In Non-Contradiction/autodiffr: Automatic Differentiation in R through 'Julia'
knitr::opts_chunk$set(echo = TRUE)
Introduction
autodiffr is an R package to exploit the Automatic Differentiation facilities
of the Julia computation system. Automatic Differentiation allows the computation of
derivatives of a function to the accuracy of analytic expressions. Moreover, it allows
the user to avoid the effort, difficulty and tedium of developing derivatives by hand.
Prerequisites
Whatever system you use, to use autodiffr you need to install Julia,
and indeed a (possibly the) correct version of Julia.
Go to https://julialang.org/downloads and choose the correct download
for your system, and the correct version (at the time of writing 0.6.4)
and install as appropriate. Note that the version of Julia is important,
and the version available in the package repository of some systems may
be inappropriate. Particular instructions are given in the sections below
for different OS families.
We also need a sufficiently up-to-date version of R, though at the time
of writing, the most recent lacks some tools on some operating systems.
For example, there is not yet an appropriate version of the Rtools software
for Windows 7 to accompany R-3.5.1 at 2018-7-27. This collection of software
is largely provided in most *nix systems, including Mac OS.
Linux
Installation on Linux has some dependencies that may
cause trouble. Some distros may not have the necessary
versions of libraries, and cannot run autodiffr.
We will try to suggest what these may be, but since there
are many distros, with variations on their library infrastructure,
this list will be haphazard.
Moreover, as with the version of Julia, we may need correct
versions of some system libraries.
We have not tested if a 64 bit distribution is required. That
is, 32 bit distros have not been tried.
We have verified that the following distros support autodiffr
- Linux Mint 18.3 Sylvia
- Ubuntu 18.04 (HWB? Did you test)
- Debian Stretch (9.5)
- Linux Mint 19 Mate "Tara" (To get
devtools to install, it was necessary to
install the Debian/Ubuntu "Bionic" packages libssl-dev and libcurl4-openssl-dev.
Installing R package codetools was done following a note when Rcpp was
installed as a devtools dependency. Then JuliaCall and autodiffr were installed
with install_github().)
The following distros do not appear to work:
- Bunsenlabs Deuterium 64 bit. (libstdc++.so.6 does not have GLIBC in a sufficiently recent version). This is based on Debian Jessie, and it is likely that distros in that family will not support
autodiffr.
- Elementary OS Freya 0.3.2 (same complaint about GLIBC)
Virtualbox special needs - Debian or Ubuntu based distros
If you install in virtualbox, you will want to install the Guest Additions.
In a running Virtualbox guest OS
you need to choose Devices / Insert Guest Additions CD
This will likely auto-mount (if not, click on it in a file manager GUI).
To install the guest additions, you will likely need to install the
build-essential
package (Debian and Ubuntu family). You will also need the right linux-headers.
Use the command uname -a to get the linux kernel verion, then install (again
Debian or Ubuntu) linux-headers-version where version is something like
4.10.0-38. It may be necessary to reboot here so these headers are recognized.
Certainly a new terminal will be needed.
Then either sudo the following commands or su - first.
cd /media/cdrom
sh ./VBoxLinuxAdditions.run
You can also set up a shared folder to allow for easy recording and transfer
of files between host and guest OS environments.
In Debian/Ubuntu systems, to install an up-to-date R, first get the signature key
for Johannes Ranke
apt-key adv --keyserver keys.gnupg.net --recv-key 'E19F5F87128899B192B1A2C2AD5F960A256A04AF'
NOTE: you may need to install the dirmngr package first in some distros.
In debian or ubuntu, as root,
apt install r-base-dev r-recommended
will install R.
To install Julia, in a web browser, go to https://julialang.org and download
the 0.6.4 version tarball.
Extract this in your home directory. Rename julia-xxxxx, where xxxxx is the
suffix to the name on your syste, to j.
Then as root
ln -s /home/(you)/j/bin/julia /usr/bin/julia
You can test julia with the command julia in a terminal window.
Getting autodiffr working
Start R in a terminal and run
install.packages("devtools")
On some systems, however, you may need to install some packages.
Do this in a separate terminal window. Note that the messages from
install.packages suggest the names of packages to install for major distros.
autodiffr needs a fairly recent version of the package JuliaCall (??ref by Changcheng).
It can be installed by
devtools::install_github("Non-Contradiction/JuliaCall")
Once this succeeds, autodiffr can be installed by:
devtools::install_github("Non-Contradiction/autodiffr")
After the installation is finished, we can run the package.
Note that the first time ad_setup() is run during a session,
it will take considerable time, but the next time it should be better.
?? Is this only slow the first execution overall, or does each
session start take the same time??
require(autodiffr)
ad_setup()
Error in juliacall_initialize(.julia$dll_file) :
C:\Users\john\AppData\Local\JULIA-~1.4\bin\libjulia.dll
- The specified procedure could not be found.
There is a msg box that gives the entry point that could not be found, with
a reference to libstdc++-6.dll
knitr::include_graphics("JuliaWin32ErrorBox.png")
Macintosh
On Macintosh (Hans W Borchers)
Any newer macOS operating system should be sufficient, at least
version 10.8; here we will use macOS "High Sierra" (that is 10.13.6).
