[R-SIG-Mac] Too many dependencies with BioConductor packages?

[Simon Urbanek]

On Jun 23, 2009, at 10:38 , David Winsemius wrote:

>
> On Jun 23, 2009, at 9:29 AM, Gad Abraham wrote:
>
>> On 23/6/09 11:12 PM, Simon Urbanek wrote:
>>>>>
>>>> Does "install dependencies" only install Depends or also Suggests?
>>>>
>>>
>>> It's equivalent to install.packages(dependencies=TRUE)
>>>
>>
>> OK, so it installs all the Depends, Imports, Suggests, and Enhances.
>
> But does the non-checked Package Installer action mimic what the  
> default "dependencies=NA" does?
>

No, in that case it's FALSE. The checkbox directly corresponds to the  
"dependencies" parameter - hence the name. (It precedes the time of  
"Enhanced" and friends so the life was simple back then ;)).

Cheers,
Simon


> "The default, NA, means c("Depends", "Imports") if lib is  
> unambiguous, and FALSE otherwise."
>
> (I think that means FALSE in my case, since I do not have an  
> unambiguous .libPaths(), but for most people it appears to mean NA. )
>>
>> Maybe this should be mentioned somehow in the GUI (rename "install  
>> dependencies"?). It's not too much of an issue for CRAN, but in  
>> BioConductor there are large packages (10s of MB at least) which  
>> can get downloaded needlessly but are only Suggested.
>>
>>
> As another user I vote "opposed" to that suggestion. I would want  
> the Install dependencies check box to do what the naive-user like  
> myself would expect, which is to mimic the action of  
> install.packages(<pkg>, <url>, dependencies=TRUE).
>
> As an alternative suggestion, it might be helpful to have the R  
> command echoed to the GUI console so that a syntactically correct  
> template could be modified.
>
> David Winsemius, MD
> Heritage Laboratories
> West Hartford, CT
>
>