removed warning in store_alt_alleles when none of the alt_alleles in a set is on reference sequence - this is a valid case and doesn't require a warning

Assignment bug; we did not want the cache assigned to the ID of the seq region but wanted the seq region id into a local variable. Fixed.

Asking if a dereferenced array ref is undefined is not testing what you think it is and is deprecated in Perl 5.16.0. You are seeing if the first element was defined when in actual fact you needed to assert if the array reference was undefined and act accordingly.

Simplification of the select. The query for select used fields which are not part of the unique key. We also get the API to issue a throw the moment we miss the xref id retrieval. This should be an impossible scenario to get into though

translated sequence, not transcript sequence.

phase change was removed because it meant that an exon whose sequence
and position had not changed would become a candidate for similarity
mapping.  This caused lots of criss-crossing stable ID history entries
between genes/transcripts/translations in some gene families.

Better version. We now avoid gzipping which should give us a good speed boost and bring us closer to the original speed

Switching the logic around here. It makes sense if we are trying to read from a file which hasn't closed yet then maybe we're not going to get all the data into the file.

sequence.  This means that the translation stable ID version might be
incremented independently of the transcript stable ID version if there
was a phase change in an exon.