[Bioperl-pipeline] some transformer stuff

[Elia Stupka]

> 	my head is cloudy with jet lag but I had playing around with 
> transformers
> on the plane and had some small proposals. I am doing some multiple 
> alignments
> for phylogenetic tree building.

I wonder if Shawn ever takes a break ;)

> So to read an alignment file I'm using AlignIO. We all know that 
> IOHandlers do not
> handle iterator type calls. So what I know have is another transfomer 
> that sits
> between the iohandler and the thre alignment editor transformer.

This is excellent, as we get more and more people out there saying 
"Iterators are cool" while in BioPipe we can't stand them... I am sure 
we will end up reusing it quite often. I think it will also be 
applicable to event-driven parsers by iterating over an event.

> So that works. The issue is whether we want to store full file paths 
> in the input table
> for each input name, which is messy if you are copying file inputs and 
> you want to figure out
> the base file name. Alternatively, what I think might be better is to 
> add a column file_dir in the stream
> adaptor table. So if its present, and the inputs are files, it will 
> concatenate it.

Seems like a very sensible idea. Again, it will probably help for all 
file-based pipelines. For example, if you want to extend the file-based 
blast pipeline by adding a downstream analysis or transformer you'd 
have to specify the dir and run through the files. Also one could 
actually just use the file_dir to let the next module figure out what 
it has to do (i.e. list files in directory and do what you are meant to 
do with them). Thumbs up :)

I am hopefully emerging from tons of annoying burocracy and will be 
playing with biopipe, biosql, promoter analysis,etc. a lot more from 
when I move to TLL.... now writing in a cab off to SIR to get a new EP 
hopefully....

Elia