We assume you have R version 3.4.1 or higher installed, but preferable
R >= 3.5.1 as the 3.5 series is significantly faster (on Mac), I saw
speed improvements of a factor of 2 and more in some of my
applications.
From the 'downloads' page of Julia,
https://julialang.org/downloads/, get the disk image file for Julia
version 0.6.4 (works also with 0.6.3) and install it on your Mac
computer in the usual way. You will find more platform-specific
instructions on https://julialang.org/downloads/platform.html but in
principle it is as easy as this.
If you did not install the Julia image in the /Applications folder,
you need to add the path to the Julia binary to your PATH variable or
-- what I found easier -- as superuser create a symbolic link in the
/usr/local/bin directory with
sudo ln -s path/to/julia-0.6.app/Contents/Resources/julia/bin/julia
/usr/local/bin/julia
To see if Julia is installed correctly, open a Terminal window and
type in the Julia command julia. You will see the usual prelude and
the Julia prompt.
_
_ _ _(_)_ | A fresh approach to technical computing
(_) | (_) (_) | Documentation: https://docs.julialang.org
_ _ _| |_ __ _ | Type "?help" for help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 0.6.4 (2018-07-09 19:09 UTC)
_/ |\__'_|_|_|\__'_| | Official http://julialang.org/ release
|__/ | x86_64-apple-darwin14.5.0
The autodiffr start-up procedure ad_setup() will install necessary
Julia packages. Because this can take quite a long time, pre-install
some of these packages directly from Julia, e.g. the important
RCall.jl package.
$ julia
julia> Pkg.add("RCall")
After some minutes about 30 or more packages are installed on your
computer. You can find them in your Home directory at
~/.julia/v0.6/. If you like, add also the forward and backward
automatic differentiation packages "ForwardDiff" and "ReverseDiff".
Some other depending packages will be installed, too.
Now it's time to install the R packages JuliaCall and autodiffr.
Both are available on CRAN, but it is better to get their newest
versions from Github, utilizing the devtools package.
devtools::install_github("Non-Contradiction/JuliaCall")
devtools::install_github("Non-Contradiction/autodiffr")
When loading the library autodiffr it is still necessary to create a
process running Julia, to connect to it, to check whether all needed
Julia packages are installed resp. to install them in case they are
missing. All this is done by the ad_setup function. If everything is
in place, it will finish in short time, otherwise it may take several
minutes (the first time it is called).
library(autodiffr)
## Attaching package: ‘autodiffr’
## The following object is masked from ‘package:stats’:
## deriv
ad_setup()
## Julia version 0.6.4 at location [...]/julia/bin will be used.
## Loading setup script for JuliaCall...
Finish loading setup script for JuliaCall.
INFO: Precompiling module DiffResults.
INFO: Precompiling module ForwardDiff.
INFO: Precompiling module ReverseDiff.
The first time ad_setup() is called on your system, these three
Julia modules / packages are being precompiled, later on this will not
happen again, except when the julia packages get installed anew (e.g.,
a new version has appeared). Please note that ad_setup is required
whenever you load the autodiffr package into R.
Now automatic differentiation is ready to get applied. As a test,
define an R function and take its gradient (or Jacobian or Hessian).
For instance, the following function will have a quite well known
derivative.
f <- function(x) exp(x)
exp(1.0); grad(f, 1.0)
## [1] 2.718281828459045
## [1] 2.718281828459045
We see the result is accurate to the last digit. Compare this with a
numerical derivative, such as is provided through the numDeriv
package.
numDeriv::grad(f, 1.0)
## [1] 2.718281828442632
Function names grad, jacobian, hessian are used in the
numDeriv (and pracma) package as well, so have to be qualified
when used in parallel with the autodiffr package.
IMPORTANT: Do not install the upcoming Julia 0.7 as there are
breaking changes between the 0.6 and 0.7/1.0 versions. After Julia 0.7
has become official and all relevant Julia packages have been updated
to the new version, autodiffr will be adapted and published on CRAN.
Windows
On some Windows systems, antivirus protection may need to be disabled
so the install can be accomplished. For example, though the ultimate
installation of autodiffr failed, on a Windows 7 32bit Virtual Machine
under VirtualBox, AVG antivirus (Free version) had to be disabled
(protection turned off) to enable the Julia install, which was to
C:\Users\john\AppData\Local\Julia-0.6.4. In the bin\ subdirectory, there
is julia.exe and this can be run in a terminal, or double-clicked to open.
Julia runs, and can be terminated by exit() or quit(). Remember to
re-enable antivirus protection once the installation of Julia (or other
necessary software) is complete.
Windows 10
Asus UX303 -- worked. What were steps?
Install Julia.
Install R
Need Rtools.
Install devtools and use it to install_github()
JuliaCall and autodiffr.
Note that PATH needs editing. (Search in Control Panel for
Environmental Variables to get the edit panel). Path for R
seems to be set automatically, but Julia path (to the "bin"
subdirectory) should be added. In my case this was
C:\Users\J\AppData\Local\Julia-0.6.4\bin
Windows 7 - 64 bit
Asus EEE 1225B
Installed Julia 0.6.4 and tested in CMD. OK
Installed R-3.5.1, devtools.
