 Status  : Public

  $db->load_gff("lynx -dump -source http://stein.cshl.org/fasta_test.fa |");

 Title   : load_gff
 Usage   : $db->load_gff($file|$directory|$filehandle);
 Function: load GFF data into database
 Returns : count of records loaded
 Args    : a directory, a file, a list of files, 
           or a filehandle
 Status  : Public

  $db->load_gff("lynx -dump -source http://stein.cshl.org/gff_test |");

 Title   : load_gff_line
 Usage   : $db->load_gff_line(@args)
 Function: called to load one parsed line of GFF
 Returns : true if successfully inserted
 Args    : see below
 Status  : abstract

 {ref    => $ref,
  class  => $class,
  source => $source,
  method => $method,
  start  => $start,
  stop   => $stop,
  score  => $score,
  strand => $strand,
  phase  => $phase,
  gclass => $gclass,
  gname  => $gname,
  tstart => $tstart,
  tstop  => $tstop,
  attributes  => $attributes}

 Title   : load_sequence
 Usage   : $db->load_sequence($handle [,$id])
 Function: load a FASTA data stream
 Returns : number of sequences
 Args    : a filehandle and optionally the ID of
  the first sequence in the stream.
 Status  : protected

 Title   : load_sequence_string
 Usage   : $db->load_sequence_string($id,$dna)
 Function: load a single DNA entry
 Returns : true if successfully loaded
 Args    : a raw sequence string (DNA, RNA, protein)
 Status  : Public

 Title   : lock_on_load
 Usage   : $lock = $db->lock_on_load([$lock])
 Function: set write locking during load
 Returns : current value of lock-on-load flag
 Args    : new value of lock-on-load-flag
 Status  : Public

 Title   : make_feature
 Usage   : $db->make_feature(@args)
 Function: Create a Bio::DB::GFF::Feature object from string data
 Returns : a Bio::DB::GFF::Feature object
 Args    : see below
 Status  : internal

 This takes 14 arguments (really!):

  $parent                A Bio::DB::GFF::RelSegment object
  $group_hash            A hashref containing unique list of GFF groups
  $refname               The name of the reference sequence for this feature
  $refclass              The class of the reference sequence for this feature
  $start                 Start of feature
  $stop                  Stop of feature
  $source                Feature source field
  $method                Feature method field
  $score                 Feature score field
  $strand                Feature strand
  $phase                 Feature phase
  $group_class           Class of feature group
  $group_name            Name of feature group
  $tstart                For homologies, start of hit on target
  $tstop                 Stop of hit on target

 Title   : make_match_sub
 Usage   : $db->make_match_sub($types)
 Function: creates a subroutine used for filtering features
 Returns : a code reference
 Args    : a list of parsed type names
 Status  : protected

 Title   : make_object
 Usage   : $db->make_object($class,$name,$start,$stop)
 Function: creates a feature object
 Returns : a feature object
 Args    : see below
 Status  : protected

  $name      database ID for object
  $class     class of object
  $start     for similarities, start of match inside object
  $stop      for similarities, stop of match inside object

 Title   : meta
 Usage   : $value = $db->meta($name [,$newval])
 Function: get or set a meta variable
 Returns : a string
 Args    : meta variable name and optionally value
 Status  : abstract

 Title   : new
 Usage   : my $db = new Bio::DB::GFF(@args);
 Function: create a new Bio::DB::GFF object
 Returns : new Bio::DB::GFF object
 Args    : lists of adaptors and aggregators
 Status  : Public

 -adaptor      Name of the adaptor module to use.  If none
               provided, defaults to "dbi::mysqlopt".

 -aggregator   Array reference to a list of aggregators
               to apply to the database.  If none provided,
	       defaults to ['processed_transcript','alignment'].

        Any other named argument pairs are passed to
               the adaptor for processing.

  my $transcript = 
     Bio::DB::Aggregator::transcript->new(-sub_parts=>[qw(exon intron utr
                                                          polyA spliced_leader)]);

  my $db = Bio::DB::GFF->new(-aggregator=>[$transcript,'clone','alignment],
                             -adaptor   => 'dbi::mysql',
                             -dsn      => 'dbi:mysql:elegans42');

  my $new_agg = 'transcript{exon,intron,utr,polyA,spliced_leader}';
  my $db      = Bio::DB::GFF->new(-aggregator=>[$new_agg,'clone','alignment],
                                  -adaptor   => 'dbi::mysql',
                                  -dsn       => 'dbi:mysql:elegans42');

  Argument       Description
  --------       -----------

  -dsn           the DBI data source, e.g. 'dbi:mysql:ens0040'
                 If a partial name is given, such as "ens0040", the
                 "dbi:mysql:" prefix will be added automatically.

  -user          username for authentication

  -pass          the password for authentication

  -refclass      landmark Class; defaults to "Sequence"

  Argument       Description
  --------       -----------

  -fasta         path to a directory containing FASTA files for the DNA
                 contained in this database (e.g. "/usr/local/share/fasta")

  -acedb         an acedb URL to use when converting features into ACEDB
                    objects (e.g. sace://localhost:2005)

 Title   : overlapping_features
 Usage   : $db->overlapping_features(@args)
 Function: get features that overlap the indicated range
 Returns : a list of Bio::DB::GFF::Feature objects
 Args    : see below
 Status  : public

  -refseq    ID of the reference sequence
  -class     Class of the reference sequence
  -start     Start of the desired range in refseq coordinates
  -stop      Stop of the desired range in refseq coordinates
  -types     List of feature types to return.  Argument is an array
	     reference containing strings of the format "method:source"
  -parent    A parent Bio::DB::GFF::Segment object, used to create
	     relative coordinates in the generated features.
  -rare      Turn on an optimization suitable for a relatively rare feature type,
             where it will be faster to filter by feature type first
             and then by position, rather than vice versa.
  -merge     Whether to apply aggregators to the generated features.
  -iterator  Whether to return an iterator across the features.

