Eric Archer - NOAA Federal
2016-Jul-01 18:23 UTC
[R] C stack error in as.vector() starting in R 3.3.0
Apologies for the long post. This is an issue I have been struggling with and I have tried to be as complete, to the point, and reproducible as possible. In documenting a package with roxygen2, I have come across an error that does not occur in R 3.2.4 revised, but does occur in R 3.3.0 and 3.3.1. Using traceback() and debug(), I've traced the error to a call made to as.vector(x, "character") that seems to get stuck in a loop which culminates in this error: Error: C stack usage 7970892 is too close to the limit The object that causes this error is of a signature type for a method. With some playing around, I've been able to work out that the error is actually associated with the method that roxygen2 creates when doing its magic. Something happens when this method definition with its associated environment is in the workspace that causes the error. At this point, I should stress again that the error does NOT occur in R 3.2.4 revised or earlier, but does occur in R 3.3.0 and 3.3.1. I have also tested this with several versions of roxygen2 and that does not make a difference. Thus, my ultimate question is what has changed in R 3.3.0 that would lead to this so that the roxygen2 maintainers can correct it? As a test, I created a signature identical to the one that causes the error and tried the as.vector() command that would generate the loop. I don't get the error in my command line generated object, nor with the object that is extracted from the method definition. However, when I load the method definition into the workspace, the error will happen on either my command line generated object, or the one extracted from the method definition and will not stop happening until I restart R. I have reported this error as an issue on the roxygen2 GitHub repository and it has been crossposted by another user who had a similar experience with a different package on the devtools repository. Those posts, which contain more information are here: https://github.com/klutometis/roxygen/issues/475 https://github.com/hadley/devtools/issues/1234 Below is the result of sessionInfo() and the output of a session demonstrating the effect. The files used are in a zip file here: https://github.com/klutometis/roxygen/files/335417/error.test.rdata.files.zip > sessionInfo() R version 3.3.0 (2016-05-03) Platform: x86_64-apple-darwin13.4.0 (64-bit) Running under: OS X 10.11.5 (El Capitan) locale: [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 attached base packages: [1] stats graphics grDevices utils datasets methods base loaded via a namespace (and not attached): [1] tools_3.3.0> rm(list = ls()) > > # Create test class and method > setClass(Class = "gtypes", slots = c(loci = "data.frame", ploidy "numeric"), package = "roxygen_devtest") > setGeneric("nInd", function(x, ...) standardGeneric("nInd"), package "adegenet")[1] "nInd"> setMethod("nInd", "gtypes", function(x, ...) nrow(x at loci) / x at ploidy)[1] "nInd"> > test.method <- getMethod("nInd", "gtypes") > str(test.method)Formal class 'MethodDefinition' [package "methods"] with 4 slots ..@ .Data :function (x, ...) ..@ target :Formal class 'signature' [package "methods"] with 3 slots .. .. ..@ .Data : chr "gtypes" .. .. ..@ names : chr "x" .. .. ..@ package: chr "roxygen_devtest" ..@ defined:Formal class 'signature' [package "methods"] with 3 slots .. .. ..@ .Data : chr "gtypes" .. .. ..@ names : chr "x" .. .. ..@ package: chr "roxygen_devtest" ..@ generic: atomic [1:1] nInd .. ..- attr(*, "package")= chr "adegenet"> > # No error: > as.vector(test.method at defined, "character")[1] "gtypes"> > # This is the method that generates the error > load("problem.method.rdata") > > # Notice the difference in the environments of the two functions: > test.method at .Datafunction (x, ...) nrow(x at loci)/x at ploidy> problem.method at .Datafunction (x, ...) nrow(x at loci)/x at ploidy <environment: 0x101f85040>> > # Swap the problem function for the one I created: > problem.method at .Data <- test.method at .Data > save(problem.method, file = "fixed.method.rdata") > > # Here's the error (both with my method and the original): > as.vector(test.method at defined, "character")Error: C stack usage 7970892 is too close to the limit> as.vector(problem.method at defined, "character")Error: C stack usage 7970892 is too close to the limit Restarting R session...