[R] library sets: A & EMPTY does not work; gset_intersection(A,EMPTY) works

David Meyer david.meyer at wu.ac.at
Wed Apr 7 15:12:29 CEST 2010


> Apparently memberships values are identical, tags of set members are identical (letters),
> but "q == q" still generates FALSE...

The equality method for gsets performs some computations which might 
yield these results in case of numeric problems. As I wrote, I will 
think about using all.equal by default to avoid these confusions (and in 
particular to make sure that x == x gives always TRUE).

David

> 
> Best regards,
> Ryszard
> 
> p <- runif(length(letters))
> q <- gset(letters, p)
> summary(p == p) # gives all TRUE values, but
> q == q # gives FALSE
> summary(p == gset_memberships(q)) # and this gives all TRUE values
> 
>> p <- runif(length(letters))
>> q <- gset(letters, p)
>> summary(p == p) # gives all TRUE values, but
>    Mode    TRUE
> logical      26
>> q == q # gives FALSE
> [1] FALSE
>> summary(p == gset_memberships(q)) # and this gives all TRUE values
>    Mode    TRUE
> logical      26
> 
> --------------------------------------------------------------------------
> Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.
>  
> -----Original Message-----
> From: David Meyer [mailto:david.meyer at wu.ac.at] 
> Sent: Wednesday, April 07, 2010 8:25 AM
> To: Czerminski, Ryszard
> Cc: Peter Ehlers; R-help at R-project.org; David Meyer
> Subject: Re: [R] library sets: A & EMPTY does not work; gset_intersection(A,EMPTY) works
> 
>>> This is rather by chance due to numeric instabilities, I think:
>> It looks like something else is going on in addition to potential problem
>> with numerical instabilities:
>>
>> p <- runif(length(letters))
>> q <- gset(letters, p)
>> all.equal(p,p)
>> all.equal(q,q)
>> q == q
>>
>>> p <- runif(length(letters))
>>> q <- gset(letters, p)
>>> all.equal(p,p)
>> [1] TRUE
>>> all.equal(q,q)
>> [1] "target and current have different memberships"
>>> q == q
>> [1] FALSE
> 
> On my system:
> 
>  > p <- runif(length(letters))
>  > q <- gset(letters, p)
>  > all.equal(p,p)
> [1] TRUE
>  > all.equal(q,q)
> [1] TRUE
>  > q == q
> [1] TRUE
> 
> This is the same problem, since the all.equal-method for gsets does not 
> use all.equal internally (which *is* a bug, of course!). Will be fixed 
> ASAP ...
> 
> Thanks
> David
> 
> 
> 
>>
>> --------------------------------------------------------------------------
>> Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.
>>  
>> -----Original Message-----
>> From: David Meyer [mailto:david.meyer at wu.ac.at] 
>> Sent: Tuesday, April 06, 2010 6:03 PM
>> To: Czerminski, Ryszard
>> Cc: Peter Ehlers; R-help at R-project.org; David Meyer
>> Subject: Re: [R] library sets: A & EMPTY does not work; gset_intersection(A,EMPTY) works
>>
>> This is rather by chance due to numeric instabilities, I think:
>>
>>  > uv <- c('a', 'b', 'c'); s <- gset(uv, runif(length(uv)))
>>  > s
>> {"a" [0.7533966], "b" [0.968067], "c" [0.2494876]}
>>  > s == s
>> [1] TRUE
>>  > gset_is_equal(s,s)
>> [1] TRUE
>>
>> But using all.equal instead of `==` for the memberships might by a good 
>> idea, I will have a look at it.
>>
>> Thanks
>> David
>>
>> Czerminski, Ryszard wrote:
>>> Hi Peter,
>>>
>>> This looks like another one: "gset_is_equal(X,X)" and "X == X" evaluate to FALSE ?
>>>
>>>> uv <- c('a', 'b', 'c'); s <- gset(uv, runif(length(uv)))
>>>> s
>>> {"a" [0.0811552], "b" [0.3552998], "c" [0.996772]}
>>>> gset_is_equal(s, s)
>>> [1] FALSE
>>>> s == s
>>> [1] FALSE
>>>> class(s)
>>> [1] "gset" "cset"
>>>
>>> Best regards,
>>> Ryszard
>>>
>>>
>>> --------------------------------------------------------------------------
>>> Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.
>>>  
>>> -----Original Message-----
>>> From: Peter Ehlers [mailto:ehlers at ucalgary.ca] 
>>> Sent: Wednesday, March 31, 2010 2:03 PM
>>> To: Czerminski, Ryszard
>>> Cc: R-help at r-project.org; David Meyer
>>> Subject: Re: [R] library sets: A & EMPTY does not work; gset_intersection(A,EMPTY) works
>>>
>>> Ryszard,
>>>
>>> You've made me take a closer look and now I do think that
>>> you've found a bug.
>>>
>>> After a quick look at the package vignette, I see that the
>>> authors have indeed overloaded "&" and so it should work for
>>> your example. The problem seems to be the order of the class
>>> attribute which is used to call the relevant 'Ops' function:
>>>
>>> class(A)
>>> #[1] "gset" "cset"
>>>
>>> class(B)
>>> #[1] "gset" "cset"
>>>
>>> class(E <- A - A)
>>> #[1] "set"  "gset" "cset"
