On Wed, 2009-04-22 at 23:53 -0500, Bo Peng wrote:
> On Wed, Apr 22, 2009 at 10:33 PM, laurent <lgaut...@gmail.com> wrote:
> > On Wed, 2009-04-22 at 12:16 -0500, Bo Peng wrote:
> > (...)
>
> I apologize for the wording of my previous email, which was sent, less
> than politely, out of my extreme frustration over a software I had
> high expectation. However, if you consider all the frustrations
> misuses of rpy2 and do not see bugs or feature requests there, I will
> not bother you again with them.
I am sorry to have to say that what seems mostly like unstructured and
public venting of frustration (or 無的放矢 if you prefer) might not meet
lengthy diplomatic answers, if any.
> >> If the correct answer is r('R.Version()').subset('svn rev')[0][0], I
> >> have to say that rpy2 is not for me.
> >
> > 1- There is not "one correct answer". There are several ways to achieve
> > it. Other options are :
> >
> > ro.r('R.Version()[["svn rev"]]')[0]
> > ro.r('R.Version()').r["svn rev"][0][0]
>
> These look as ugly as r('R.Version()').subset('svn rev')[0][0]. I do
> not want to go into all the technical complexities here because they
> are irrelevant in terms of user expectation. Essentially speaking,
> R.Version() returns a list of strings identifiable by keys so users
> expect something like a dictionary, even if it is something wrapped
> around a R object.
An R list is not a dictionary and cannot be reliably represented as such
for a) it is ordered, b) names do not have to be unique.
Spending time on the archives for this list could have been an
alternative to writing the above:
http://www.mail-archive.com/rpy-list@lists.sourceforge.net/msg01337.html
Otherwise, customization is straightforward:
http://rpy.sourceforge.net/rpy2/doc/html/robjects_convert.html#a-simple-example
> -- If R.Version()[["svn rev"]] returns '12345' in R, users will expect
> '12345' in rpy. The [0] part is simply redundant and makes me wonder
> what is [1].
>
> -- If r('R.Version()') returns the whole dictionary-like object, it is
> natural to expect r('R.Version()')['svn rev']. Users will have to know
> a lot about rpy2 to write something like r["svn rev"][0][0].
>
> Compared to r('R.Version()')['svn rev'] in rpy, the syntax in rpy2 is
> extremely difficult to understand.
I do not think so. The choice made is the one of consistency for the
python __getitem__/"["; R lists are recursive structures (having two
ways to be subset "[", and "[[").
A short section in the documentation describing particularities
about a list is what I mostly see missing in the short term.
"robjects" is otherwise still evolving and likely getting better.
It is one high-level interface, and there is one known alternative to it
(rnumpy).
> > 2- rpy is still around.
>
> I hope that it will continue to be maintained for future versions of
> Python and R.
>
> >> 1. rpy_classic is not the same as rpy.
> >
> > This is not a complete implementation of the rpy API because of:
> > - limited resources
> > - apparently a limited demand
> > Contributions have not been many.
>
> As the maintainer of another open source object, I do understand all
> these but please indicate clearly the differences between rpy_classic
> and rpy in the documentation.
The first sentence at
http://rpy.sourceforge.net/rpy2/doc/html/rpy_classic.html
appears to be:
"This module provides an API similar to the one in RPy-1.x (rpy)."
"Similar" is used here for "some of the same characteristics".
> >> 2. The R vector etc are too complicated to use, and I do not see any
> >> benefit of using them.
> >
> > Watch closer, I'd like to say.
>
> I looked through the rpy2 documentation and I could not find an answer
> to questions such as what is lacking in rpy, what problems rpy2 is
> trying to address, what are the advantages of rpy2 over rpy, and in
> which way users will benefit from rpy2. Instead, I saw you were
> motivated by some alternative implementation (??) of rpy. My guess is
> that this implementation allows rpy2 to expose internal R objects as
> native Python objects. These are of course not a bad thing but I would
> like to know how exactly normal users would benefit from this
> approach, at the cost of all the syntax complexities/oddities.
Read more, guess less.
- A low-level interface upon which high-level interface can be built
http://rpy.sourceforge.net/rpy2/doc/html/overview.html#design-notes
http://rpy.sourceforge.net/rpy2/doc/html/robjects.html#overview
- Better similarity between R and rpy code:
http://rpy.sourceforge.net/rpy2/doc/html/introduction.html#principal-component-analysis
- No more CONVERSION/NO_CONVERSION poking.
http://rpy.sourceforge.net/rpy2/doc/html/robjects.html#overview
> >> 3. The rpy_object stuff is almost useless without support for the . to
> >> _ conversion, because there are so many functions in R with .
> >> (dev.off(), as.xxxx()...).
>
> > It's straightforward to implement for yourself. Where is the/your
> > problem ?
>
> I had contributed to rpy before and I am still listed as one of the
> developers. However, I need to be persuaded that rpy2 is better than
> rpy to invest my time on it. Right now, I am afraid that rpy2 might be
> designed around technical possibilities, not users' need.
Thanks for sharing this with us, but whether rpy or rpy2 suits best your
needs is mostly for yourself to decide.
Also, since "users' needs" seems to be used in place of "my needs", may
be you can tell us a little more about what you are doing ?
L.
> Cheers,
> Bo
>
> ------------------------------------------------------------------------------
> Stay on top of everything new and different, both inside and
> around Java (TM) technology - register by April 22, and save
> $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco.
> 300 plus technical and hands-on sessions. Register today.
> Use priority code J9JMT32. http://p.sf.net/sfu/p
> _______________________________________________
> rpy-list mailing list
> rpy-list@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/rpy-list
------------------------------------------------------------------------------
Register Now & Save for Velocity, the Web Performance & Operations
Conference from O'Reilly Media. Velocity features a full day of
expert-led, hands-on workshops and two days of sessions from industry
leaders in dedicated Performance & Operations tracks. Use code vel09scf
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
rpy-list mailing list
rpy-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rpy-list
