[BioC] Need better quality control for reliable installation of bioconductor packages?

Robert Gentleman rgentlem at fhcrc.org
Sat Dec 11 21:52:56 CET 2004


On Dec 11, 2004, at 6:43 AM, Gordon K Smyth wrote:

> Robert Gentleman rgentlem at fhcrc.org wrote
>> On Dec 10, 2004, at 10:21 AM, Glynn, Earl wrote:
>>> Under Windows I just downloaded the new 2.0.1 version of R and
>>> installed
>>> all the packages:
>>>
>>> 	Packages | Install package(s) from CRAN ...
>>>
>>> I select all the packages for installation and that worked fine.   
>>> This
>>> took about an hour.
>>>
>>> So with the new version of R, I thought I'd re-install all the
>>> Bioconductor packages. Like several times in the past, when I try to
>>> install all the Bioconductor packages I encounter problems:
>>>
>>> 	Packages | Install package(s) from Bioconductor ...
>>>
>>> I select all the packages for installation.  The download take a long
>>> time but then the installation of packages fails.  I wish the name of
>>> the temporary download directory was displayed earlier to make it
>>> easier
>>> to find.  But I found the Bioconductor ZIP files and then tried
>>> Packages
>>> | Install Package(s) from local zip files ...
>>>
>>> Why is the installation of so many Bioconductor packages failing?
>>>
>>> For example:
>>>
>>
>>  But you have not shown us what commands you are using?
>
> Effectively he has.  He has explained that he is using the drop-down  
> menus under the Windows Rgui.
>  This completely specifies the commands involved.
>
>>  You can (and possibly should) use the reposTools library and the  
>> tools
>> that we have built for dealing with the package dependencies that  
>> exist
>> in and between many Bioconductor packages.
>
> If a drop-down menu is provided for installing Bioconductor packages,  
> it seems not unreasonable
> that users might use it.  If you don't want people to use the menu, it  
> should probably be removed
> from R.
>
>>> install.packages(choose.files('',filters=Filters[c('zip','All'),]),
>>> .libPaths()[1], CRAN = NULL)
>>> 	Error in file(file, "r") : unable to open connection
>>> 	In addition: Warning messages:
>>> 	1: error 1 in extracting from zip file
>>> 	2: cannot open file `twilight/DESCRIPTION'
>>> 	>
>>> install.packages(choose.files('',filters=Filters[c('zip','All'),]),
>>> .libPaths()[1], CRAN = NULL)
>>> 	Error in file(file, "r") : unable to open connection
>>> 	In addition: Warning messages:
>>> 	1: error 1 in extracting from zip file
>>> 	2: cannot open file `iSPlot/DESCRIPTION'
>>> 	>
>>
>>   Did you look to see what the problem is? Try just installing a  
>> single
>> package, giving the correct path. These are just zip files open them
>> and see if they are complete (it isn't very hard).
>> Do you have Gtk properly installed? If not why would you want a  
>> package
>> that relies on it?
>>
>>> install.packages(choose.files('',filters=Filters[c('zip','All'),]),
>>> .libPaths()[1], CRAN = NULL)
>>> 	Error in file(file, "r") : unable to open connection
>>> 	In addition: Warning messages:
>>> 	1: error 1 in extracting from zip file
>>> 	2: cannot open file `affycomp/DESCRIPTION'
>>> 	>
>>> install.packages(choose.files('',filters=Filters[c('zip','All'),]),
>>> .libPaths()[1], CRAN = NULL)
>>> 	Error in file(file, "r") : unable to open connection
>>> 	In addition: Warning messages:
>>> 	1: error 1 in extracting from zip file
>>> 	2: cannot open file `RdbiPgSQL/DESCRIPTION'
>>>
>>
>>   Do you have Postgres properly installed on your computer? If not,  
>> why
>> would you expect this package to install or work? I don't think that
>> blindly trying to install every package, regardless of whether you  
>> want
>> or need to use it makes very much sense. As for why these particular
>> ones have failed, one needs to look and see if they downloaded (maybe
>> there was a time out), what the name of the downloaded file is, does  
>> it
>> match the name in the command you issued, and then you can look at the
>> zip file and see if it is ok. I suspect that there is a failure
>> somewhere in one of those three places (but I could be wrong).
>>
>>> This installation process is quite frustrating since one selects all
>>> the
>>> packages to install and then it fails.  This must be repeated every
>>> time
>>> for a failure.  So far I've seen at least 9 failures.  What wrong  
>>> with
>>> these DESCRIPTION directories?
>>>
>>
>>   You have chosen to install the development version of these  
>> packages,
>> that means that you are working with unstable software.
>
> No he hasn't.  There is in fact no way to ask for developmental  
> packages using the Windows menu.
> The menu downloads Bioconductor Release 1.5.
>
>> If you want
>> stable software you can use the release version, or you can pay  
>> someone
>> (like Insightful for that). But as I noted above, I suspect that the
>> problem lies with the downloaded files, and that is the first place to
>> start looking.
>
> It seems fair to compare Bioconductor with R/CRAN, which is also free.
>
>>> Why isn't the installation of bioconductor packages more reliable?
>>> (I've
>>> seen this same problem several times before).  What quality control
>>> measures can be put in place to catch these problems and fix them
>>> automatically?
>>
>>   Well, we would like to make it more stable, and accurate and
>> informative user reports help us to do that. You could help, by trying
>> to find out just what the problem is, and no we can't tell what is
>> happening on your computer, so you have to be slightly more
>> descriptive.
>
> It doesn't take much checking to see that most of the problem is on  
> Bioconductor rather than on
> his computer.  None of the packages twilight, iSPlot or RdbiPgSQL have  
> Windows versions available
> as part of the Bioconductor 1.5 release.
>