Tried to install_github("Non-contradiction/Julia_Call") -- FAILED
Needs Rtools. Installed Rtools35.exe. But when tried R said this version NOT compatible.
(In Win10 found Rtools34.exe worked. However, in Win10, was able to install autodiff and devtools
WITHOUT Rtools. Possibly don't need the underscore.)
Following is partial record of some things tried. Apologies for messiness.
JuliaCall::julia_command("@suppress begin function Base.float{V,D,O}(t::ReverseDiff.TrackedReal{V,D,O}) t end end")
apiFuncs()
.AD$inited <- TRUE
}
}
<environment: namespace:autodiffr>
> ad_setup()
Julia version 0.6.4 at location C:\Users\mary\AppData\Local\JULIA-~1.4\bin will be used.
Loading setup script for JuliaCall...
Warning message:
running command '"C:\Users\mary\AppData\Local\JULIA-~1.4\bin/julia" "C:/Users/mary/Documents/R/win-library/3.4/JuliaCall/julia/install_dependency.jl" "C:/PROGRA~1/R/R-34~1.4"' had status 5
> .AD$julia <- JuliaCall::julia_setup(...)
Error: '...' used in an incorrect context
> .AD$julia <- JuliaCall::julia_setup()
Julia version 0.6.4 at location C:\Users\mary\AppData\Local\JULIA-~1.4\bin will be used.
Loading setup script for JuliaCall...
Error in .julia$cmd(paste0("include(\"", system.file("julia/setup.jl", :
Error happens when you try to execute command include("C:/Users/mary/Documents/R/win-library/3.4/JuliaCall/julia/setup.jl") in Julia.
To have more helpful error messages,
you could considering running the command in Julia directly
In addition: Warning message:
running command '"C:\Users\mary\AppData\Local\JULIA-~1.4\bin/julia" "C:/Users/mary/Documents/R/win-library/3.4/JuliaCall/julia/install_dependency.jl" "C:/PROGRA~1/R/R-34~1.4"' had status 1
>
Run in Julia and try to build CodecZlib
Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.
C:\Users\mary>cd Documents
C:\Users\mary\Documents>cd R
C:\Users\mary\Documents\R>cd win-library
C:\Users\mary\Documents\R\win-library>cd 3.4
C:\Users\mary\Documents\R\win-library\3.4>cd JuliaCall
C:\Users\mary\Documents\R\win-library\3.4\JuliaCall>cd julia
C:\Users\mary\Documents\R\win-library\3.4\JuliaCall\julia>julia
_
_ _ _(_)_ | A fresh approach to technical computing
(_) | (_) (_) | Documentation: https://docs.julialang.org
_ _ _| |_ __ _ | Type "?help" for help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 0.6.4 (2018-07-09 19:09 UTC)
_/ |\__'_|_|_|\__'_| | Official http://julialang.org/ release
|__/ | x86_64-w64-mingw32
julia> ./setup.jl
ERROR: syntax: "./" is not a unary operator
julia> setup.jl
ERROR: UndefVarError: setup not defined
julia> source("setup.jl")
ERROR: UndefVarError: source not defined
julia> include("./setup.jl")
INFO: Precompiling module RCall.
ERROR: LoadError: CodecZlib.jl is not installed properly, run Pkg.build("CodecZl
ib") and restart Julia.
Stacktrace:
[1] error(::String) at .\error.jl:21
[2] include_from_node1(::String) at .\loading.jl:576
[3] include(::String) at .\sysimg.jl:14
[4] anonymous at .\<missing>:2
while loading C:\Users\mary\.julia\v0.6\CodecZlib\src\CodecZlib.jl, in expressio
n starting on line 38
ERROR: LoadError: LoadError: Failed to precompile CodecZlib to C:\Users\mary\.ju
lia\lib\v0.6\CodecZlib.ji.
Stacktrace:
[1] compilecache(::String) at .\loading.jl:710
[2] _require(::Symbol) at .\loading.jl:463
[3] require(::Symbol) at .\loading.jl:405
[4] include_from_node1(::String) at .\loading.jl:576
[5] include(::String) at .\sysimg.jl:14
[6] include_from_node1(::String) at .\loading.jl:576
[7] include(::String) at .\sysimg.jl:14
[8] anonymous at .\<missing>:2
while loading C:\Users\mary\.julia\v0.6\DataFrames\src\deprecated.jl, in express
ion starting on line 26
while loading C:\Users\mary\.julia\v0.6\DataFrames\src\DataFrames.jl, in express
ion starting on line 106
ERROR: LoadError: Failed to precompile DataFrames to C:\Users\mary\.julia\lib\v0
.6\DataFrames.ji.
Stacktrace:
[1] compilecache(::String) at .\loading.jl:710
[2] _require(::Symbol) at .\loading.jl:463
[3] require(::Symbol) at .\loading.jl:405
[4] include_from_node1(::String) at .\loading.jl:576
[5] include(::String) at .\sysimg.jl:14
[6] anonymous at .\<missing>:2
while loading C:\Users\mary\.julia\v0.6\RCall\src\RCall.jl, in expression starti
ng on line 8
ERROR: LoadError: Failed to precompile RCall to C:\Users\mary\.julia\lib\v0.6\RC
all.ji.
