[R] Odd Behavior Out of setdiff(...) - addition of duplicate entries is not identified

[G. Jay Kerns]

Jason,

On Fri, May 29, 2009 at 5:58 PM, Jason Rupert <jasonkrupert at yahoo.com> wrote:
>
> Jay,
>
>
> Thanks much for the reply.    I think you are right about the prob. Unfortunately, I was not able to find the old emails I had discussing the use of the more powerful setdiff that essentially inherits from the base class R setdiff functionality but extends that functionality by now working with data.frames instead of just a simple array of values.  Love this functionality.


Your previous post is here

[1]  http://tolstoy.newcastle.edu.au/R/e6/help/09/03/7781.html

and my earlier post is here:

[2]  https://stat.ethz.ch/pipermail/r-devel/2007-December/047706.html

(please note that the link in [1] referring to [2] is now broken).
As mentioned in [2], the notions of "set" and "element" are ambiguous
in the data frame case... what is an element...? a row, a column, or a
single entry?


>
> However, for the following example,
> Test1_DF<-data.frame(HouseSize=c(1:100), LandLocation=c("Here"))
> Test1_DF<-data.frame(HouseSize=c(1:100), LandLocation=c("Here"), Price = c("Low"))
> Test2_DF<-rbind(Test1_DF, Test1_DF)
> setdiff(Test1_DF, Test2_DF)
> [1] HouseSize    LandLocation Price
> <0 rows> (or 0-length row.names)
>> setdiff(Test2_DF, Test1_DF)
> [1] HouseSize    LandLocation Price
> <0 rows> (or 0-length row.names)
>
> I was hoping for this example one of the setdiff's would have returned essentially Test1_DF, since it is duplicated and that is what is different between the two dataframes.
>
> So, I guess I am trying to figure out a way to truely diff the dataframes, i.e. determine when two data.frames are different from one another and then receive the output of the results.
>
> Does this capability exist in a function within a current R package or does it exist within a typically used pattern to create this functionality?
>
> Thanks again for any feedback you can provide.




Your question speaks to the ambiguity above.  For instance, your 2nd
example would be solved by a setdiff for data frames that operates
column-wise.  If that is all you want, then IIRC there are at least 3
independent solutions in [2] to the row-wise problem.  It should be
easy enough to tweak one of them to operate on columns instead.

For an efficient setdiff() for data frames that can decipher
on-the-fly which of row/column/entry is desired, I am going to have to
defer to the aforementioned Ninjas.  :-)


>
> Also, I tried to determine my Session Info and the packages I have loaded, but I received the following:
>> sessionInfo()
> Error in x$Priority : $ operator is invalid for atomic vectors


Ninjas.


Hope this helps,
Jay