 Title   : parse_types
 Usage   : $db->parse_types(@args)
 Function: parses list of types
 Returns : an array ref containing ['method','source'] pairs
 Args    : a list of types in 'method:source' form
 Status  : internal

 Title   : segment
 Usage   : $db->segment(@args);
 Function: create a segment object
 Returns : segment object(s)
 Args    : numerous, see below
 Status  : public

 -name         ID of the landmark sequence.

 -class        Database object class for the landmark sequence.
               "Sequence" assumed if not specified.  This is
               irrelevant for databases which do not recognize
               object classes.

 -start        Start of the segment relative to landmark.  Positions
               follow standard 1-based sequence rules.  If not specified,
               defaults to the beginning of the landmark.

 -end          Stop of the segment relative to the landmark.  If not specified,
               defaults to the end of the landmark.

 -stop         Same as -end.

 -offset       For those who prefer 0-based indexing, the offset specifies the
               position of the new segment relative to the start of the landmark.

 -length       For those who prefer 0-based indexing, the length specifies the
               length of the new segment.

 -refseq       Specifies the ID of the reference landmark used to establish the
               coordinate system for the newly-created segment.

 -refclass     Specifies the class of the reference landmark, for those databases
               that distinguish different object classes.  Defaults to "Sequence".

 -absolute
               Return features in absolute coordinates rather than relative to the
               parent segment.

 -nocheck      Don't check the database for the coordinates and length of this
               feature.  Construct a segment using the indicated name as the
               reference, a start coordinate of 1, an undefined end coordinate,
               and a strand of +1.

 -force        Same as -nocheck.

 -seq,-sequence,-sourceseq   Aliases for -name.

 -begin,-end   Aliases for -start and -stop

 -off,-len     Aliases for -offset and -length

 -seqclass     Alias for -class

  my $db = Bio::DB::GFF->new(-dsn => 'dbi:mysql:human',-adaptor=>'dbi::mysql');

  my $segment = $db->segment(-name=>'A0000182',-class=>'Accession');

  my $sourceseq = $segment->sourceseq;

  my $start = $segment->abs_start;
  my $stop = $segment->abs_stop;

  $segment->refseq($s2);

  my $rel_start = $segment->start;

 Title   : setup_load
 Usage   : $db->setup_load
 Function: called before load_gff_line()
 Returns : void
 Args    : none
 Status  : abstract

 Title   : split_group
 Usage   : $db->split_group($group_field,$gff3_flag)
 Function: parse GFF group field
 Returns : ($gclass,$gname,$tstart,$tstop,$attributes)
 Args    : the gff group column and a flag indicating gff3 compatibility
 Status  : internal

 Title   : strict_bounds_checking
 Usage   : $flag = $db->strict_bounds_checking([$flag])
 Function: gets/sets strict bounds checking
 Returns : current setting of bounds checking flag
 Args    : new setting for bounds checking flag
 Status  : public

 Title   : types
 Usage   : $db->types(@args)
 Function: return list of feature types in range or database
 Returns : a list of Bio::DB::GFF::Typename objects
 Args    : see below
 Status  : public

  -ref        ID of reference sequence
  -class      class of reference sequence
  -start      start of segment
  -stop       stop of segment
  -enumerate  if true, count the features

Methods code

sub _delete {

  my $self = shift;
  my $delete_options = shift;
  $self->throw('_delete is not implemented in this adaptor');

}

sub _delete_features {

  my $self = shift;
  my @feature_ids = @_;
  $self->throw('_delete_features is not implemented in this adaptor');

}

sub _delete_groups {

  my $self = shift;
  my @group_ids = @_;
  $self->throw('_delete_groups is not implemented in this adaptor');

}

sub _feature_by_attribute {

  my $self = shift;
  my ($attributes,$callback) = @_;
  $self->throw("_feature_by_name() must be implemented by an adaptor");

}

sub _feature_by_id {

  my $self = shift;
  my ($ids,$type,$callback) = @_;
  $self->throw("_feature_by_id() must be implemented by an adaptor");

}

sub _feature_by_name {

  my $self = shift;
  my ($class,$name,$location,$callback) = @_;
  $self->throw("_feature_by_name() must be implemented by an adaptor");

}

sub _features {

  my $self = shift;
  my ($search,$options,$parent) = @_;
  (@{$search}{qw(start stop)}) = (@{$search}{qw(stop start)})
    if defined($search->{start}) && $search->{start} > $search->{stop};

  my $types = $self->parse_types($search->{types});  # parse out list of types
  my @aggregated_types = @$types;         # keep a copy

