[Phylobase-devl] several proposed additions

[Ben Bolker]

Hilmar Lapp wrote:
> Hi Jim -
> 
> this sounds great! I can't speak on behalf of the phylobase leads, but  
> can you commit all your changes to a branch? That would make it  
> easiest for people to check out and play with your code. The unit  
> tests sound cool - haven't been aware of this, but if it is what the  
> name suggests it will be highly useful.
> 
> 	-hilmar
> 
> On Jun 24, 2009, at 3:45 AM, Jim Regetz wrote:
> 
>> Hi all,
>>
>> I'd love to get feedback on any/all of the following. Below, when I  
>> talk
>> about things I've done, I mean in my local sources -- I haven't  
>> actually
>> committed any of the following. But it is all more or less ready to  
>> go,
>> albeit with a couple of minor rough edges that we can discuss  
>> further if
>> folks want to move ahead on any of these:
>>
>> * NAMESPACE: I've created one. Seems to work fine. I'm basically
>> exporting all functions and methods now. I figure it's easier to start
>> with a liberal approach, and later remove whatever items folks agree  
>> are
>> only required internally.

  We've had a lot of arguments about whether to use namespaces or
not (I have been very frustrated by namespaces in the past, partly
because I don't really understand them), but I don't think it's
a bad idea.

>>
>> * Unit tests: I've created a handful of RUnit tests, and added the
>> necessary "run tests" machinery as documented here:
>> http://wiki.r-project.org/rwiki/doku.php?id=developers:runit
>>
>> The tests are not even remotely comprehensive, but they do test  
>> against
>> the bug fixes I've implemented in the last week or so. I find unit  
>> tests
>> to be hugely helpful in making sure that fixed bugs stay fixed, and  
>> that
>> code changes in one area don't unexpectedly break things somewhere  
>> else.
>> And I find RUnit testing to provide better structure and test
>> expressivity than you get with unstructured test scripts in the pkg/ 
>> test
>> directory.
>>

  sounds great.
>> I'm not 100% certain that R-forge would do the right thing with these
>> tests without some additional tweaking, but they are successfully  
>> run by
>> R CMD CHECK on my local machine (at least as long as phylobase is
>> already installed).
>>
>> * phylo4 generic: Any reason why there has been a generic for phylo4d,
>> but not phylo4? I've now created a phylo4 generic, along with two
>> methods: one with signature x="matrix" (replacing the original phylo4
>> constructor function), and one with signature x="phylo" (imports phylo
>> to phylo4). I updated the associated doc pages to match, plus fixed a
>> few places in the examples and sources where phylo(edge=...) was  
>> called
>> explicitly (the first argument is now the more generic 'x', not  
>> 'edge').
>>
>> * Dealing with messy node labels: My motivation for creating the  
>> phylo4
>> generic was to expose an option allowing node labels to be dropped  
>> (or,
>> in the case of phylo4d, added to the tree data) during phylo import.
>> This is basically just a modified implementation of what François  
>> did to
>> deal with MrBayes sorts of trees, except now focused specifically on
>> phylo import (it could be added elsewhere too), and now with user- 
>> level
>> control of the behavior. I'm glossing over the details, but for those
>> not paying attention to the bug tracker, see more discussion here:
>> https://r-forge.r-project.org/tracker/index.php?func=detail&aid=466&group_id=111&atid=488
>>
>> FYI, R CMD CHECK passes on all the above changes (well, aside from a
>> latex documentation issue that I think has been around for some time).
>>
>> I'll be traveling much of the next month, but would be happy to  
>> discuss
>> any of this further. And of course, I can push some/all of this up to
>> the repository at any time, or send relevant code to the list if  
>> requested.
>>

  Please do push.  We need to move ahead.  I want to spend more time
working this summer (maybe with Francois), but probably won't happen
until second half of July ...

  Ben

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 260 bytes
Desc: OpenPGP digital signature
Url : http://lists.r-forge.r-project.org/pipermail/phylobase-devl/attachments/20090624/d18ade14/attachment.pgp