Stacktrace:
[1] compilecache(::String) at .\loading.jl:710
[2] _require(::Symbol) at .\loading.jl:497
[3] require(::Symbol) at .\loading.jl:405
[4] include_from_node1(::String) at .\loading.jl:576
[5] include(::String) at .\sysimg.jl:14
while loading C:\Users\mary\Documents\R\win-library\3.4\JuliaCall\julia\setup.jl
, in expression starting on line 19
julia>
julia> Pkg.build("CodecZlib")
INFO: Building CodecZlib
Info: Downloading https://github.com/bicycle1885/ZlibBuilder/releases/download/v
1.0.0/Zlib.x86_64-w64-mingw32.tar.gz to C:\Users\mary\.julia\v0.6\CodecZlib\deps
\usr\downloads\Zlib.x86_64-w64-mingw32.tar.gz...
Exception setting "SecurityProtocol": "Cannot convert null to type "System.Net.
SecurityProtocolType" due to invalid enumeration values. Specify one of the fol
lowing enumeration values and try again. The possible enumeration values are "S
sl3, Tls"."
At line:1 char:35
+ [System.Net.ServicePointManager]:: <<<< SecurityProtocol =
+ CategoryInfo : InvalidOperation: (:) [], RuntimeException
+ FullyQualifiedErrorId : PropertyAssignmentException
Exception calling "DownloadFile" with "2" argument(s): "The underlying connecti
on was closed: An unexpected error occurred on a send."
At line:4 char:24
+ $webclient.DownloadFile <<<< ("https://github.com/bicycle1885/ZlibBuilder/rel
eases/download/v1.0.0/Zlib.x86_64-w64-mingw32.tar.gz", "C:\Users\mary\.julia\v0
.6\CodecZlib\deps\usr\downloads\Zlib.x86_64-w64-mingw32.tar.gz")
+ CategoryInfo : NotSpecified: (:) [], MethodInvocationException
+ FullyQualifiedErrorId : DotNetMethodException
==============================[ ERROR: CodecZlib ]==============================
LoadError: Could not download https://github.com/bicycle1885/ZlibBuilder/release
s/download/v1.0.0/Zlib.x86_64-w64-mingw32.tar.gz to C:\Users\mary\.julia\v0.6\Co
decZlib\deps\usr\downloads\Zlib.x86_64-w64-mingw32.tar.gz
while loading C:\Users\mary\.julia\v0.6\CodecZlib\deps\build.jl, in expression s
tarting on line 92
================================================================================
================================[ BUILD ERRORS ]================================
WARNING: CodecZlib had build errors.
- packages with build errors remain installed in C:\Users\mary\.julia\v0.6
- build the package(s) and all dependencies with `Pkg.build("CodecZlib")`
- build a single package by running its `deps/build.jl` script
================================================================================
Windows 7 - 32 bit
Tried on VM, but did not get satisfactory outcome.
> ad_setup()
Julia version 0.6.4 at location C:\Users\john\AppData\Local\JULIA-~1.4\bin will be used.
Loading setup script for JuliaCall...
Error in .julia$cmd(paste0("include(\"", system.file("julia/setup.jl", :
Error happens when you try to execute command include("C:/Users/john/Documents/R/win-library/3.5/JuliaCall/julia/setup.jl") in Julia.
To have more helpful error messages,
you could considering running the command in Julia directly
In addition: Warning message:
In system2(file.path(.julia$bin_dir, "julia"), shQuote(command), :
running command '"C:\Users\john\AppData\Local\JULIA-~1.4\bin/julia" "C:/Users/john/Documents/R/win-library/3.5/JuliaCall/julia/install_dependency.jl" "C:/PROGRA~1/R/R-35~1.1"' had status 1
In Command window:
Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.
C:\Users\john>cd Documents
C:\Users\john\Documents>cd R
C:\Users\john\Documents\R>cd win-library
C:\Users\john\Documents\R\win-library>cd 3.5
C:\Users\john\Documents\R\win-library\3.5>cd JuliaCall
C:\Users\john\Documents\R\win-library\3.5\JuliaCall>cd julia
C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia>ls
JuliaObject.jl convert.jl incomplete_console.jl libjulia.jl
REPLhook.jl dispatch.jl install_dependency.jl setup.jl
asR.jl display interface1.jl
C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia>julia
_
_ _ _(_)_ | A fresh approach to technical computing
(_) | (_) (_) | Documentation: https://docs.julialang.org
_ _ _| |_ __ _ | Type "?help" for help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 0.6.4 (2018-07-09 19:09 UTC)
_/ |\__'_|_|_|\__'_| | Official http://julialang.org/ release
|__/ | i686-w64-mingw32
julia> include("setup.jl")
INFO: Precompiling module RCall.
ERROR: LoadError: RCall not properly installed. Please run Pkg.build("RCall")
Stacktrace:
[1] error at .\error.jl:21
[2] include_from_node1 at .\loading.jl:576
[3] include at .\sysimg.jl:14
while loading C:\Users\john\.julia\v0.6\RCall\src\RCall.jl, in expression starti
ng on line 40
ERROR: LoadError: Failed to precompile RCall to C:\Users\john\.julia\lib\v0.6\RC
all.ji.
