[Bioc-devel] NAMESPACE questions

[Seth Falcon]

On 21 Nov 2005, ddiez at iib.uam.es wrote:
>> If in package A we wanted to call function baz from package B then
>> we could:
>>
>> - use B::baz()
>> - import baz from B in the NAMESPACE file and just call baz()
>
> This is only possible if package B has a NAMESPACE, isn't it?

Yes, that is correct.  You cannot use the import directive in a
NAMESPACE file for a package that does not have its own NAMESPACE
file.  In such a case you can:

 a) write the the package maintainer and ask them to add a name space?
    ;-)

 b) Add the package to the Depends field in DESCRIPTION and live
    without the protection a name space provides.

> And what if package A and B have either function foo() and B call
> also function foo() from package A. I defined a function log2 to
> manage log2 for some type of data and inside uses log2 from base
> package. But instead of that it tries to use log2 from my package.
> This is because my log2 if first in the search path. I could solve
> this using base::log2 in case log2 has a NAMESPACE but I prefer to
> rename my function. But developers deal normally in this situations?

Yes, both are good solutions.  In general, one should think carefully
before creating a function with the same name as a basic math function
in base.  There are times when this may be reasonable, but realize
that an end user who loads up your package (even with NAMESPACE) will
have a strange experience with log2.

> Another question is that I have some S4 classes and methods. I make
> the corresponding entry in the NAMESPACE file but the documentation
> tell something about .onLoad call that must be present. I see that
> in package Biobase is in file zzz.R in R directory but I don't known
> how to call it in my case. In fact it is not actually and my package
> compiles and works without warning.

First, you don't call .onLoad yourself.  All you do is define it.  The
.onLoad function is a "hook" function.  When your package is loaded, R
will look for a .onLoad function in your package.  If found, it will
be called.

As to whether or not you need one, I'm not sure.  It is true that the
Extension Manual says to include a .onLoad function that does
require(methods) if your package has S4 classes and a NAMESPACE.
However, I *suspect* this is no longer needed as long as you have
methods in Depends.

> By the way I think it really clarifies things. I have solved the
> issues with my package and now has a  working NAMESPACE file (maeby
> with some bugs?)

Excellent.  Thanks for asking some good questions.  I hope others will
find the discussion useful as well.

Best,

+ seth