On Mon, Mar 23, 2009 at 4:59 AM, Gary Strangman
<str...@nmr.mgh.harvard.edu> wrote:
>
> The rnumpy concept looks great, and addresses some annoyances I've
> encountered working at the interface between R and numpy.
Great!
> I hope to test
> rnumpy this week, but in the meantime ... the text suggests that R data
> frames can be manipulated/sliced using rnumpy. Can they be created
> directly as well? Is that what is meant by "numpy arrays are mapped
> correctly by default"? If so, what's the syntax?
For creating data frames, there are few different easy options:
1) If you have a numpy record array, and pass it to any R function or
to rnumpy.rcopy, then it will be mapped to an R data frame. So you can
just say
r.lm("y ~ x", my_recarray)
2) If you have a Python dict, then it will be mapped to an R list, and
it's easy to turn an R list into a data frame. So I usually end up
building data frames like:
d = {"x": [1, 2, 3], "y": [10, 11, 12]}
r.as_data_frame(d)
(Also, many functions like lm are happy to accept a raw list, so you
can make calls like 'r.lm("y ~ x", d)' directly.)
3) You can drop to the lower-level, more-explicit way of calling R
functions, ".rcall". It looks like this:
r.data_frame.rcall([("x", [1, 2, 3]), ("y", [10, 11, 12])])
(If you check out the definition of RClosure.__call__, then you'll see
it's just a thin convenience wrapper around .rcall. This identical to
r.data_frame(x=[1, 2,3], y=[10, 11, 12]), except that you control the
order of arguments, and the argument names don't undergo underscore
conversion.)
(1) is super-easy if you have a record array, but I find record arrays
a bit cumbersome to construct in the first place. (2) takes the least
typing, but you can't control which order your columns end up in. (I
find I usually don't care, but sometimes it matters.) (3) Gives you
explicit control over the column order, without much extra typing.
> My particular use case is
> conversion of flexible-type numpy arrays or numpy arrays of type object.
I'm not sure what exactly you mean by flexible-type numpy arrays.
Numpy arrays of type object are converted by:
-- we call the conversion routine recursively to map each element to R
-- then we bundle them up in an R list (the only R container type
that can hold items that have different types from each other)
What sort of object arrays are you working with, and what would you
want to happen to them?
> Also, can one return a data frame that has been manipulated in R to a
> numpy (flexible-type) array?
There isn't much explicit support for copying data out of R and into
Python, ATM. This lack has been annoying me too; I'm just not sure
exactly how it should work. Collecting use cases would be a good
step...
By "flexible-type" do you mean record array?
-- Nathaniel
------------------------------------------------------------------------------
Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are
powering Web 2.0 with engaging, cross-platform capabilities. Quickly and
easily build your RIAs with Flex Builder, the Eclipse(TM)based development
software that enables intelligent coding and step-through debugging.
Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com
_______________________________________________
rpy-list mailing list
rpy-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rpy-list