Stacktrace:
[1] compilecache(::String) at .\loading.jl:710
[2] _require(::Symbol) at .\loading.jl:497
[3] require(::Symbol) at .\loading.jl:405
[4] include_from_node1 at .\loading.jl:576
[5] include at .\sysimg.jl:14
while loading C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\setup.jl
, in expression starting on line 19
julia> Pkg.build("RCall")
INFO: Building CodecZlib
INFO: Building SpecialFunctions
INFO: Building Conda
INFO: Building RCall
INFO: Using R at c:\Program Files\R\R-3.5.1 and libR at c:\Program Files\R\R-3.5
.1\bin\i386\R.dll.
julia> include("setup.jl")
WARNING: replacing module JuliaCall
INFO: Precompiling module RCall.
JuliaCall
julia>
-------------------------------
But running again in julia, after ad_setup() in R gives
julia> include("setup.jl")
WARNING: replacing module JuliaCall
WARNING: Method definition sexp(Tuple) in module JuliaCall at C:\Users\john\Docu
ments\R\win-library\3.5\JuliaCall\julia\convert.jl:8 overwritten in module Julia
Call at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\convert.jl:8.
WARNING: Method definition rcopy(Type{Tuple}, Ptr{RCall.VecSxp}) in module Julia
Call at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\convert.jl:14
overwritten in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\Jul
iaCall\julia\convert.jl:14.
WARNING: Method definition rcopytype(Type{RCall.RClass{:JuliaTuple}}, Ptr{RCall.
VecSxp}) in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\JuliaC
all\julia\convert.jl:17 overwritten in module JuliaCall at C:\Users\john\Documen
ts\R\win-library\3.5\JuliaCall\julia\convert.jl:17.
WARNING: Method definition rcopytype(Type{RCall.RClass{:JuliaObject}}, Ptr{RCall
.EnvSxp}) in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\Julia
Call\julia\JuliaObject.jl:92 overwritten in module JuliaCall at C:\Users\john\Do
cuments\R\win-library\3.5\JuliaCall\julia\JuliaObject.jl:92.
WARNING: Method definition sexp(Any) in module JuliaCall at C:\Users\john\Docume
nts\R\win-library\3.5\JuliaCall\julia\JuliaObject.jl:96 overwritten in module Ju
liaCall at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\JuliaObject
.jl:96.
WARNING: Method definition sexp(AbstractArray{Any, N} where N) in module JuliaCa
ll at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\JuliaObject.jl:1
05 overwritten in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\
JuliaCall\julia\JuliaObject.jl:105.
JuliaCall
julia>
Running in a fresh instance of julia gives the same "RCall not properly
installed" error from before.
Testing the autodiffr installation
And we can use package numDeriv to compare with autodiffr.
require(numDeriv)
extra material
Note that later we will probably need to adjust some or other PATH variables, and
possibly set up links to ensure R can find Julia.
To get Julia to start in R, C:\Users\john\AppData\Local\Julia-0.6.4\bin needs
to be in the PATH. Then a CMD box will start Julia with the command julia.
However, after require(autodiffr), the ad_setup() call gives an error
Julia version 0.6.4 at location
C:\Users\john\AppData\Local\JULIA-~1.4\bin will be used.
Error in juliacall_initialize(.julia$dll_file) :
C:\Users\john\AppData\Local\JULIA-~1.4\bin\libjulia.dll
- The specified procedure could not be found.
Thus for Windows, there must be another setting somewhere needed.
May be due to R 3.4.4??
then download Helium 64 and install, Guest Addiaions download
and try to install, todos, Allowed upgrade, added java support. Backports
from testing, leafpad, build-essential
sudo apt-get install linux-headers-$(uname -r) ## to get right headers
"Insert" VBoxGuestAdditions, cd /media/cdrom
sudo sh ./VBoxLinuxAdditions.run
Reboot.
Could not get larger. xrandr shows 1024 is largest screen. Also had difficulty
getting R 3.5.1 -- some key problem.
Set up Debian Stretch. installed build-essential, checked uname -a, then installed
linux-headers for correct version. Then su - and install VBoxLinuxAdditions.run.
Had to do 'xrandr --output VGA-1 --mode "1440x900" to get large display.
Installed leafpad. Added ... /debian stretch-cran35/ to sources.list
Note: needed to install dirmngr. But stretch-cran35 does not have a Release file.
Turned out it was "cdrom" that was problem. -- just comment it out in sources.list
Downloaded julia 0.6.4 from julialang.org, unpacked in ~, renamed julia-xxxxx to j
As root set symlink in /usr/bin to j/bin/julia. Tested it works.
start R. install("devtools"), but needed git2r, httr (which needed system libssl-dev)
and libcurl-openssl-dev. devtools in.
require(devtools) / install_github("Non-contradiction/JuliaCall")
install_github("Non-contradiction/autodiffr")
require(autodiffr) / ad_setup() ## takes a very long time and disk is flashing
## Is it trying to locate a library file??
Eventually finished (20 mins?)
Then test <- function(x) { sum(x*x) }
testg <- makeGradFunc(test)
x0 <- c(1,3,5)
test(x0) ## 35
testg(x0) ## 2 6 10
Later re-installed mintupdate and virtualbox-guest-dkms, ...x11 and one other removed,
then screen sized to max automatically.
