Added dumping of "orphan" xrefs that aren't associated with primary_xrefs or dependent_xrefs - this is currently just Interpro xrefs, but we dump them all just in case.

Now dumps GO linkage types to go_xref and uploads this if required, allowing display of GO xrefs & related information on website, assuming ensembl_go database is present and configured.

Added generation of gene descriptions. Previously these were done via a separate script, now they are generated as part of the post-xref-mapping phase.

New implementation uses the same precedence rules for deciding which xref description to assign to a gene as the old script.

fix bug that if -dumpcheck set and files not found then the script failed instead of dumping the files

Make all external_synonym entries unique.

Corrected location from which display_xref .sql files are uploaded.

Fixed calculation of transcript display_xrefs; wasn't considering translation-transcript relationship properly.

Now use | as separator character in composite hash keys; : was a bad choice since it appears in GO xrefs.

Don't write xref/object_xrefs for sources that have no corresponding external_db entry (but warn about this).

Set label to accession if it's not set otherwise (so that something gets displayed on the web site).

Check all object_xrefs are unique.

Still problem with gene display_xrefs.

No longer write external_db table from source table, since external_db is defined in CVS. Now builds source_idd->external_db_id mapping and writes external_db_id to xref.txt.

Added external_db and version to core xref output.

Renamed dump_xrefs to dump_core_xrefs to avoid conflict with similarly-named but differntly-functioning dump_xref