[R-sig-ME] Trouble getting development version of lme4

[Ross Boylan]

On 2/27/2013 7:59 PM, Ben Bolker wrote:
> Ross Boylan <ross at ...> writes:
>
>> Running on Windows 7, R 2.15.2, I have been unable to get the
>> development version of lme4 working.  I'd appreciate any assistance.
>>
>> Following the instructions at http://lme4.r-forge.r-project.org/
>>
>>   1.
>>
>>      install.packages("lme4",repos="http://r-forge.r-project.org")
>>      Installing package(s) into
> 'c:/Users/rdboylan/Documents/R/R-2.15.2/site-library'
>>      (as 'lib' is unspecified)
>>      Warning message:
>>      package 'lme4' is not available (for R version 2.15.2)
>>
>>   2.
>>
>>      install.packages("lme4",repos="http://lme4.r-forge.r-project.org/repos")
>   [snip]
>>      Warning: dependencies 'minqa', 'Rcpp', 'RcppEigen' are not available
>>
>>         There is a binary version available (and will be installed) but the
>>         source version is later:
>>                   binary        source
>>      lme4 0.999902344-0 0.999902345-0
>>
>>      Warning: dependencies 'minqa', 'Rcpp', 'RcppEigen' are not available
>   [snip]
>
>>      Content type 'application/zip' length 2027695 bytes (1.9 Mb)
>>      opened URL
>>      downloaded 1.9 Mb
>>
>>      package 'lme4' successfully unpacked and MD5 sums checked
>>      Warning: cannot remove prior installation of package 'lme4
>> ## I had installed the one from CRAN, but detached and unloaded
>> it before attempting installation
>>      # Do I need to manually delete the existing install?
>    You shouldn't need to.  Normally this happens when the package
> is still loaded while you're trying to install, but ...
>
>>   3. Then I tried to get the dependencies listed above from CRAN. This
>>      mostly worked, but
>>
>>      package 'minga' is not available (for R version 2.15.2)
>>
>>   4. The end result:
>>
>>      > library(lme4)
>>      Error in library(lme4) : there is no package called 'lme4'
>>
>>      Despite the message in 2, it appears lme4 was mostly removed, but
>>      there is still a dll there.  I thought detach  unload was supposed
>>      to take care of that.
>>
>> My guess is exiting ESS and deleting the installed lme4 files will fix
>> some of the problems, but it looks as if one required package, minga, is
>> unavailable.
>     Well, the major problem with this is that the package is called
> 'minqa', not 'minga'.
Well, that's a  good reason it didn't work!  Unfortunately, I still have 
problems with it, perhaps because minqa and lme4 are coming from 
different sources built against different versions of R.
I got minqa from
'http://r-forge.r-project.org/bin/windows/contrib/2.15/minqa_1.2.1.zip'
and lme4 from
'http://lme4.r-forge.r-project.org/repos/bin/windows/contrib/2.15/lme4_0.999902344-0.zip'
Neither is available on the other's repository.
Then
 > library(lme4)
Loading required package: lattice
Loading required package: Matrix
Error in inDL(x, as.logical(local), as.logical(now), ...) :
   unable to load shared object 
'c:/Users/rdboylan/Documents/R/R-2.15.2/site-library/minqa/libs/i386/minqa.dll':
   LoadLibrary failure:  Invalid access to memory location.

Error: package/namespace load failed for 'lme4'

BTW, ending the R session and exiting ess (but not emacs) did allow me 
to delete the residual old dll and install the new package.
>
>    I do apologize that things are a bit of a mess right now; we've
> been having trouble with the R-forge build because of some version/
> dependency problems with Rcpp.
>
>    A couple of general thoughts:
>
>   * development is now happening on github.  If you want to install
> the very latest version, and have the tools for compiling for source,
> try
>
> install.packages(c("Rcpp","RcppEigen","minqa","devtools"))
> library("devtools")
> install_github("lme4",user="lme4")
>
>    Do please let us/me know, if you try it, whether that works
> and/or whether you encounter problems.
>
>    In general for installing from R-forge but wanting to be
> able to get the dependencies from CRAN as well:
>
>   install.packages("lme4",repos=c("http://r-forge.r-project.org",
>        getOption("repos")))
>
>    In the very near future we hope to
> (1) build some updated binaries and put them up at
> http://lme4.r-forge.r-project.org/repos
That would be great.
> (2) update the installation instructions (we may stop referring
> to the automatically built r-forge binaries, since these will
> start falling behind the version on github)
The binaries are a great convenience.

The best news is that the CRAN version of lme4 seems able to do the 
poisson with my data as long as I don't specify nAGQ=.  Since my 
immediate concern is using this inside an imputation procedure, great 
precision is probably not that important.