> # *** Restart R for this to work *** > rm(list = ls()) > > load("fixed.method.rdata") > as.vector(problem.method at defined, "character")[1] "gtypes" ---- *Eric Archer, Ph.D.* Southwest Fisheries Science Center NMFS, NOAA 8901 La Jolla Shores Drive La Jolla, CA 92037 USA 858-546-7121 (work) 858-546-7003 (FAX) Marine Mammal Genetics Group: swfsc.noaa.gov/mmtd-mmgenetics ETP Cetacean Assessment Program: swfsc.noaa.gov/mmtd-etp https://github/ericarcher " *The universe doesn't care what you believe. The wonderful thing about science is that it doesn't ask for your faith, it just asks for your eyes.*" - Randall Munroe "*Lighthouses are more helpful than churches.*" - Benjamin Franklin "*...but I'll take a GPS over either one.*" - John C. "Craig" George [[alternative HTML version deleted]]
luke-tierney at uiowa.edu
2016-Jul-06 23:06 UTC
[R] C stack error in as.vector() starting in R 3.3.0
I cannot reproduce this. My best guess is that there is a problem,
maybe a version incompatibility, with one of the packages loaded when
load("problem.method.rdata"). This includes
adegenet
apex
copula
Rcpp
DBI
sp
colorspace
and any dependencies these are bringing in.
Best,
luke
On Fri, 1 Jul 2016, Eric Archer - NOAA Federal wrote:
> Apologies for the long post. This is an issue I have been struggling with
> and I have tried to be as complete, to the point, and reproducible as
> possible.
>
> In documenting a package with roxygen2, I have come across an error that
> does not occur in R 3.2.4 revised, but does occur in R 3.3.0 and 3.3.1.
> Using traceback() and debug(), I've traced the error to a call made to
> as.vector(x, "character") that seems to get stuck in a loop which
> culminates in this error:
>
> Error: C stack usage 7970892 is too close to the limit
>
> The object that causes this error is of a signature type for a method. With
> some playing around, I've been able to work out that the error is
actually
> associated with the method that roxygen2 creates when doing its magic.
> Something happens when this method definition with its associated
> environment is in the workspace that causes the error.
>
> At this point, I should stress again that the error does NOT occur in R
> 3.2.4 revised or earlier, but does occur in R 3.3.0 and 3.3.1. I have also
> tested this with several versions of roxygen2 and that does not make a
> difference. Thus, my ultimate question is what has changed in R 3.3.0 that
> would lead to this so that the roxygen2 maintainers can correct it?
>
> As a test, I created a signature identical to the one that causes the error
> and tried the as.vector() command that would generate the loop. I don't
get
> the error in my command line generated object, nor with the object that is
> extracted from the method definition. However, when I load the method
> definition into the workspace, the error will happen on either my command
> line generated object, or the one extracted from the method definition and
> will not stop happening until I restart R.
>
> I have reported this error as an issue on the roxygen2 GitHub repository
> and it has been crossposted by another user who had a similar experience
> with a different package on the devtools repository. Those posts, which
> contain more information are here:
>
> https://github.com/klutometis/roxygen/issues/475
> https://github.com/hadley/devtools/issues/1234
>
> Below is the result of sessionInfo() and the output of a session
> demonstrating the effect. The files used are in a zip file here:
>
https://github.com/klutometis/roxygen/files/335417/error.test.rdata.files.zip
>
> > sessionInfo()
> R version 3.3.0 (2016-05-03)
> Platform: x86_64-apple-darwin13.4.0 (64-bit)
> Running under: OS X 10.11.5 (El Capitan)
>
> locale:
> [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
>
> attached base packages:
> [1] stats graphics grDevices utils datasets methods base
>
> loaded via a namespace (and not attached):
> [1] tools_3.3.0
>
>> rm(list = ls())
>>
>> # Create test class and method
>> setClass(Class = "gtypes", slots = c(loci =
"data.frame", ploidy > "numeric"), package =
"roxygen_devtest")
>> setGeneric("nInd", function(x, ...)
