[Bioc-devel] HDF5Array/DelayedArray matrix operations fail on windows

[Kevin Wang]

Dear Herve,

Thank you for your help. I did not manage the replicate the error on my Windows VM. However, I did comment out the only offending line of code which used parallelisation and manually set the BiocParallel backend to SerialParam during testthat and the error was gone. Thanks a lot for your help!

Best Wishes
Kevin
The University of Sydney

On 24 Oct 2019, at 4:28 am, Pages, Herve <hpages using fredhutch.org<mailto:hpages using fredhutch.org>> wrote:

Hi Kevin,

Running this particular code (e.g. the code in the test-sparsematrix.R
file) also works for me on our Windows builders. However, trying to run
the **full suite** of tests does trigger the error. So it seems to be a
situation where some code in some other tests is altering the state of
the session in a way that breaks the `scMerge()` calls in
test-hdf5array.R and test-sparsematrix.R.

I see that you use parallelization a lot. Could it be that some code at
some point alters the state of the current parallelization backend? All
these socketConnection() warnings makes me suspect that. This would also
explain why we only see the problem on Windows. Parallelization problems
are very platform specific and tend to occur more often on Windows where
backends like MulticoreParam are not supported.

Best,
H.


On 10/21/19 23:32, Kevin Wang wrote:
Hi, I am trying to make my package scMerge to be compatible with HDF5Array. While all the checks passed on Linux and Mac, it errored on Windows (https://protect-au.mimecast.com/s/-9LqC81Zj6tK3KPLtnsVJT?domain=urldefense.proofpoint.com ). Upon closer inspection, the failures seem to come a test for compatibility of HDF5Array
(the script can be viewed here https://protect-au.mimecast.com/s/VjnLC91ZkQtpnpN9FEm10J?domain=urldefense.proofpoint.com ). The error message was "non-numeric argument to binary operator” and calls to DelayedArray. I have tried to run this specific line on a Windows machine with Rtools35 installed and there was no error. A similar scenario occurred for sparse matrix (https://protect-au.mimecast.com/s/Z1HxC0YZWVFLvLgZh2bUev?domain=urldefense.proofpoint.com ) where the error was "'a' must be a numeric matrix”.

Any help is greatly appreciated.

Thank you
Best Wishes
Kevin
The University of Sydney

[[alternative HTML version deleted]]

_______________________________________________
Bioc-devel using r-project.org<mailto:Bioc-devel using r-project.org> mailing list
https://protect-au.mimecast.com/s/V8VRCgZowLH767wpH3Aqcg?domain=urldefense.proofpoint.com


--
Hervé Pagès

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages using fredhutch.org<mailto:hpages using fredhutch.org>
Phone:  (206) 667-5791
Fax:    (206) 667-1319


	[[alternative HTML version deleted]]