?? need to check path
?? symlink
Julia.exe is in C:\Users\john\AppData\Local\Julia-0.6.4 after installation
( 32 bit -- the extra steps were not necessary as far as I can tell)
C:\Users\john\AppData\Local\Julia-0.6.4\bin was added to the PATH.
CMD box then runs julia fine.
Non-Contradiction/autodiffr documentation built on May 10, 2019, 8:04 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
knitr::opts_chunk$set(echo = TRUE)
Introduction
autodiffr is an R package to exploit the Automatic Differentiation facilities
of the Julia computation system. Automatic Differentiation allows the computation of
derivatives of a function to the accuracy of analytic expressions. Moreover, it allows
the user to avoid the effort, difficulty and tedium of developing derivatives by hand.
Prerequisites
Whatever system you use, to use autodiffr you need to install Julia,
and indeed a (possibly the) correct version of Julia.
Go to https://julialang.org/downloads and choose the correct download for your system, and the correct version (at the time of writing 0.6.4) and install as appropriate. Note that the version of Julia is important, and the version available in the package repository of some systems may be inappropriate. Particular instructions are given in the sections below for different OS families.
We also need a sufficiently up-to-date version of R, though at the time
of writing, the most recent lacks some tools on some operating systems.
For example, there is not yet an appropriate version of the Rtools software
for Windows 7 to accompany R-3.5.1 at 2018-7-27. This collection of software
is largely provided in most *nix systems, including Mac OS.
Linux
Installation on Linux has some dependencies that may
cause trouble. Some distros may not have the necessary
versions of libraries, and cannot run autodiffr.
We will try to suggest what these may be, but since there
are many distros, with variations on their library infrastructure,
this list will be haphazard.
Moreover, as with the version of Julia, we may need correct
versions of some system libraries.
We have not tested if a 64 bit distribution is required. That is, 32 bit distros have not been tried.
We have verified that the following distros support autodiffr
- Linux Mint 18.3 Sylvia
- Ubuntu 18.04 (HWB? Did you test)
- Debian Stretch (9.5)
- Linux Mint 19 Mate "Tara" (To get
devtoolsto install, it was necessary to install the Debian/Ubuntu "Bionic" packages libssl-dev and libcurl4-openssl-dev. Installing R packagecodetoolswas done following a note when Rcpp was installed as a devtools dependency. Then JuliaCall and autodiffr were installed with install_github().)
The following distros do not appear to work:
- Bunsenlabs Deuterium 64 bit. (libstdc++.so.6 does not have GLIBC in a sufficiently recent version). This is based on Debian Jessie, and it is likely that distros in that family will not support
autodiffr. - Elementary OS Freya 0.3.2 (same complaint about GLIBC)
Virtualbox special needs - Debian or Ubuntu based distros
If you install in virtualbox, you will want to install the Guest Additions.
In a running Virtualbox guest OS
you need to choose Devices / Insert Guest Additions CD
This will likely auto-mount (if not, click on it in a file manager GUI).
To install the guest additions, you will likely need to install the
build-essential
package (Debian and Ubuntu family). You will also need the right linux-headers.
Use the command uname -a to get the linux kernel verion, then install (again
Debian or Ubuntu) linux-headers-version where version is something like
4.10.0-38. It may be necessary to reboot here so these headers are recognized.
Certainly a new terminal will be needed.
Then either sudo the following commands or su - first.
cd /media/cdrom sh ./VBoxLinuxAdditions.run
You can also set up a shared folder to allow for easy recording and transfer of files between host and guest OS environments.
In Debian/Ubuntu systems, to install an up-to-date R, first get the signature key for Johannes Ranke
apt-key adv --keyserver keys.gnupg.net --recv-key 'E19F5F87128899B192B1A2C2AD5F960A256A04AF'
NOTE: you may need to install the dirmngr package first in some distros.
In debian or ubuntu, as root,
apt install r-base-dev r-recommended
will install R.
To install Julia, in a web browser, go to https://julialang.org and download the 0.6.4 version tarball.
Extract this in your home directory. Rename julia-xxxxx, where xxxxx is the
suffix to the name on your syste, to j.
Then as root
ln -s /home/(you)/j/bin/julia /usr/bin/julia
You can test julia with the command julia in a terminal window.
Getting autodiffr working
Start R in a terminal and run
install.packages("devtools")
On some systems, however, you may need to install some packages.
Do this in a separate terminal window. Note that the messages from
install.packages suggest the names of packages to install for major distros.
autodiffr needs a fairly recent version of the package JuliaCall (??ref by Changcheng).
It can be installed by
devtools::install_github("Non-Contradiction/JuliaCall")
Once this succeeds, autodiffr can be installed by:
devtools::install_github("Non-Contradiction/autodiffr")
After the installation is finished, we can run the package.
Note that the first time ad_setup() is run during a session,
it will take considerable time, but the next time it should be better.
?? Is this only slow the first execution overall, or does each
session start take the same time??
require(autodiffr) ad_setup()
Error in juliacall_initialize(.julia$dll_file) : C:\Users\john\AppData\Local\JULIA-~1.4\bin\libjulia.dll - The specified procedure could not be found.
