[Bioperl-l] *major* error in genbank parser or am i just insane?

[Chris Mungall]

bear in mind the situation I'm talking about is not for weirdo isolated
transplicing cases or 20 year old records entered by some crazed lunatic -
this bug happens if you go to ncbi download the human assembly build29 and
parse it: half of the mRNAs are wrong, unless i'm doing something
fundamentally wrong

we need a good long term solution that is robust for all of genbank, but
we need a short term fix for the standard situation even more - shall i
commit my chnange or will this mess things up more?

all it is

$location->strand($strand)

in FTHelper.pm

On Wed, 7 Aug 2002, Elia Stupka wrote:

> > I doubt that the cross-product of location types and genbank entries
> > has been tested in its entirety, so something may have easily
> > escaped.
>
> Definitely not. Long time ago when I had written the Genbank parser I had
> done a few file-parser-file cycles to see that the files spit out would be
> identical.
>
> As we all know public databases will never cease to amaze us, so the only
> way to be bug-free on this would be to parse all of Genbank on a regular
> basis, spit it out again in Genbank format and write a log of the diff and
> look into that diff.
>
> Sorry for dropping the standard of the e-mail conversation from grammar
> discussion to this idiotic paranoyed level of checking, but it is the
> only way to do it.
>
> Since we keep public DBs in biosql and we will need more and more public
> dbs in biosql, I can easily put a hook in to spit out the files again and
> diff them, and on a monthly basis for example post the "findings" so we
> can then all go off and fix bugs...
>
> Does it sound right?
>
> By the way, I posted a fix for missing GIs, has it landed on the bioperl
> list? I assume all is fine, just checking.
>
> Elia
>
> ********************************
> * http://www.fugu-sg.org/~elia *
> * tel:    +65 6874 1467        *
> * mobile: +65 9030 7613        *
> * fax:    +65 6777 0402        *
> ********************************
>
>
>