the feature specific part.

implem'ted.

Do some argument checking in get_bins().

Move some code around.

and ProteinAlignFeature.  Currently gets all data from the
{dna,protein}_align_feature tables without no join to any other table.

  Flushes (deletes) all cached data not associated with the current
  slice.

Also store the projection segments and the mapping from seq_region_ids
to the projection segments in the class.

Added external_db_id and hcoverage, for the 2 new columns in dna_align_feature and protein_align_feature.

lines shorter.

Add documentation on binning methods.

Still need to implement the 'coverage' binning method.

the private methods, yet.

new(), lightweight().

one of the various binning routines.

Renamed method collection() into entries().

Carp.

The constructor now takes two named arguments:

  -light (boolean, optional), determines whether the populate() method
  should populate the collection with feature-specific data or not by
  default.

  -slice (a slice, not optional), the slice we're working on.

The is_light() method was renamed into lightweight().

The populate() method takes three named arguments (all optional):

  -sorted (boolean), determines if the results should be sorted or not.

  -sortsub (coderef), reference to subroutine that will be used in
  sorting the collection (for sorting on feature size, score, whatever,
  rather than on start position).

  -light (boolean), overrides the corresponding value given to the
  constructor at the creation of the collection.

A gene entry in a colleciton now contains:

  index contents

  5     gene.biotype
  6     gene.status
  7     gene_stable_id.stable_id
  8     xref.display_xref_id

(apart from the usual seq_region stuff in index 0-4).

Allow the user to specify whether the populate() method should sort the
feature collection or not.

Sort the seq_region_ids in multi-segment queries.

RepeatFeature adaptor uses it.  Don't seem to make a difference to
timing results though.

collection entries.  Fixes problem with binning introduced with the
previous check-in.

  If a slice, for a particular coordinate system, projects onto more
  than two segments, do not involve seq_region_start and seq_region_end
  in the SQL for the "middle slices".

The compact representation of a feature now also contains the
seq_region_id as the second array entry.

a slice fetches.

Added kill_reason and strip_UTR - two attribs used to show why transcripts were
removed/changed, for genebuilder use in databases of killed structures

  collection()

A feature collection is a compact representation of the features of a
certain type on a particular slice.

This is work in progress.