standardGeneric("nInd"), package > "adegenet")
> [1] "nInd"
>> setMethod("nInd", "gtypes", function(x, ...) nrow(x
at loci) / x at ploidy)
> [1] "nInd"
>>
>> test.method <- getMethod("nInd", "gtypes")
>> str(test.method)
> Formal class 'MethodDefinition' [package "methods"] with
4 slots
> ..@ .Data :function (x, ...)
> ..@ target :Formal class 'signature' [package "methods"]
with 3 slots
> .. .. ..@ .Data : chr "gtypes"
> .. .. ..@ names : chr "x"
> .. .. ..@ package: chr "roxygen_devtest"
> ..@ defined:Formal class 'signature' [package "methods"]
with 3 slots
> .. .. ..@ .Data : chr "gtypes"
> .. .. ..@ names : chr "x"
> .. .. ..@ package: chr "roxygen_devtest"
> ..@ generic: atomic [1:1] nInd
> .. ..- attr(*, "package")= chr "adegenet"
>>
>> # No error:
>> as.vector(test.method at defined, "character")
> [1] "gtypes"
>>
>> # This is the method that generates the error
>> load("problem.method.rdata")
>>
>> # Notice the difference in the environments of the two functions:
>> test.method at .Data
> function (x, ...)
> nrow(x at loci)/x at ploidy
>> problem.method at .Data
> function (x, ...)
> nrow(x at loci)/x at ploidy
> <environment: 0x101f85040>
>>
>> # Swap the problem function for the one I created:
>> problem.method at .Data <- test.method at .Data
>> save(problem.method, file = "fixed.method.rdata")
>>
>> # Here's the error (both with my method and the original):
>> as.vector(test.method at defined, "character")
> Error: C stack usage 7970892 is too close to the limit
>> as.vector(problem.method at defined, "character")
> Error: C stack usage 7970892 is too close to the limit
>
> Restarting R session...
>
>> # *** Restart R for this to work ***
>> rm(list = ls())
>>
>> load("fixed.method.rdata")
>> as.vector(problem.method at defined, "character")
> [1] "gtypes"
>
> ----
>
> *Eric Archer, Ph.D.*
> Southwest Fisheries Science Center
> NMFS, NOAA
> 8901 La Jolla Shores Drive
> La Jolla, CA 92037 USA
> 858-546-7121 (work)
> 858-546-7003 (FAX)
>
> Marine Mammal Genetics Group: swfsc.noaa.gov/mmtd-mmgenetics
> ETP Cetacean Assessment Program: swfsc.noaa.gov/mmtd-etp
> https://github/ericarcher
>
> "
>
>
> *The universe doesn't care what you believe. The wonderful thing about
> science is that it doesn't ask for your faith, it just asks for
your
> eyes.*" - Randall Munroe
>
> "*Lighthouses are more helpful than churches.*"
> - Benjamin Franklin
>
> "*...but I'll take a GPS over either one.*"
> - John C. "Craig" George
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.
>
--
Luke Tierney
Ralph E. Wareham Professor of Mathematical Sciences
University of Iowa Phone: 319-335-3386
Department of Statistics and Fax: 319-335-3017
Actuarial Science
241 Schaeffer Hall email: luke-tierney at uiowa.edu
Iowa City, IA 52242 WWW: http://www.stat.uiowa.edu
Eric Archer - NOAA Federal
2016-Jul-07 04:02 UTC
[R] C stack error in as.vector() starting in R 3.3.0
Luke,
Thanks for this point out. I started removing packages one-by-one in R
v.3.3.1. Oddly enough, it went away after I removed the Matrix package.
When I restarted R, the same version of Matrix (1.2-6) had been
"reinstalled" (?), but now no C stack error. Lets see if the others
that
were having a similar problem have the same result.