  # allow the aggregators to operate on the original
  my @aggregators;
  if ($options->{automerge}) {
    for my $a ($self->aggregators) {
      $a = $a->clone if $options->{iterator};
      unshift @aggregators,$a
	if $a->disaggregate(\@aggregated_types,$self);
    }
  }

  if ($options->{iterator}) {
    my @accumulated_features;
    my $callback = $options->{automerge} ? sub { $self->make_aggregated_feature(\@accumulated_features,$parent,\@aggregators,@_) }
                                         : sub { [$self->make_feature($parent,undef,@_)] };
    return $self->get_features_iterator({ %$search, 
					  types =>\@ aggregated_types  },
					{ %$options,
					 sort_by_group => $options->{automerge}  },
					$callback
				       );
  }

  my %groups;         # cache the groups we create to avoid consuming too much unecessary memory
  my $features = [];

  my $callback = sub { push @$features,$self->make_feature($parent,\%groups,@_) };
  $self->get_features({ %$search,
			types  =>\@ aggregated_types },
		        $options,
		        $callback);

  if ($options->{automerge}) {
    warn "aggregating...\n" if $self->debug;
    foreach my $a (@aggregators) {  # last aggregator gets first shot
      warn "Aggregator $a:\n" if $self->debug;
      $a->aggregate($features,$self);
    }
  }

  @$features;

}

sub _multiple_return_args {

  my $self = shift;
  my @args = @_;
  if (@args == 0) {
    return;
  } elsif (@args == 1) {
    return $args[0];
  } elsif (wantarray) { # more than one reference sequence
    return @args;
  } else {
    $self->error($args[0]->name,
		 " has more than one reference sequence in database.  Please call in a list context to retrieve them all.");
    $self->throw('multiple segment exception');
    return;
  }

}

# backward compatibility -- don't use!
# (deliberately undocumented too)

}

sub _split_gff2_group {

  my $self = shift;
  my @groups = @_;

  my ($gclass,$gname,$tstart,$tstop,@attributes);

  for (@groups) {

    my ($tag,$value) = /^(\S+)(?:\s+(.+))?/;
    $value ||= '';
    if ($value =~ /^\"(.+)\"$/) {  #remove quotes
      $value = $1;
    }
    $value =~ s/\\t/\t/g;
    $value =~ s/\\r/\r/g;

    # Any additional groups become part of the attributes hash
    # For historical reasons, the tag "Note" is treated as an
    # attribute, even if it is the only group.
    $tag ||= '';
    if ($tag eq 'Note' or ($gclass && $gname)) {
      push @attributes,[$tag => $value];
    }

    # if the tag eq 'Target' then the class name is embedded in the ID
    # (the GFF format is obviously screwed up here)
    elsif ($tag eq 'Target' && /([^:\"\s]+):([^\"\s]+)/) {
      ($gclass,$gname) = ($1,$2);
      ($tstart,$tstop) = / (\d+) (\d+)/;
    }

    elsif (!$value) {
      push @attributes,[Note => $tag];  # e.g. "Confirmed_by_EST"
    }

    # otherwise, the tag and value correspond to the
    # group class and name
    else {
      ($gclass,$gname) = ($tag,$value);
    }
  }

  return ($gclass,$gname,$tstart,$tstop,\@attributes);

}

sub _split_gff3_group {

  my $self   = shift;
  my @groups = @_;
  my ($gclass,$gname,$tstart,$tstop,@attributes);

  for my $group (@groups) {
    my ($tag,$value) = split /=/,$group;
    $tag             = unescape($tag);
    my @values       = map {unescape($_)} split /,/,$value;
    if ($tag eq 'Parent') {
      $gclass = 'Sequence';
      $gname  = shift @values;
    }
    elsif ($tag eq 'ID') {
      $gclass = 'Sequence';
      $gname  = shift @values;
    }
    elsif ($tag eq 'Target') {
      $gclass = 'Sequence';
      ($gname,$tstart,$tstop) = split /\s+/,shift @values;
    }
    push @attributes,[$tag=>$_] foreach @values;
  }
  return ($gclass,$gname,$tstart,$tstop,\@attributes);

}

sub abs_segment {

  my $self = shift;
  return $self->segment($self->setup_segment_args(@_),-absolute=>1);

}

sub abscoords {

  my $self = shift;
  my ($name,$class,$refseq) = @_;
  $class ||= $self->{default_class};
  $self->get_abscoords($name,$class,$refseq);

}

sub absolute {

  my $self = shift;
  my $d = $self->{absolute};
  $self->{absolute} = shift if @_;
  $d;

}

sub add_aggregator {

  my $self       = shift;
  my $aggregator = shift;
  my $list = $self->{aggregators} ||= [];
  if (ref $aggregator) { # an object
    @$list = grep {$_->get_method ne $aggregator->get_method} @$list;
    push @$list,$aggregator;
  }

  elsif ($aggregator =~ /^(\w+)\{([^\/\}]+)\/?(.*)\}$/) {
    my($agg_name,$subparts,$mainpart) = ($1,$2,$3);
    my @subparts = split /,\s*/,$subparts;
    my @args = (-method      => $agg_name,
		-sub_parts   =>\@ subparts);
    push @args,(-main_method => $mainpart) if $mainpart;
    warn "making an aggregator with (@args), subparts = @subparts" if $self->debug;
    push @$list,Bio::DB::GFF::Aggregator->new(@args);
  }

  else {
    my $class = "Bio::DB::GFF::Aggregator::\L${aggregator}\E";
    eval "require $class";
    $self->throw("Unable to load $aggregator aggregator: $@") if $@;
    push @$list,$class->new();
  }

}

sub aggregators {

  my $self = shift;
  my $d = $self->{aggregators};
  if (@_) {
    $self->clear_aggregators;
    $self->add_aggregator($_) foreach @_;
  }
  return unless $d;
  return @$d;

