[BioC] limma: read.imagene patch

[Yves Bastide]

Gordon Smyth wrote:
> Dear Yves,
> 
> Thanks for the suggested code changes. Before I make commit changes to 
> the code, would you mind showing me an example of some headers which 
> change length? I know that different versions of Imagene produce 
> different headers, but read.imagene() like read.maimages() is designed 
> to read a batch of arrays which should in principle have the same 
> layout, so I would be uncomfortable about users reading in unrelated 
> arrays. I just want to see in what ways the header can change within one 
> experiment.

You're right, this patch is way too broad; I don't think the headers for 
the cy3 and cy5 files of a measure can change.

> 
> Note also that if you have two batches of arrays scanned with different 
> settings, you can read them into two RGList objects and then combine the 
> objects using cbind.

This is the case I met, with slightly different quality settings between 
the first and subsequent scans.  I think it'd be convenient having 
read.imagene() handle this.

Sketch:

for (i in 1:narrays) {
   fullname <- ...
+  headers <- getImageneHeaders(fullname)
+  if (headers$Field.Dimensions[...] != printer) error(...)
+  skip <- headers$Begin.Raw.Data
   obj<-read.table(...)

> 
> Cheers
> Gordon
> 

Regards,

yves