[Biopython-dev] Active projects list? (BOSC Biopython Project Update)

[Peter]

On Tue, Jul 20, 2010 at 5:07 PM, Andrea Pierleoni
<andrea at biocomp.unibo.it> wrote:
>
>>
>> I wouldn't want to ship web2py with Biopython - we'd just list it as
>> another optional package you might want to install for use with BioSQL
>> (as we do with MySQLdb etc).
>>
>
> that sounds reasonable.
>
>>
>> There are already several query methods, but more wouldn't be a bad
>> idea. I was thinking we could implement dictionary like access, and
>> support for iterator over all the records.
>>
>
> dictionary and iterators would be very pythonic, and useful. are you
> working on it?

Not right now, no - if you want to try soon please go ahead.

> correct me if I'm wrong, but the standard policy in Biopython BioSQL to
> update a bioentry record is to delete the old one and create a new one
> (or make a new version). wouldn't be useful to enable in biopython some
> minor modifications to a bioentry like adding/removing features and
> qualifiers? maybe I can help with this.

The current functionality is limited to loading and retrieving records
(and retreiving is done in a lazy or on demand way which saves
memory and DB access). As a consequence, if you want to edit a
record in the database you have to either do it directly (bypass our
BioSQL code) or load a new record.

The BioSQL schema doesn't have any sort of audit trail (unlike CHADO
if I remember correctly), so for many uses this almost read only setup
is actually a plus point. Here we use BioSQL essentially as a container
for NCBI GenBank / RefSeq dumps - although we do add additional
annotations on top.

I can see advantages in allowing the DBSeqRecord to write back
to the database - it would need a lots of refactoring through (e.g. most
of the loader code would get moved). I would start by creating a read
only proxy for the DBSeqFeature (something I think Leighton Pritchard
did in some of his code) because editing feature annotations would
be an important part of this.

Peter