methods signature

[lgatto]

As an example, the peaksCount generic is defined in mzR as

setGeneric("peaksCount", function(object,scans,...) standardGeneric("peaksCount"))

I would suggest to only implement methods for object = "OnDiskMSnExp", scans = "missing" (and ignore scans = "numeric"). If only some peaks count needs to be computer, it seems more consistent to subset prior to calling the methods, like

peaksCount(x[scans])

[lgatto]

Done in 08385ed.

[lgatto]

NB: I will keep the 2 in-memory methods for now, as the one with numeric scans should be deprecated prior to defunct. Will do that later.

[lgatto]

And by the way, the performance is hardly different:

> f <- proteomics(full = TRUE)[1]
> suppressMessages(xx <- MSnbase:::readMSData2(f))
> microbenchmark(suppressMessages(spectra(xx[1:2])), 
+                suppressMessages(spectra(xx, scans = 1:2)))
Unit: milliseconds
                                       expr      min       lq     mean   median
         suppressMessages(spectra(xx[1:2])) 187.1789 191.2811 200.8911 197.4171
 suppressMessages(spectra(xx, scans = 1:2)) 183.9336 186.9881 198.0007 192.6416
       uq      max neval
 203.5280 250.7263   100
 199.1815 256.3811   100

[jorainer]

I just added some additional functionality to peaksCount that prevents the full data to be imported and peak counts being calculated if the data was not processed (returning the originalPeaksCount in that case).
Pushed in 617c477

[lgatto]

Ok, thanks.

Regarding you comment here

## @lgatto: I suppose we will do the subsetting all on the object itself, thus
## we don't need the subsetting code below, right?
## Sub-set the feature data based on the arguments.
+    fd <- .subsetFeatureDataBy(fData(object), index=index, scanIdx=scanIdx,
+                               name=name, rtlim=rtlim)

Indeed, I think we can remove this. Will do after pulling/merging.