Slight change in procedure: a 'master' database must now be specified. The external_dbs.txt file is loaded into this, and this in turn is compared (id and name) with the contents of the external_db table in each of the databases to be updated.

If the database to be updated has any extra rows, a warning is given and the database is skipped.

This should help avoid unilateral changes to external_db which haven't been commited to the external_dbs.txt file in CVS being silently overwritten and causing problems down the line.

fetch_all_by_biotype  added

Change in UNIQUE index on xref results in many more GO terms (apparently) being projected, but this is because the same GO term projected from multiple sources now gets multiple rows in the xref table, whereas before, because info_type/info_text wasn't being considered, only one row would have been created.

length as the original exon, returning undef if not (can happen when
only part of the exon could be transformed)

committing intermediate state of module before rewriting it to use lightweight feature object for caching

feature itself does not transform cleanly and the co-ordinates of the
transformed gene are inferred from the transformed transcripts/exons.

FlyBase genes which have names