[Bioperl-l] DB.t failures

[Chris Fields]

On Feb 15, 2007, at 9:44 AM, Chris Fields wrote:

>
> On Feb 15, 2007, at 8:37 AM, Sendu Bala wrote:
>
>> Chris Fields wrote:
>>>
>>> On Feb 15, 2007, at 7:37 AM, Sendu Bala wrote:
>>>
>>>> Jay Hannah wrote:
>>>>> On Feb 14, 2007, at 8:02 AM, Sendu Bala wrote:
>>>>>> DB.t is failing with BIOPERLDEBUG set. Apparently, we are no
>>>>>> longer
>>>>>> getting sequences back from NCBI in the order we requested  
>>>>>> them in
>>>>>> batch mode.
>>>
>>> Okay, I committed a fix for that.  I hope there are many users who
>>> depend on the returned sequence order for anything!
>>
>> s/are/aren't/ ?
>
> Yes, my oops.
>
>> I suspect there might be, and its certainly a reasonable  
>> assumption to
>> make. Did you not see an easy way of maintaining the order?
>
> I haven't looked (been busy the last few days), but I think there is
> a way via efetch.
>
> We could add in something to the default base URL if there is
> something or (probably better) add a sort_order() method to designate
> a particular sort order, defaulting to the old order if not set.
>
> chris

Delving in to it further, the problem only occurs when using  
get_seq_stream() directly in batch mode, which is likely only used by  
developers for testing.  The sort issue only pops up when eposting  
IDs using that mode; retrieved seqs are returned in a different order  
than through a direct efetch query (the default with get_Stream* or  
get_Seq* methods).  No use of the 'sort' parameter works to get  
around that problem, not a complete surprise since it is supposed to  
only work for PubMed, but since the method is rarely used I'll just  
leave the bullet-proofed tests alone.

chris