There is a msg box that gives the entry point that could not be found, with a reference to libstdc++-6.dll
knitr::include_graphics("JuliaWin32ErrorBox.png")
Macintosh
On Macintosh (Hans W Borchers)
Any newer macOS operating system should be sufficient, at least version 10.8; here we will use macOS "High Sierra" (that is 10.13.6). We assume you have R version 3.4.1 or higher installed, but preferable R >= 3.5.1 as the 3.5 series is significantly faster (on Mac), I saw speed improvements of a factor of 2 and more in some of my applications.
From the 'downloads' page of Julia, https://julialang.org/downloads/, get the disk image file for Julia version 0.6.4 (works also with 0.6.3) and install it on your Mac computer in the usual way. You will find more platform-specific instructions on https://julialang.org/downloads/platform.html but in principle it is as easy as this.
If you did not install the Julia image in the /Applications folder,
you need to add the path to the Julia binary to your PATH variable or
-- what I found easier -- as superuser create a symbolic link in the
/usr/local/bin directory with
sudo ln -s path/to/julia-0.6.app/Contents/Resources/julia/bin/julia
/usr/local/bin/julia
To see if Julia is installed correctly, open a Terminal window and
type in the Julia command julia. You will see the usual prelude and
the Julia prompt.
_ _ _ _(_)_ | A fresh approach to technical computing (_) | (_) (_) | Documentation: https://docs.julialang.org _ _ _| |_ __ _ | Type "?help" for help. | | | | | | |/ _` | | | | |_| | | | (_| | | Version 0.6.4 (2018-07-09 19:09 UTC) _/ |\__'_|_|_|\__'_| | Official http://julialang.org/ release |__/ | x86_64-apple-darwin14.5.0
The autodiffr start-up procedure ad_setup() will install necessary
Julia packages. Because this can take quite a long time, pre-install
some of these packages directly from Julia, e.g. the important
RCall.jl package.
$ julia julia> Pkg.add("RCall")
After some minutes about 30 or more packages are installed on your
computer. You can find them in your Home directory at
~/.julia/v0.6/. If you like, add also the forward and backward
automatic differentiation packages "ForwardDiff" and "ReverseDiff".
Some other depending packages will be installed, too.
Now it's time to install the R packages JuliaCall and autodiffr. Both are available on CRAN, but it is better to get their newest versions from Github, utilizing the devtools package.
devtools::install_github("Non-Contradiction/JuliaCall")
devtools::install_github("Non-Contradiction/autodiffr")
When loading the library autodiffr it is still necessary to create a
process running Julia, to connect to it, to check whether all needed
Julia packages are installed resp. to install them in case they are
missing. All this is done by the ad_setup function. If everything is
in place, it will finish in short time, otherwise it may take several
minutes (the first time it is called).
library(autodiffr) ## Attaching package: ‘autodiffr’ ## The following object is masked from ‘package:stats’: ## deriv ad_setup() ## Julia version 0.6.4 at location [...]/julia/bin will be used. ## Loading setup script for JuliaCall... Finish loading setup script for JuliaCall. INFO: Precompiling module DiffResults. INFO: Precompiling module ForwardDiff. INFO: Precompiling module ReverseDiff.
The first time ad_setup() is called on your system, these three
Julia modules / packages are being precompiled, later on this will not
happen again, except when the julia packages get installed anew (e.g.,
a new version has appeared). Please note that ad_setup is required
whenever you load the autodiffr package into R.
Now automatic differentiation is ready to get applied. As a test, define an R function and take its gradient (or Jacobian or Hessian). For instance, the following function will have a quite well known derivative.
f <- function(x) exp(x) exp(1.0); grad(f, 1.0) ## [1] 2.718281828459045 ## [1] 2.718281828459045
We see the result is accurate to the last digit. Compare this with a numerical derivative, such as is provided through the numDeriv package.
numDeriv::grad(f, 1.0) ## [1] 2.718281828442632
Function names grad, jacobian, hessian are used in the
numDeriv (and pracma) package as well, so have to be qualified
when used in parallel with the autodiffr package.
IMPORTANT: Do not install the upcoming Julia 0.7 as there are breaking changes between the 0.6 and 0.7/1.0 versions. After Julia 0.7 has become official and all relevant Julia packages have been updated to the new version, autodiffr will be adapted and published on CRAN.
Windows
On some Windows systems, antivirus protection may need to be disabled
so the install can be accomplished. For example, though the ultimate
installation of autodiffr failed, on a Windows 7 32bit Virtual Machine
under VirtualBox, AVG antivirus (Free version) had to be disabled
(protection turned off) to enable the Julia install, which was to
C:\Users\john\AppData\Local\Julia-0.6.4. In the bin\ subdirectory, there
is julia.exe and this can be run in a terminal, or double-clicked to open.
Julia runs, and can be terminated by exit() or quit(). Remember to
re-enable antivirus protection once the installation of Julia (or other
necessary software) is complete.
Windows 10
Asus UX303 -- worked. What were steps?
Install Julia.
Install R
Need Rtools.
Install devtools and use it to install_github()
JuliaCall and autodiffr.
