new.env does not recognize parents from subclassesof "environment"

Looks like those ETH pages were nnot updated after the R-2.12.0 release
(they do not show current R-tached / R-devel), hence CCing Martin Maechler.

Best,
Uwe Ligges





On 22.10.2010 21:54, Vitally S. wrote:

John Chambers<jmc4 at stanford.edu> writes:

You need to update your version of R (r-devel or 2.12 patched) to rev
53385 or later, and read NEWS, particularly the line:

- Assignment of an environment to functions or as an attribute to other
objects now works for subclasses of "environment".

I am following the news in daily snapshots from here
ftp://ftp.stat.math.ethz.ch/Software/R
and the above line is still not in the NEWS of today's version.

Thanks for the patch. Looking forward to the stable release.

Vitally.

On 10/22/10 10:20 AM, Vitally S. wrote:

Yet another inconsistency. environment<- does not work with S4:

setClass("myenv", contains = "environment")

[1] "myenv"

env<- new("myenv")
tf<- function(x){x}
environment(tf)<- env

Error in environment(tf)<- env :
replacement object is not an environment


Vitally.


John Chambers<jmc4 at stanford.edu> writes:

This is a problem related to the introduction of exact= into the [[
and [[<- functions. As Bill says, the

current

method misuses eval.parent() when that argument is added.

However, a simpler and more efficient solution is to migrate the
checks for subclasses of "environment" used in
other base code into the code for [[<- (and for $<-), at which
point the methods for these functions are no

longer

needed.

A solution on these lines is being tested now and will find its way
into r-devel and 2.12 patched.

One other point about the original posting:

Please don't use constructions like env at .xData. This depends on the
current implementation and is not part of

the

user-level definition. Use as(env, "environment") or equivalent.
(In this case, the assignment of the object's

own

environment was irrelevant to the error.)

John Chambers

On 10/21/10 9:21 AM, William Dunlap wrote:

The traceback looks very similar to a problem
in R 2.11.1 reported earlier this month by Troy Robertson.

From: r-devel-bounces at r-project.org
[mailto:r-devel-bounces at r-project.org] On Behalf Of Troy Robertson
Sent: Wednesday, October 06, 2010 6:13 PM
To: 'r-devel at R-project.org'
Subject: Re: [Rd] Recursion error after upgrade to
R_2.11.1[Sec=Unclassified]

It was due to a miscount of how many frames to go
up before evaluating an expression in
getMethod("[[<-",".environment") because setMethod()
introduced a local function in the new method.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: r-devel-bounces at r-project.org
[mailto:r-devel-bounces at r-project.org] On Behalf Of Vitally S.
Sent: Thursday, October 21, 2010 9:00 AM
To: John Chambers
Cc: r-devel at r-project.org
Subject: Re: [Rd] new.env does not recognize parents from
subclassesof "environment"



Here is an infinite recursion error which occurs only with S4
subclasses assignment.

setClass("myenv", contains = "environment")
#[1] "myenv"
env<- new("myenv")
env[[".me"]]<- ?
#Error: evaluation nested too deeply: infinite recursion /
options(expressions=)?


With basic types it works as expected:

env1<- new.env()
env1[[".me"]]<- env1

May be this is related to active bindings that you mentioned,
but I am still
reporting it here.

Vitally.

Thanks for the report. Should now be fixed in r-devel and

2.12 patched (rev 53383).

Please do report any cases where a subclass of environment

doesn't work. There are some known cases in locking and

active binding, that will be fixed in due course.

The workaround for any such problem is usually as.environment().

On 10/20/10 3:17 AM, Vitaly S. wrote:

Dear Developers,

A lot has been changed in the R12.0 with respect to

behavior of "environment"

subclasses. Many thanks for that.

One small irregularity, though; new.env does not allow the

parent to be from S4

subclass.

setClass("myenv", contains="environment")

[1] "myenv"

new.env(parent=new("myenv"))

Error in new.env(parent = new("myenv")) : 'enclos' must be

an environment

I wonder if this is a "planed" behavior.

The use of .xData slot obviously works:

new.env(parent=new("myenv")@.xData)

<environment: 063bb9e8>
Thanks,
Vitaly.

R-devel at r-project.org mailing list

R-devel at r-project.org mailing list

R-devel at r-project.org mailing list