R devel - Jan 2013 - Regarding the recent changes to @<-

Hello all,

  In one of the packages (phylobase) I'm contributing to, we define a class
as follows:

setClass("phylo4",
         representation(edge = "matrix",
                        edge.length = "numeric",
                        label = "character",
                        edge.label = "character",
                        order = "character",
                        annote = "list"),
         prototype = list(
                        edge = matrix(nrow = 0, ncol = 2,
                            dimname = list(NULL, c("ancestor",
"descendant"))),
                        edge.length = numeric(0),
                        label = character(0),
                        edge.label = character(0),
                        order = "unknown",
                        annote = list()
                       ),
         validity = checkPhylo4)

  Using today's SVN checkout (r61643), building an object of the class
'phylo4' like this:

phylo4(foo, annote="bar")

  doesn't return an error as it used to (given that annote is not a list).

  What is now the preferred way of checking that annote is a actually a
list?

  Thanks,
  -- François

	[[alternative HTML version deleted]]

Thanks.  No need for a complicated example actually:
> C1 <- setClass("C1", slots = c(x = "numeric"))
> c1 <- C1()
> c1 at x <- "Hello"
> validObject(c1)
Error in validObject(c1) : 
  invalid class "C1" object: invalid object for slot "x" in
class "C1": got class "character", should be or extend class
"numeric"

In making @<- substantially more efficient in r-devel, we lost some error
checking.  That's too efficient.

This will be fixed, but given that code has migrated from an R function to C and
that I'm not currently at home, it may take a little while.

John


On Jan 13, 2013, at 3:34 PM, Fran?ois Michonneau wrote:
> Hello all,
> 
>  In one of the packages (phylobase) I'm contributing to, we define a
class
> as follows:
> 
> setClass("phylo4",
>         representation(edge = "matrix",
>                        edge.length = "numeric",
>                        label = "character",
>                        edge.label = "character",
>                        order = "character",
>                        annote = "list"),
>         prototype = list(
>                        edge = matrix(nrow = 0, ncol = 2,
>                            dimname = list(NULL, c("ancestor",
> "descendant"))),
>                        edge.length = numeric(0),
>                        label = character(0),
>                        edge.label = character(0),
>                        order = "unknown",
>                        annote = list()
>                       ),
>         validity = checkPhylo4)
> 
>  Using today's SVN checkout (r61643), building an object of the class
> 'phylo4' like this:
> 
> phylo4(foo, annote="bar")
> 
>  doesn't return an error as it used to (given that annote is not a
list).
> 
>  What is now the preferred way of checking that annote is a actually a
> list?
> 
>  Thanks,
>  -- Fran?ois
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

>>>>> Fran?ois Michonneau <francois.michonneau at
gmail.com>
>>>>>     on Sun, 13 Jan 2013 18:34:36 -0500 writes:
    > Hello all,
    > In one of the packages (phylobase) I'm contributing to, we define a
class
    > as follows:

> setClass("phylo4",
>          representation(edge = "matrix",
>                         edge.length = "numeric",
>                         label = "character",
>                         edge.label = "character",
>                         order = "character",
>                         annote = "list"),
>          prototype = list(
>                         edge = matrix(nrow = 0, ncol = 2,
>                             dimname = list(NULL, c("ancestor",
> "descendant"))),
>                         edge.length = numeric(0),
>                         label = character(0),
>                         edge.label = character(0),
>                         order = "unknown",
>                         annote = list()
>                        ),
>          validity = checkPhylo4)
>   Using today's SVN checkout (r61643), building an object of the class
> 'phylo4' like this:
> phylo4(foo, annote="bar")
>   doesn't return an error as it used to (given that annote is not a
list).
Well, the above is not reproducible
{foo  and  checkPhylo4   are missing !}.

If I make it reproducible,
it does work, i.e., give an error for me (rev 61644):

phylo4 <- setClass("phylo4",
         representation(edge = "matrix",
                        edge.length = "numeric",
                        label = "character",
                        edge.label = "character",
                        order = "character",
                        annote = "list"),
         prototype = list(edge = matrix(nrow = 0, ncol = 2,
                                        dimnames = list(NULL,
                                        c("ancestor",
"descendant"))),
                          edge.length = numeric(0),
                          label = character(0),
                          edge.label = character(0),
                          order = "unknown",
                          annote = list()
         )
         )##               ),
         ## validity = checkPhylo4)

foo <- phylo4()
phylo4(foo, annote="bar")

-----------

Gives

Error in validObject(.Object) : 
  invalid class "phylo4" object: invalid object for slot
"annote" in class "phylo4": got class "character",
should be or extend class "list"

as it should.

Can you try providing a simple reproducible example?

>   What is now the preferred way of checking that annote is a actually a
> list?
>   Thanks,
>   -- Fran?ois
> 	[[alternative HTML version deleted]]