[Bioperl-l] PullHSP bug

[Chris Fields]

On Dec 3, 2008, at 5:38 AM, Sendu Bala wrote:

> Chris Fields wrote:
>> The short and easy fix is either having get_nse() or blast_pull  
>> fall back to '', but both are technically wrong as there isn't a  
>> true ID.  Ideas?  Maybe a fallback 'unknown' ID (which seems to be  
>> used as a ).

...fallback in some cases (Can tell I was getting tired).

> I'd say for now at least the fix would be to alter get_nse() to fall  
> back to ''. You don't want the work-around code duplicated in  
> multiple places. Indeed, the normal blast parser should be altered  
> to return undef as well.

That solution requires get_nse to silently convert this; the exception  
is raised b/c the N/S-E is used for indexing (so lack of any one of  
those may result in indexing issues, such as replacing sequences with  
same ID but no S-E, etc.).

> Then if a more desirable solution is thought of in the future,  
> there's just one place to implement it: in get_nse().

We don't want the exception turned off unless it is explicitly known  
the sequence instance lacks any or all of the N/S-E, such as in this  
case.  Setting an explicit LocatableSeq flag for such instances  
(force_nse), then having get_nse() DTRT internally would be easy to  
implement.

Odd, list mail hasn't been coming through (bioperl-gut either).  I'll  
post something to support.

chris