}

sub all_seqfeatures {

  my $self = shift;
  my ($automerge,$iterator)= rearrange([
					[qw(MERGE AUTOMERGE)],
					'ITERATOR'
				       ],@_);
  my @args;
  push @args,(-merge=>$automerge)   if defined $automerge;
  push @args,(-iterator=>$iterator) if defined $iterator;
  $self->features(@args);

}

sub attributes {

  my $self = shift;
  my ($id,$tag) = @_;
  my @result = $self->do_attributes($id,$tag) or return;
  return @result if wantarray;

  # what to do in an array context
  return $result[0] if $tag;
  my %result;
  while (my($key,$value) = splice(@result,0,2)) {
     push @{$result{$key}},$value;
  }
  return\% result;

}

sub automerge {

  my $self = shift;
  my $g = $self->{automerge};
  $self->{automerge} = shift if @_;
  $g;

}

sub classes {

  my $self = shift;
  return ();

}

sub clear_aggregators {

 shift->{aggregators} = []

}

sub contained_features {

  my $self = shift;
  $self->features_in_range(-range_type=>'contains',@_);

}

sub contained_in {

  my $self = shift;
  $self->features_in_range(-range_type=>'contained_in',@_);

}

sub debug {

  my $self = shift;
  my $g = $self->{debug};
  $self->{debug} = shift if @_;
  $g;

}

sub default_aggregators {

  my $self = shift;
  return ['processed_transcript','alignment'];

}

sub default_class {

   my $self = shift;
   my $d = exists($self->{default_class}) ? $self->{default_class} : 'Sequence';
   $self->{default_class} = shift if @_;
   $d;
}

# gets name of the reference sequence, and returns its class
# currently just calls default_class

}

sub default_meta_values {

  my $self = shift;
  return ();

}

sub delete {

  my $self = shift;
  my @args = $self->setup_segment_args(@_);
  my ($name,$class,$start,$end,$offset,$length,$type,$force,$range_type) =
    rearrange([['NAME','REF'],'CLASS','START',[qw(END STOP)],'OFFSET',
	       'LENGTH',[qw(TYPE TYPES)],'FORCE','RANGE_TYPE'],@args);
  $offset = 0 unless defined $offset;
  $start = $offset+1 unless defined $start;
  $end   = $start+$length-1 if !defined $end and $length;
  $class ||= $self->default_class;

  my $types = $self->parse_types($type);  # parse out list of types

  $range_type ||= 'overlaps';
  $self->throw("range type must be one of {".
	       join(',',keys %valid_range_types).
	       "}\n")
    unless $valid_range_types{lc $range_type};


  my @segments;
  if (defined $name && $name ne '') {
    my @args = (-name=>$name,-class=>$class);
    push @args,(-start=>$start) if defined $start;
    push @args,(-end  =>$end)   if defined $end;
    @segments = $self->segment(@args);
    return unless @segments;
  }
  $self->_delete({segments   =>\@ segments,
		  types      => $types,
		  range_type => $range_type,
		  force      => $force}
		);

}

sub delete_features {

  my $self = shift;
  my @features_or_ids = @_;
  my @ids = map {UNIVERSAL::isa($_,'Bio::DB::GFF::Feature') ? $_->id : $_} @features_or_ids;
  return unless @ids;
  $self->_delete_features(@ids);

}

sub delete_groups {

  my $self = shift;
  my @features_or_ids = @_;
  my @ids = map {UNIVERSAL::isa($_,'Bio::DB::GFF::Feature') ? $_->group_id : $_} @features_or_ids;
  return unless @ids;
  $self->_delete_groups(@ids);

}

sub dna {

  my $self = shift;
  my ($id,$start,$stop,$class)  = rearrange([
					     [qw(NAME ID REF REFSEQ)],
					     qw(START),
					     [qw(STOP END)],
    					    'CLASS',
					   ],@_);
# return unless defined $start && defined $stop;
  $self->get_dna($id,$start,$stop,$class);

}

sub dna_chunk_size {

  return 0;

}

sub do_attributes {

  my $self = shift;
  my ($id,$tag) = @_;
  return ();

}

sub do_initialize {

    shift->throw('do_initialize(): must be implemented by an adaptor');

}

sub do_load_gff {

  my $self      = shift;
  my $io_handle = shift;

  local $self->{gff3_flag} = 0;
  $self->setup_load();

  my $fasta_sequence_id;

  while (<$io_handle>) {
    chomp;
    $self->{gff3_flag}++ if /^\#\#gff-version\s+3/;
    if (/^>(\S+)/) {  # uh oh, sequence coming
      $fasta_sequence_id = $1;
      last;
    }
    if (/^\#\#\s*sequence-region\s+(\S+)\s+(\d+)\s+(\d+)/i) { # header line
      $self->load_gff_line(
			   {
			    ref    => $1,
			    class  => 'Sequence',
			    source => 'reference',
			    method => 'Component',
			    start  => $2,
			    stop   => $3,
			    score  => undef,
			    strand => undef,
			    phase  => undef,
			    gclass => 'Sequence',
			    gname  => $1,
			    tstart => undef,
			    tstop  => undef,
			    attributes  => [],
			   }
			  );
      next;
    }

    next if /^\#/;
    my ($ref,$source,$method,$start,$stop,$score,$strand,$phase,$group) = split "\t";
    next unless defined($ref) && defined($method) && defined($start) && defined($stop);
    foreach (\$score,\$strand,\$phase) {
      undef $$_ if $$_ eq '.';
    }

    my ($gclass,$gname,$tstart,$tstop,$attributes) = $self->split_group($group,$self->{gff3_flag});