Cheers,
Eric
----
*Eric Archer, Ph.D.*
Southwest Fisheries Science Center
NMFS, NOAA
8901 La Jolla Shores Drive
La Jolla, CA 92037 USA
858-546-7121 (work)
858-546-7003 (FAX)
Marine Mammal Genetics Group: swfsc.noaa.gov/mmtd-mmgenetics
ETP Cetacean Assessment Program: swfsc.noaa.gov/mmtd-etp
https://github/ericarcher
"
*The universe doesn't care what you believe. The wonderful thing about
science is that it doesn't ask for your faith, it just asks for your
eyes.*" - Randall Munroe
"*Lighthouses are more helpful than churches.*"
- Benjamin Franklin
"*...but I'll take a GPS over either one.*"
- John C. "Craig" George
On Wed, Jul 6, 2016 at 4:06 PM, <luke-tierney at uiowa.edu> wrote:
> I cannot reproduce this. My best guess is that there is a problem,
> maybe a version incompatibility, with one of the packages loaded when
> load("problem.method.rdata"). This includes
>
> adegenet
> apex
> copula
> Rcpp
> DBI
> sp
> colorspace
>
> and any dependencies these are bringing in.
>
> Best,
>
> luke
>
> On Fri, 1 Jul 2016, Eric Archer - NOAA Federal wrote:
>
> Apologies for the long post. This is an issue I have been struggling with
>> and I have tried to be as complete, to the point, and reproducible as
>> possible.
>>
>> In documenting a package with roxygen2, I have come across an error
that
>> does not occur in R 3.2.4 revised, but does occur in R 3.3.0 and 3.3.1.
>> Using traceback() and debug(), I've traced the error to a call made
to
>> as.vector(x, "character") that seems to get stuck in a loop
which
>> culminates in this error:
>>
>> Error: C stack usage 7970892 is too close to the limit
>>
>> The object that causes this error is of a signature type for a method.
>> With
>> some playing around, I've been able to work out that the error is
actually
>> associated with the method that roxygen2 creates when doing its magic.
>> Something happens when this method definition with its associated
>> environment is in the workspace that causes the error.
>>
>> At this point, I should stress again that the error does NOT occur in R
>> 3.2.4 revised or earlier, but does occur in R 3.3.0 and 3.3.1. I have
also
>> tested this with several versions of roxygen2 and that does not make a
>> difference. Thus, my ultimate question is what has changed in R 3.3.0
that
>> would lead to this so that the roxygen2 maintainers can correct it?
>>
>> As a test, I created a signature identical to the one that causes the
>> error
>> and tried the as.vector() command that would generate the loop. I
don't
>> get
>> the error in my command line generated object, nor with the object that
is
>> extracted from the method definition. However, when I load the method
>> definition into the workspace, the error will happen on either my
command
>> line generated object, or the one extracted from the method definition
and
>> will not stop happening until I restart R.
>>
>> I have reported this error as an issue on the roxygen2 GitHub
repository
>> and it has been crossposted by another user who had a similar
experience
>> with a different package on the devtools repository. Those posts, which
>> contain more information are here:
>>
>> https://github.com/klutometis/roxygen/issues/475
>> https://github.com/hadley/devtools/issues/1234
>>
>> Below is the result of sessionInfo() and the output of a session
>> demonstrating the effect. The files used are in a zip file here:
>>
>>
https://github.com/klutometis/roxygen/files/335417/error.test.rdata.files.zip
>>
>> > sessionInfo()
>> R version 3.3.0 (2016-05-03)
>> Platform: x86_64-apple-darwin13.4.0 (64-bit)
>> Running under: OS X 10.11.5 (El Capitan)
>>
>> locale:
>> [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
>>
>> attached base packages:
>> [1] stats graphics grDevices utils datasets methods base
>>
>> loaded via a namespace (and not attached):
>> [1] tools_3.3.0
>>
>> rm(list = ls())
>>>
>>> # Create test class and method
>>> setClass(Class = "gtypes", slots = c(loci =
"data.frame", ploidy >>>
>> "numeric"), package = "roxygen_devtest")
>>
>>> setGeneric("nInd", function(x, ...)
