[BioC] limma reading genpix files

[Gordon Smyth]

Dear Dave,

Thanks for the diagnosis. You are quite correct that read.maimages() does 
assume that all the gpr files in a batch have headers of the same number of 
lines, and this assumption should be relaxed. As it happens, the assumption 
has always been true for genepix data sets that I have seen.

Gordon

At 03:53 AM 14/04/2004, David Nelson wrote:
>Hello all,
>
>I've just started plowing thru some two-color arrays and ran across the 
>same bug.
>
>The problem I found is due to an efficiency hack that doesn't always work: 
>read.maimages assumes that all the genepix files it reads have the same 
>number of header records. That is, rather than taking the number of header 
>records directly from each .gpr file, it counts the number of lines until 
>the header line in the *first file only*, and then assumes that all the 
>rest of the files have the same number of records.
>
>If not, then the skip= parameter is incorrect, read.table() starts in the 
>wrong place, and results are, as they say, unpredictable.
>
>The "right" way to do this is to read the first lines of each .gpr file, 
>get the number of header records, and then use read.table with the right 
>number of header records. A quick hack is just to search for the header in 
>each file.
>
>I've attached a modification to read.maimages() that does exactly that, at 
>least the three times I've tried it :-)
>
>
>Cheers,
>
>Dave Nelson