    # no standard way in the GFF file to denote the class of the reference sequence -- drat!
    # so we invoke the factory to do it
    my $class = $self->refclass($ref);

    # call subclass to do the dirty work
    if ($start > $stop) {
      ($start,$stop) = ($stop,$start);
      if ($strand eq '+') {
	$strand = '-';
      } elsif ($strand eq '-') {
	$strand = '+';
      }
    }
    $self->load_gff_line({ref    => $ref,
			  class  => $class,
			  source => $source,
			  method => $method,
			  start  => $start,
			  stop   => $stop,
			  score  => $score,
			  strand => $strand,
			  phase  => $phase,
			  gclass => $gclass,
			  gname  => $gname,
			  tstart => $tstart,
			  tstop  => $tstop,
			  attributes  => $attributes}
			);
  }

  my $result = $self->finish_load();
  $result += $self->load_sequence($io_handle,$fasta_sequence_id) 
    if defined $fasta_sequence_id;
  $result;

}

sub error {

  my $self = shift;
  my $g = $self->{error};
  $self->{error} = join '',@_ if @_;
  $g;

}

sub fast_queries {

  my $self = shift;
  my $d = $self->{fast_queries};
  $self->{fast_queries} = shift if @_;
  $d;

}

sub features {

  my $self = shift;
  my ($types,$automerge,$sparse,$iterator,$other);
  if (defined $_[0] && 
      $_[0] =~ /^-/) {
    ($types,$automerge,$sparse,$iterator,$other) = rearrange([
							      [qw(TYPE TYPES)],
							      [qw(MERGE AUTOMERGE)],
							      [qw(RARE SPARSE)],
							      'ITERATOR'
							     ],@_);
  } else {
    $types =\@ _;
  }

  # for whole database retrievals, we probably don't want to automerge!
  $automerge = $self->automerge unless defined $automerge;
  $other ||= {};
  $self->_features({
		    rangetype => 'contains',
		    types     => $types,
		   },
		   { sparse    => $sparse,
		     automerge => $automerge,
		     iterator  =>$iterator,
		     %$other,
		   }
		   );

}

sub features_in_range {

  my $self = shift;
  my ($range_type,$refseq,$class,$start,$stop,$types,$parent,$sparse,$automerge,$iterator,$other) =
    rearrange([
	       [qw(RANGE_TYPE)],
	       [qw(REF REFSEQ)],
	       qw(CLASS),
	       qw(START),
	       [qw(STOP END)],
	       [qw(TYPE TYPES)],
	       qw(PARENT),
	       [qw(RARE SPARSE)],
	       [qw(MERGE AUTOMERGE)],
	       'ITERATOR'
	      ],@_);
  $other ||= {};
  $automerge = $types && $self->automerge unless defined $automerge;
  $self->throw("range type must be one of {".
	       join(',',keys %valid_range_types).
	       "}\n")
    unless $valid_range_types{lc $range_type};
  $self->_features({
		    rangetype => lc $range_type,
		    refseq    => $refseq,
		    refclass  => $class,
		    start     => $start,
		    stop      => $stop,
		    types     => $types },
		   {
		    sparse    => $sparse,
		    automerge => $automerge,
		    iterator  => $iterator,
		    %$other,
		   },
		   $parent);

}

sub finish_load {

  # default, do nothing

}

sub get_Seq_by_accession {

  my $self = shift;
  my $id = shift;
  my $stream = $self->get_Stream_by_accession($id);
  return $stream->next_seq;

}

sub get_Seq_by_id {

  my $self = shift;
  my $id = shift;
  my $stream = $self->get_Stream_by_id($id);
  return $stream->next_seq;

}

sub get_Stream_by_group {

  my $self = shift;
  my @ids  = @_;
  my $id = ref($ids[0]) ? $ids[0] :\@ ids;
  Bio::DB::GFF::ID_Iterator->new($self,$id,'group');

}

sub get_Stream_by_id {

  my $self = shift;
  my @ids  = @_;
  my $id = ref($ids[0]) ? $ids[0] :\@ ids;
  Bio::DB::GFF::ID_Iterator->new($self,$id,'feature');

}

sub get_Stream_by_name {

  my $self = shift;
  my @ids  = @_;
  my $id = ref($ids[0]) ? $ids[0] :\@ ids;
  Bio::DB::GFF::ID_Iterator->new($self,$id,'name');

}

sub get_abscoords {

  my $self = shift;
  my ($name,$class,$refseq) = @_;
  $self->throw("get_abscoords() must be implemented by an adaptor");

}

sub get_dna {

  my $self = shift;
  my ($id,$start,$stop,$class,) = @_;
  $self->throw("get_dna() must be implemented by an adaptor");

}

sub get_feature_by_attribute {

  my $self = shift;
  my %attributes = ref($_[0]) ? %{$_[0]} : @_;

  # we need to refactor this... It's repeated code (see above)...
  my @aggregators;
  if ($self->automerge) {
    for my $a ($self->aggregators) {
      unshift @aggregators,$a if $a->disaggregate([],$self);
    }
  }

  my %groups;         # cache the groups we create to avoid consuming too much unecessary memory
  my $features = [];
  my $callback = sub { push @$features,$self->make_feature(undef,\%groups,@_) };
  $self->_feature_by_attribute(\%attributes,$callback);

  warn "aggregating...\n" if $self->debug;
  foreach my $a (@aggregators) {  # last aggregator gets first shot
      $a->aggregate($features,$self) or next;
  }