standardGeneric("nInd"), package >>>
>> "adegenet")
>> [1] "nInd"
>>
>>> setMethod("nInd", "gtypes", function(x, ...)
nrow(x at loci) / x at ploidy)
>>>
>> [1] "nInd"
>>
>>>
>>> test.method <- getMethod("nInd", "gtypes")
>>> str(test.method)
>>>
>> Formal class 'MethodDefinition' [package "methods"]
with 4 slots
>> ..@ .Data :function (x, ...)
>> ..@ target :Formal class 'signature' [package
"methods"] with 3 slots
>> .. .. ..@ .Data : chr "gtypes"
>> .. .. ..@ names : chr "x"
>> .. .. ..@ package: chr "roxygen_devtest"
>> ..@ defined:Formal class 'signature' [package
"methods"] with 3 slots
>> .. .. ..@ .Data : chr "gtypes"
>> .. .. ..@ names : chr "x"
>> .. .. ..@ package: chr "roxygen_devtest"
>> ..@ generic: atomic [1:1] nInd
>> .. ..- attr(*, "package")= chr "adegenet"
>>
>>>
>>> # No error:
>>> as.vector(test.method at defined, "character")
>>>
>> [1] "gtypes"
>>
>>>
>>> # This is the method that generates the error
>>> load("problem.method.rdata")
>>>
>>> # Notice the difference in the environments of the two functions:
>>> test.method at .Data
>>>
>> function (x, ...)
>> nrow(x at loci)/x at ploidy
>>
>>> problem.method at .Data
>>>
>> function (x, ...)
>> nrow(x at loci)/x at ploidy
>> <environment: 0x101f85040>
>>
>>>
>>> # Swap the problem function for the one I created:
>>> problem.method at .Data <- test.method at .Data
>>> save(problem.method, file = "fixed.method.rdata")
>>>
>>> # Here's the error (both with my method and the original):
>>> as.vector(test.method at defined, "character")
>>>
>> Error: C stack usage 7970892 is too close to the limit
>>
>>> as.vector(problem.method at defined, "character")
>>>
>> Error: C stack usage 7970892 is too close to the limit
>>
>> Restarting R session...
>>
>> # *** Restart R for this to work ***
>>> rm(list = ls())
>>>
>>> load("fixed.method.rdata")
>>> as.vector(problem.method at defined, "character")
>>>
>> [1] "gtypes"
>>
>> ----
>>
>> *Eric Archer, Ph.D.*
>> Southwest Fisheries Science Center
>> NMFS, NOAA
>> 8901 La Jolla Shores Drive
>> La Jolla, CA 92037 USA
>> 858-546-7121 (work)
>> 858-546-7003 (FAX)
>>
>> Marine Mammal Genetics Group: swfsc.noaa.gov/mmtd-mmgenetics
>> ETP Cetacean Assessment Program: swfsc.noaa.gov/mmtd-etp
>> https://github/ericarcher
>>
>> "
>>
>>
>> *The universe doesn't care what you believe. The wonderful thing
about
>> science is that it doesn't ask for your faith, it just asks for
your
>> eyes.*" - Randall Munroe
>>
>> "*Lighthouses are more helpful than churches.*"
>> - Benjamin Franklin
>>
>> "*...but I'll take a GPS over either one.*"
>> - John C. "Craig" George
>>
>> [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> 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.
>>
>>
> --
> Luke Tierney
> Ralph E. Wareham Professor of Mathematical Sciences
> University of Iowa Phone: 319-335-3386
> Department of Statistics and Fax: 319-335-3017
> Actuarial Science
> 241 Schaeffer Hall email: luke-tierney at uiowa.edu
> Iowa City, IA 52242 WWW: http://www.stat.uiowa.edu
>
[[alternative HTML version deleted]]