  Thanks for clarifying that. We will check and fix that problem. I am  
not sure how packages that do not get built for windows ended up in the  
manifest. There were however other problems mentioned and these were  
primarily due to failure to download packages in their entirety. That  
is not something that we can deal with. Now, you might expect, as Earl  
did, that the downloading tools in R would report a failure. And  
perhaps they should, but that is an R question not a Bioconductor one,  
since it is R that is being used to download software.

  Earl was also unhappy about the error reporting capabilities of  
install.packages. Again, I don't disagree, but that is not a failure of  
Bioconductor, he is using standard R tools and if they are not doing  
what he wants, then he needs to put in feature requests, probably on  
R-devel.

   We do want to provide a good software environment but in this, as in  
any project bugs exist.  There is a bug in Bioconductor, the manifest  
contains things it should not. We will fix it.

> So the question is, why are these packages included in the Windows  
> menu list if they are not
> actually available for Windows?  CRAN has a system in place whereby  
> packages are listed only if
> they are actually available for the operating system being used.  Why  
> not Bioconductor?
>

    We should be doing what CRAN is doing, and will be soon, I hope.

  Robert

> Gordon
>
>>   As for what could make things easier, well, contributing money is  
>> one
>> if you send us a lot we can fix it really fast. Otherwise, you can
>> supply informative bug reports that can be used to figure out what has
>> gone wrong.
>>
>>   Robert
>>
>>>
>>> efg
>>> --
>>> Earl F. Glynn
>>> Scientific Programmer
>>> Bioinformatics Department
>>> Stowers Institute for Medical Research
>>>
>>>
>>> 	[[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> Bioconductor mailing list
>>> Bioconductor at stat.math.ethz.ch
>>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>>>
>>>
>> +--------------------------------------------------------------------- 
>> --
>> ----+
>> | Robert Gentleman              phone: (206) 667-7700            |
>> | Head, Program in Computational Biology   fax:
>>     |
>> | Division of Public Health Sciences       office: M2-B865
>>     |
>> | Fred Hutchinson Cancer Research Center   email: rgentlem at  
>> fhcrc.org
>>     |
>> +--------------------------------------------------------------------- 
>> --
>> ----+
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
>
>
+----------------------------------------------------------------------- 
----+
| Robert Gentleman              phone: (206) 667-7700            |
| Head, Program in Computational Biology   fax:                          
     |
| Division of Public Health Sciences       office: M2-B865               
     |
| Fred Hutchinson Cancer Research Center   email: rgentlem at fhcrc.org     
     |
+----------------------------------------------------------------------- 
----+



More information about the Bioconductor mailing list