  @$features;
}

# more indecision...
*fetch_feature_by_attribute =\& get_feature_by_attribute;

}

sub get_feature_by_gid {

  my $self = shift;
  my $id   = ref($_[0]) eq 'ARRAY' ? $_[0] :\@ _;
  my %groups;         # cache the groups we create to avoid consuming too much unecessary memory
  my $features = [];
  my $callback = sub { push @$features,$self->make_feature(undef,\%groups,@_) };
  $self->_feature_by_id($id,'group',$callback);
  return wantarray ? @$features : $features->[0];
}
*fetch_feature_by_gid =\& get_feature_by_gid;

}

sub get_feature_by_id {

  my $self = shift;
  my $id   = ref($_[0]) eq 'ARRAY' ? $_[0] :\@ _;
  my %groups;         # cache the groups we create to avoid consuming too much unecessary memory
  my $features = [];
  my $callback = sub { push @$features,$self->make_feature(undef,\%groups,@_) };
  $self->_feature_by_id($id,'feature',$callback);
  return wantarray ? @$features : $features->[0];
}
*fetch_feature_by_id =\& get_feature_by_id;

}

sub get_feature_by_name {

  my $self = shift;
  my ($gclass,$gname,$automerge,$ref,$start,$end);
  if (@_ == 1) {
    $gclass = $self->default_class;
    $gname  = shift;
  } else  {
    ($gclass,$gname,$automerge,$ref,$start,$end) = rearrange(['CLASS','NAME','AUTOMERGE',
							      ['REF','REFSEQ'],
							      'START',['STOP','END']
							     ],@_);
    $gclass ||= $self->default_class;
  }
  $automerge = $self->automerge unless defined $automerge;

  # we need to refactor this... It's repeated code (see below)...
  my @aggregators;
  if ($automerge) {
    for my $a ($self->aggregators) {
      push @aggregators,$a if $a->disaggregate([],$self);
    }
  }

  my %groups;         # cache the groups we create to avoid consuming too much unecessary memory
  my $features = [];
  my $callback = sub { push @$features,$self->make_feature(undef,\%groups,@_) };
  my $location = [$ref,$start,$end] if defined $ref;
  $self->_feature_by_name($gclass,$gname,$location,$callback);

  warn "aggregating...\n" if $self->debug;
  foreach my $a (@aggregators) {  # last aggregator gets first shot
      $a->aggregate($features,$self) or next;
  }

  @$features;
}

# horrible indecision regarding proper names!
*fetch_group   = *fetch_feature = *fetch_feature_by_name =\& get_feature_by_name;
*segments      =\& segment;

}

sub get_feature_by_target {

  shift->get_feature_by_name(@_);

}

sub get_features {

  my $self = shift;
  my ($search,$options,$callback) = @_;
  $self->throw("get_features() must be implemented by an adaptor");

}

sub get_features_iterator {

  my $self = shift;
  my ($search,$options,$callback) = @_;
  $self->throw('feature iteration is not implemented in this adaptor');

}

sub get_seq_stream {

  my $self = shift;
  my @args = !defined($_[0]) || $_[0] =~ /^-/ ? (@_,-iterator=>1)
                                              : (-types=>\@_,-iterator=>1);
  $self->features(@args);
}

*get_feature_stream =\& get_seq_stream;

}

sub get_types {

  my $self = shift;
  my ($refseq,$class,$start,$stop,$count,$types) = @_;
  $self->throw("get_types() must be implemented by an adaptor");

}

sub initialize {

  my $self = shift;
  #$self->do_initialize(1) if @_ == 1 && $_[0];
  #why was this line (^) here?  I can't see that it actually does anything
  #one option would be to execute the line and return, but I don't know
  #why you would want to do that either.
 
  my ($erase,$meta) = rearrange(['ERASE'],@_);
  $meta ||= {};

  # initialize (possibly erasing)
  return unless $self->do_initialize($erase);
  my @default = $self->default_meta_values;

  # this is an awkward way of uppercasing the 
  # even-numbered values (necessary for case-insensitive SQL databases)
  for (my $i=0; $i<@default; $i++) {
    $default[$i] = uc $default[$i] if !($i % 2);
  }

  my %values = (@default,%$meta);
  foreach (keys %values) {
    $self->meta($_ => $values{$_});
  }
  1;

}

sub insert_sequence {

  my $self = shift;
  my($id,$offset,$seq) = @_;
  $self->throw('insert_sequence(): must be defined in subclass');
}

# This is the default class for reference points.  Defaults to Sequence.

}

sub insert_sequence_chunk {

  my $self = shift;
  my ($id,$offsetp,$seqp) = @_;
  if (my $cs = $self->dna_chunk_size) {
    while (length($$seqp) >= $cs) {
      my $chunk = substr($$seqp,0,$cs);
      $self->insert_sequence($id,$$offsetp,$chunk);
      $$offsetp += length($chunk);
      substr($$seqp,0,$cs) = '';
    }
  }
  return 1;  # the calling routine may expect success or failure
}

# used to store big pieces of DNA in itty bitty pieces

}

sub load_fasta {

  my $self              = shift;
  my $file_or_directory = shift || '.';
  return $self->load_sequence($file_or_directory) if ref($file_or_directory) &&
                                                     tied *$file_or_directory;

  my $tied = tied(*STDIN);
  open SAVEIN,"<&STDIN" unless $tied;
  local @ARGV = $self->setup_argv($file_or_directory,'fa','dna','fasta') or return;  # to play tricks with reader
  my $result = $self->load_sequence('ARGV');
  open STDIN,"<&SAVEIN" unless $tied;  # restore STDIN
  return $result;