>>>
>>> If you re-order the class vector, function Ops.gset will be called
>>> to handle A and E:
>>>
>>> class(E) <- class(E)[c(2,3,1)]
>>> A & E
>>> #{}
>>>
>>> I've cc'd David Meyer.
>>>
>>>   -Peter Ehlers
>>>
>>> On 2010-03-31 10:11, Czerminski, Ryszard wrote:
>>>> It seems that "A&  B" works the same way as "gset_intersection(A,B)"
>>>> as long as A and B are not empty...
>>>> see below:
>>>>
>>>> fuzzy_logic("Yager", p=2)
>>>> support<- universe<- c('a','b','c','d')
>>>> A<- gset(support=support, memberships=c(0.2, 0.2, 0.9, 0),
>>>> universe=universe)
>>>> B<- gset(support=support, memberships=c(0.211, 0.222, 0.999, 0),
>>>> universe=universe)
>>>> A
>>>> B
>>>> A&  B
>>>> gset_intersection(A,B)
>>>> E<- A - A
>>>> A&  E
>>>> gset_intersection(A,E)
>>>>
>>>>> fuzzy_logic("Yager", p=2)
>>>>> support<- universe<- c('a','b','c','d')
>>>>> A<- gset(support=support, memberships=c(0.2, 0.2, 0.9, 0),
>>>> universe=universe)
>>>>> B<- gset(support=support, memberships=c(0.211, 0.222, 0.999, 0),
>>>> universe=universe)
>>>>> A
>>>> {"a" [0.2], "b" [0.2], "c" [0.9]}
>>>>> B
>>>> {"a" [0.211], "b" [0.222], "c" [0.999]}
>>>>> A&  B
>>>> {"c" [0.899995]}
>>>>> gset_intersection(A,B)
>>>> {"c" [0.899995]}
>>>>> E<- A - A
>>>>> A&  E
>>>> Error in A&  E :
>>>>    operations are possible only for numeric or logical types
>>>> In addition: Warning message:
>>>> Incompatible methods ("Ops.gset", "Ops.set") for "&"
>>>>> gset_intersection(A,E)
>>>> {}
>>>>
>>>>
>>>> --------------------------------------------------------------------------
>>>> Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.
>>>>
>>>> -----Original Message-----
>>>> From: Peter Ehlers [mailto:ehlers at ucalgary.ca]
>>>> Sent: Wednesday, March 31, 2010 11:43 AM
>>>> To: Czerminski, Ryszard
>>>> Cc: R-help at r-project.org
>>>> Subject: Re: [R] library sets: A&  EMPTY does not work;
>>>> gset_intersection(A,EMPTY) works
>>>>
>>>> On 2010-03-31 9:30, Peter Ehlers wrote:
>>>>> Unless I'm missing something, I don't see any method
>>>>> in pkg:sets for intersection other than gset_intersection.
>>>> Whoops, a bit quick on the draw.
>>>> There are of course also set_intersection and cset_intersection,
>>>> but not AFAICS any method for `&`.
>>>>
>>>>    -Peter Ehlers
>>>>
>>>>> So you're using the base R function `&` whose help page
>>>>> tells you that its arguments should be vectors. Yours
>>>>> aren't.
>>>>>
>>>>> -Peter Ehlers
>>>>>
>>>>> On 2010-03-31 8:50, Czerminski, Ryszard wrote:
>>>>>> When using generalized sets from "sets" library
>>>>>>
>>>>>> A&  EMPTY does not work, but gset_intersection(A,EMPTY) works: example
>>>>>> code below
>>>>>>
>>>>>> Is it a bug?
>>>>>>
>>>>>> Best regards,
>>>>>> Ryszard
>>>>>>
>>>>>> library(sets)
>>>>>> support<- universe<- c('a','b','c','d')
>>>>>> A<- gset(support=support, memberships=c(0.2, 0.2, 0.9, 0),
>>>>>> universe=universe)
>>>>>> class(A)
>>>>>> A
>>>>>> E<- A - A # create empty set
>>>>>> I<- gset_intersection(A, E) #<<<  this WORKS
>>>>>> class(I); I
>>>>>> I<- A&  E #<<<  this DOES NOT WORK
>>>>>>
>>>>>> # ---
>>>>>>
>>>>>>> support<- universe<- c('a','b','c','d')
>>>>>>> A<- gset(support=support, memberships=c(0.2, 0.2, 0.9, 0),
>>>>>> universe=universe)
>>>>>>> class(A)
>>>>>> [1] "gset" "cset"
>>>>>>> A
>>>>>> {"a" [0.2], "b" [0.2], "c" [0.9]}
>>>>>>> E<- A - A
>>>>>>> I<- gset_intersection(A, E) #<<<  this WORKS
>>>>>>> class(I); I
>>>>>> [1] "set" "gset" "cset"
>>>>>> {}
>>>>>>> I<- A&  E #<<<  this DOES NOT WORK
>>>>>> Error in A&  E :
>>>>>> operations are possible only for numeric or logical types
>>>>>> In addition: Warning message:
>>>>>> Incompatible methods ("Ops.gset", "Ops.set") for "&"
>>>>>>
>>>>>>
>>>>>>
>>>> ------------------------------------------------------------------------
>>>> --
>>>>>> Confidentiality Notice: This message is private and may
>>>> ...{{dropped:8}}
>>>>>> ______________________________________________
>>>>>> R-help at r-project.org mailing list
>>>>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>>>> PLEASE do read the posting guide
>>>>>> http://www.R-project.org/posting-guide.html
>>>>>> and provide commented, minimal, self-contained, reproducible code.
>>>>>>
>>>>>>
> 

-- 
Priv.-Doz. Dr. David Meyer
Department of Information Systems and Operations

WU
Wirtschaftsuniversität Wien
Vienna University of Economics and Business
Augasse 2-6, 1090 Vienna, Austria
Tel: +43-1-313-36-4393
Fax: +43-1-313-36-90-4393
HP:  http://ec.wu.ac.at/~meyer



More information about the R-help mailing list