[emboss-dev] Mapping feature types to Sequence Ontology (SO)

[Peter Cock]

On Wed, Aug 17, 2011 at 4:55 PM, Peter Rice <pmr at ebi.ac.uk> wrote:
> On 17/08/2011 16:48, Peter Cock wrote:
>> See also the discussion today on gmod-gbrowse / song-devel where
>> it sounds like GFF3 should have a single block of FASTA embedded
>> sequence at the end of the fine, rather than interleaved. As I suggest
>> on that thread, the practical solution for EMBOSS seqret might be to
>> omit the FASTA sequence altogether. Or cache them in memory/on
>> disk to write out at the very end of the all the features?
>
> Thanks. We already save sequences and write at the end for some
> formats so I'll add it for GFF3. We will need more work for reading
> GFF3 input though, but it may not be too bad.
>
> If we are reading it as feature input, we don't look for the sequence.
>
> If we are reading as sequence input, we need to read all the sequeces
> into memory and then go back to read the features. For streamed input
> we can buffer to make the rewind work.

I'm curious what other file formats needed this kind of work. But it
is good that you've already got some buffer/cache infrastructure
in place. Does it boil down to writing temp files in /tmp ?

Peter C.