}

sub load_gff {

  my $self              = shift;
  my $file_or_directory = shift || '.';
  return $self->do_load_gff($file_or_directory) if ref($file_or_directory) &&
                                                   tied *$file_or_directory;

  my $tied_stdin = tied(*STDIN);
  open SAVEIN,"<&STDIN" unless $tied_stdin;
  local @ARGV = $self->setup_argv($file_or_directory,'gff') or return;  # to play tricks with reader
  my $result = $self->do_load_gff('ARGV');
  open STDIN,"<&SAVEIN" unless $tied_stdin;  # restore STDIN
  return $result;
}

*load =\& load_gff;

}

sub load_gff_line {

  shift->throw("load_gff_line(): must be implemented by an adaptor");

}

sub load_sequence {

  my $self = shift;
  my $io_handle = shift;
  my $id        = shift;   # hack for GFF files that contain fasta data

  # read fasta file(s) from ARGV
  my ($seq,$offset,$loaded) = (undef,0,0);
  while (<$io_handle>) {
    chomp;
    if (/^>(\S+)/) {
      $self->insert_sequence($id,$offset,$seq) if $id;
      $id     = $1;
      $offset = 0;
      $seq    = '';
      $loaded++;
    } else {
      $seq .= $_;
      $self->insert_sequence_chunk($id,\$offset,\$seq);
    }
  }
  $self->insert_sequence($id,$offset,$seq) if $id;
  $loaded+0;

}

sub load_sequence_string {

  my $self = shift;
  my ($acc,$seq)  = @_;
  my $offset = 0;
  $self->insert_sequence_chunk($acc,\$offset,\$seq) or return;
  $self->insert_sequence($acc,$offset,$seq) or return;
  1;

}

sub lock_on_load {

  my $self = shift;
  my $d = $self->{lock};
  $self->{lock} = shift if @_;
  $d;

}

sub make_aggregated_feature {

  my $self                 = shift;
  my ($accumulated_features,$parent,$aggregators) = splice(@_,0,3);
  my $feature = $self->make_feature($parent,undef,@_);
  return [$feature] if $feature && !$feature->group;

  # if we have accumulated features and either: 
  # (1) make_feature() returned undef, indicated very end or
  # (2) the current group is different from the previous one

  local $^W = 0;  # irritating uninitialized value warning in next statement
  if (@$accumulated_features &&
      (!defined($feature) || ($accumulated_features->[-1]->group ne $feature->group))) {
    foreach my $a (@$aggregators) {  # last aggregator gets first shot
      $a->aggregate($accumulated_features,$self) or next;
    }
    my @result = @$accumulated_features;
    @$accumulated_features = $feature ? ($feature) : ();
    return unless @result;
    return\@ result ;
  }
  push @$accumulated_features,$feature;
  return;

}

sub make_feature {

  my $self = shift;
  my ($parent,$group_hash,          # these arguments provided by generic mechanisms
      $srcseq,                      # the rest is provided by adaptor
      $start,$stop,
      $source,$method,
      $score,$strand,$phase,
      $group_class,$group_name,
      $tstart,$tstop,
      $db_id,$group_id) = @_;

  return unless $srcseq;            # return undef if called with no arguments.  This behavior is used for
                                    # on-the-fly aggregation.

  my $group;  # undefined
  if (defined $group_class && defined $group_name) {
    $tstart ||= '';
    $tstop  ||= '';
    if ($group_hash) {
      $group = $group_hash->{$group_class,$group_name,$tstart,$tstop}
	||= $self->make_object($group_class,$group_name,$tstart,$tstop);
    } else {
      $group = $self->make_object($group_class,$group_name,$tstart,$tstop);
    }
  }

# fix for some broken GFF files
# unfortunately - has undesired side effects
#  if (defined $tstart && defined $tstop && !defined $strand) {
#    $strand = $tstart <= $tstop ? '+' : '-';
#  }

  if (ref $parent) { # note that the src sequence is ignored
    return Bio::DB::GFF::Feature->new_from_parent($parent,$start,$stop,
						  $method,$source,
						  $score,$strand,$phase,
						  $group,$db_id,$group_id,
						  $tstart,$tstop);
  } else {
    return Bio::DB::GFF::Feature->new($self,$srcseq,
				      $start,$stop,
				      $method,$source,
				      $score,$strand,$phase,
				      $group,$db_id,$group_id,
				      $tstart,$tstop);
  }

}

sub make_match_sub {

  my $self = shift;
  my $types = shift;

  return sub { 1 } unless ref $types && @$types;

  my @expr;
  for my $type (@$types) {
    my ($method,$source) = @$type;
    $method ||= '.*';
    $source  = $source ? ":$source" : "(?::.+)?";
    push @expr,"${method}${source}";
  }
  my $expr = join '|',@expr;
  return $self->{match_subs}{$expr} if $self->{match_subs}{$expr};

  my $sub =<<END;
sub {
  my\$ feature = shift or return;
  return\$ feature->type =~ /^($expr)\$/i;
}
END
  warn "match sub: $sub\n" if $self->debug;
  my $compiled_sub = eval $sub;
  $self->throw($@) if $@;
  return $self->{match_subs}{$expr} = $compiled_sub;

