Question

Trouble loading 'lumi' package

0

Entering edit mode

Alexander Azarian ▴ 30

@alexander-azarian-16955

Last seen 2.4 years ago

Armenia

Recently I've been having hard times trying to load the 'lumi' package. The message I am receiving is:

Error: package or namespace load failed for ‘lumi’:

.onLoad failed in loadNamespace() for 'HDF5Array', details:

call: init_global_counter(filepath)

error: file 'C:\Users\EE49~1\AppData\Local\Temp\RtmpSSfcD1/HDF5Array_dump_files_global_counter' already exists

Is there any piece of advice available on copeing with this issue?

lumi illumina hdf5array • 2.2k views

ADD COMMENT • link updated 6.3 years ago by James W. MacDonald 67k • written 6.4 years ago by Alexander Azarian ▴ 30

1

Entering edit mode

Hi Alexander,

It seems no problem for me to load both lumi and HDF5Array package on my Mac OS. I guess the issue is more to do with HDF5Array than lumi on Windows system. Did you read this thread yet?

error with HDF5Array

Best,

Lei

ADD REPLY • link 6.4 years ago lhuang7 ▴ 50

0

Entering edit mode

Hello Lei,

Thank you fo your instant reply. Unfortunately the trouble proved to be more persistant. Issue has occured right after upgrading to R v.3.5.1, ergo I assume I have to address to the authors of 'lumi' for an immediate mend.

Best regards,

Alexander

ADD REPLY • link 6.4 years ago Alexander Azarian ▴ 30

0

Entering edit mode

Hi Alexander,

I am one of the lumi package maintainers. To the best of my knowledge, lumi was originally designed to process Illumina Methylation and Expression Microarrays and does not support HDF5Array at this point. May I know the reason why you use both lumi and HDF5Array in your code? Could you provide the code snippet where you encounter this error? Please also include the session info.

Thanks,

Lei

ADD REPLY • link 6.4 years ago lhuang7 ▴ 50

score 0 · Answer 1 · 2018-08-20

@lhuang7: lumi imports methylumi, which in turn depends on minfi, which imports HDF5Array. So while your package doesn't directly need HDF5array, the package dependency tree ends up importing that package.

@Alexander: The error you show is weird - in essence it's saying that there is already the HDF5_dump_files_global_counter in your R tempdir. But the tempdir should be instantiated when you start R, and hypothetically never reused. When you start R, you should see this:

> dir(tempdir())
character(0)

And when you load lumi, you should then see this:

> library(lumi)
<snip>
> dir(tempdir())
[1] "HDF5Array_dataset_creation_global_counter"
[2] "HDF5Array_dump"                           
[3] "HDF5Array_dump_files_global_counter"      
[4] "HDF5Array_dump_log"                       
[5] "HDF5Array_dump_names_global_counter"

The fact that you get the error you see indicates that either you are somehow re-using a tempdir, or are somehow instantiating the files that HDF5Array uses, before the HDF5Array package is loaded. Both are weird things to have happen.

What happens if you start R in a clean directory (with no .RData file in it), or using R --vanilla? Can you then load lumi normally?