We did this recently in minfi and getting acceptable performance was not
trivial. We are still sorting out issues; we will eventually write about
our experiences. Pete is working on some blog posts on this and we hope to
do something more formal later.
We would all love experience from more
Thanks Hervé and Stephanie for your suggestions. I am really looking for a S4
methods solution however, given how my package is already set up. Also, I have
several functions that I need to adapt in this way, so it seems cleaner and
simpler to do the class union, which sounds like is not a
Interesting. I tried something like that in the past i.e. start with
a unary setClassUnion() but then got into problems when I tried to add
new members to the union by **extending** the union class:
https://stat.ethz.ch/pipermail/r-devel/2016-March/072489.html
So it seems like I should have
It would be great to be able to define a matrix-like abstraction
independent of 'matrix' and 'DelayedMatrix'. It could also encompass
objects from the Matrix package and potentially other things. So you
could define a parent class of 'matrix' using setClassUnion() and then
use setIs() to establish
The class union should probably be:
setClassUnion("matrixOrDelayed", c("matrix", "DelayedMatrix"))
i.e. use DelayedMatrix instead of DelayedArray.
So in addition to the class union and to Stephanie's solution, which
IMO are both valid solutions, you could also go for something like this:
Just to mention that the issue with rowSums() on a big DelayedMatrix
objects that you are referring to Tim is duly noted and will be one
of my first priorities once we're done with the release process.
Cheers,
H.
On 04/30/2018 09:57 AM, Tim Triche, Jr. wrote:
much obliged -- and the packages
much obliged -- and the packages are terrific, I am not surprised that a
big step is accompanied with some growing pains.
Thanks to you and Herve and Keegan for enthusiastically chasing down, and
spending your time fixing, this and other bugs.
Having fiddled with bigMemory and bigMatrix backends
Rather than a class union, how about an internal function that is called
by the methods for both matrix and DelayedArray:
setGeneric("myNewRowMeans", function(x,...) {
standardGeneric("myNewRowMeans")})
#' @importFrom DelayedArray rowMeans
.myNewRowMeans <- function(x,...){
# a lot of
Tim: As the developer of DelayedMatrixStats (and enthusiastic 'canary down
the coal mine' user-dev of DelayedArray) I'm obviously invested in reducing
the confusion around these packages
I'm going to write some blog posts-cum-vignettes-cum-F1000 around these
issues over the coming weeks, with the
But if you merge methods like that, the error method can be that much more
difficult to identify. It took a couple of weeks to chase that bug down
properly, and it ended up down to rowMeans2 vs rowMeans.
I suppose the merged/abstracted method allows to centralize any such
dispatch into one place
But that issue will be fixed, so Tim's advice is inappropriate.
On 04/30/2018 10:42 AM, Tim Triche, Jr. wrote:
Don't do that. Seriously, just don't.
https://github.com/Bioconductor/DelayedArray/issues/16
--t
On Mon, Apr 30, 2018 at 10:02 AM, Elizabeth Purdom <
epur...@stat.berkeley.edu>
Don't do that. Seriously, just don't.
https://github.com/Bioconductor/DelayedArray/issues/16
--t
On Mon, Apr 30, 2018 at 10:02 AM, Elizabeth Purdom <
epur...@stat.berkeley.edu> wrote:
> Hello,
>
> I am trying to extend my package to handle `HDF5Matrix` class ( or more
> generally
Hello,
I am trying to extend my package to handle `HDF5Matrix` class ( or more
generally `DelayedArray`). I currently have S4 functions for `matrix` class.
Usually I have a method for `SummarizedExperiment`, which will call call the
method on `assay(x)` and I want the method to be able to deal
13 matches
Mail list logo