}

sub make_object {

  my $self = shift;
  my ($class,$name,$start,$stop) = @_;
  return Bio::DB::GFF::Homol->new($self,$class,$name,$start,$stop)
    if defined $start and length $start;
  return Bio::DB::GFF::Featname->new($class,$name);

}

sub meta {

  my $self = shift;
  my ($name,$value) = @_;
  return;

}

sub new {

  my $package   = shift;
  my ($adaptor,$aggregators,$args,$refclass);

  if (@_ == 1) {  # special case, default to dbi::mysqlopt
    $adaptor = 'dbi::mysqlopt';
    $args = {DSN => shift};
  } else {
    ($adaptor,$aggregators,$refclass,$args) = rearrange([
							 [qw(ADAPTOR FACTORY)],
							 [qw(AGGREGATOR AGGREGATORS)],
							 'REFCLASS',
							],@_);
  }

  $adaptor    ||= 'dbi::mysqlopt';
  my $class = "Bio::DB::GFF::Adaptor::\L${adaptor}\E";
  eval "require $class" unless $class->can('new');
  $package->throw("Unable to load $adaptor adaptor: $@") if $@;

  my $self = $class->new($args);
  $self->default_class($refclass) if defined $refclass;

  # handle the aggregators.
  # aggregators are responsible for creating complex multi-part features
  # from the GFF "group" field.  If none are provided, then we provide a
  # list of the two used in WormBase.
  # Each aggregator can be a scalar or a ref.  In the former case
  # it is treated as a class name to call new() on.  In the latter
  # the aggreator is treated as a ready made object.
  $aggregators = $self->default_aggregators unless defined $aggregators;
  my @a = ref($aggregators) eq 'ARRAY' ? @$aggregators : $aggregators;
  for my $a (@a) {
    $self->add_aggregator($a);
  }

  # default settings go here.....
  $self->automerge(1);  # set automerge to true

  $self;

}

sub overlapping_features {

  my $self = shift;
  $self->features_in_range(-range_type=>'overlaps',@_);

}

sub parse_types {

  my $self  = shift;
  return []   if !@_ or !defined($_[0]);
  return $_[0] if ref $_[0] eq 'ARRAY' && ref $_[0][0];
  my @types = ref($_[0]) ? @{$_[0]} : @_;
  my @type_list = map { [split(':',$_,2)] } @types;
  return\@ type_list;

}

sub refclass {

  my $self = shift;
  my $name = shift;
  return $self->default_class;

}

sub segment {

  my $self = shift;
  my @segments =  Bio::DB::GFF::RelSegment->new(-factory => $self,
						$self->setup_segment_args(@_));
  foreach (@segments) {
    $_->absolute(1) if $self->absolute;
  }

  $self->_multiple_return_args(@segments);

}

sub setup_argv {

  my $self = shift;
  my $file_or_directory = shift;
  my @suffixes          = @_;
  no strict 'refs';  # so that we can call fileno() on the argument

  my @argv;

  if (-d $file_or_directory) {
    @argv = map { glob("$file_or_directory/*.{$_,$_.gz,$_.Z,$_.bz2}")} @suffixes;
  }elsif (my $fd = fileno($file_or_directory)) {
    open STDIN,"<&=$fd" or $self->throw("Can't dup STDIN");
    @argv = '-';
  } elsif (ref $file_or_directory) {
    @argv = @$file_or_directory;
  } else {
    @argv = $file_or_directory;
  }

  foreach (@argv) {
    if (/\.gz$/) {
      $_ = "gunzip -c $_ |";
    } elsif (/\.Z$/) {
      $_ = "uncompress -c $_ |";
    } elsif (/\.bz2$/) {
      $_ = "bunzip2 -c $_ |";
    }
  }
  @argv;

}

sub setup_load {

  # default, do nothing

}

sub setup_segment_args {

  my $self = shift;
  return @_ if defined $_[0] && $_[0] =~ /^-/;
  return (-name=>$_[0],-start=>$_[1],-stop=>$_[2]) if @_ == 3;
  return (-class=>$_[0],-name=>$_[1])              if @_ == 2;
  return (-name=>$_[0])                            if @_ == 1;

}

sub split_group {

  my $self = shift;
  my ($group,$gff3) = @_;
  if ($gff3) {
    my @groups = split /[;&]/,$group;  # so easy!
    return $self->_split_gff3_group(@groups);
  } else {
    # handle group parsing
    # protect embedded semicolons in the group; there must be faster/more elegant way
    # to do this.
    $group =~ s/\\;/$;/g;
    while ($group =~ s/( \"[^\"]*);([^\"]*\")/$1$;$2/) { 1 }
    my @groups = split(/\s*;\s*/,$group);
    foreach (@groups) { s/$;/;/g }
    return $self->_split_gff2_group(@groups);
  }

}

sub strict_bounds_checking {

  my $self = shift;
  my $d = $self->{strict};
  $self->{strict} = shift if @_;
  $d;

}

sub types {

  my $self = shift;
  my ($refseq,$start,$stop,$enumerate,$refclass,$types) = rearrange ([
								      [qw(REF REFSEQ)],
								      qw(START),
								      [qw(STOP END)],
								      [qw(ENUMERATE COUNT)],
								      [qw(CLASS SEQCLASS)],
								      [qw(TYPE TYPES)],
								     ],@_);
  $types = $self->parse_types($types) if defined $types;
  $self->get_types($refseq,$refclass,$start,$stop,$enumerate,$types);

}