Note that PATH needs editing. (Search in Control Panel for Environmental Variables to get the edit panel). Path for R seems to be set automatically, but Julia path (to the "bin" subdirectory) should be added. In my case this was
C:\Users\J\AppData\Local\Julia-0.6.4\bin
Windows 7 - 64 bit
Asus EEE 1225B
Installed Julia 0.6.4 and tested in CMD. OK
Installed R-3.5.1, devtools.
Tried to install_github("Non-contradiction/Julia_Call") -- FAILED
Needs Rtools. Installed Rtools35.exe. But when tried R said this version NOT compatible. (In Win10 found Rtools34.exe worked. However, in Win10, was able to install autodiff and devtools WITHOUT Rtools. Possibly don't need the underscore.)
Following is partial record of some things tried. Apologies for messiness.
JuliaCall::julia_command("@suppress begin function Base.float{V,D,O}(t::ReverseDiff.TrackedReal{V,D,O}) t end end") apiFuncs() .AD$inited <- TRUE } } <environment: namespace:autodiffr> > ad_setup() Julia version 0.6.4 at location C:\Users\mary\AppData\Local\JULIA-~1.4\bin will be used. Loading setup script for JuliaCall... Warning message: running command '"C:\Users\mary\AppData\Local\JULIA-~1.4\bin/julia" "C:/Users/mary/Documents/R/win-library/3.4/JuliaCall/julia/install_dependency.jl" "C:/PROGRA~1/R/R-34~1.4"' had status 5 > .AD$julia <- JuliaCall::julia_setup(...) Error: '...' used in an incorrect context > .AD$julia <- JuliaCall::julia_setup() Julia version 0.6.4 at location C:\Users\mary\AppData\Local\JULIA-~1.4\bin will be used. Loading setup script for JuliaCall... Error in .julia$cmd(paste0("include(\"", system.file("julia/setup.jl", : Error happens when you try to execute command include("C:/Users/mary/Documents/R/win-library/3.4/JuliaCall/julia/setup.jl") in Julia. To have more helpful error messages, you could considering running the command in Julia directly In addition: Warning message: running command '"C:\Users\mary\AppData\Local\JULIA-~1.4\bin/julia" "C:/Users/mary/Documents/R/win-library/3.4/JuliaCall/julia/install_dependency.jl" "C:/PROGRA~1/R/R-34~1.4"' had status 1 > Run in Julia and try to build CodecZlib Microsoft Windows [Version 6.1.7601] Copyright (c) 2009 Microsoft Corporation. All rights reserved. C:\Users\mary>cd Documents C:\Users\mary\Documents>cd R C:\Users\mary\Documents\R>cd win-library C:\Users\mary\Documents\R\win-library>cd 3.4 C:\Users\mary\Documents\R\win-library\3.4>cd JuliaCall C:\Users\mary\Documents\R\win-library\3.4\JuliaCall>cd julia C:\Users\mary\Documents\R\win-library\3.4\JuliaCall\julia>julia _ _ _ _(_)_ | A fresh approach to technical computing (_) | (_) (_) | Documentation: https://docs.julialang.org _ _ _| |_ __ _ | Type "?help" for help. | | | | | | |/ _` | | | | |_| | | | (_| | | Version 0.6.4 (2018-07-09 19:09 UTC) _/ |\__'_|_|_|\__'_| | Official http://julialang.org/ release |__/ | x86_64-w64-mingw32 julia> ./setup.jl ERROR: syntax: "./" is not a unary operator julia> setup.jl ERROR: UndefVarError: setup not defined julia> source("setup.jl") ERROR: UndefVarError: source not defined julia> include("./setup.jl") INFO: Precompiling module RCall. ERROR: LoadError: CodecZlib.jl is not installed properly, run Pkg.build("CodecZl ib") and restart Julia. Stacktrace: [1] error(::String) at .\error.jl:21 [2] include_from_node1(::String) at .\loading.jl:576 [3] include(::String) at .\sysimg.jl:14 [4] anonymous at .\<missing>:2 while loading C:\Users\mary\.julia\v0.6\CodecZlib\src\CodecZlib.jl, in expressio n starting on line 38 ERROR: LoadError: LoadError: Failed to precompile CodecZlib to C:\Users\mary\.ju lia\lib\v0.6\CodecZlib.ji. Stacktrace: [1] compilecache(::String) at .\loading.jl:710 [2] _require(::Symbol) at .\loading.jl:463 [3] require(::Symbol) at .\loading.jl:405 [4] include_from_node1(::String) at .\loading.jl:576 [5] include(::String) at .\sysimg.jl:14 [6] include_from_node1(::String) at .\loading.jl:576 [7] include(::String) at .\sysimg.jl:14 [8] anonymous at .\<missing>:2 while loading C:\Users\mary\.julia\v0.6\DataFrames\src\deprecated.jl, in express ion starting on line 26 while loading C:\Users\mary\.julia\v0.6\DataFrames\src\DataFrames.jl, in express ion starting on line 106 ERROR: LoadError: Failed to precompile DataFrames to C:\Users\mary\.julia\lib\v0 .6\DataFrames.ji. Stacktrace: [1] compilecache(::String) at .\loading.jl:710 [2] _require(::Symbol) at .\loading.jl:463 [3] require(::Symbol) at .\loading.jl:405 [4] include_from_node1(::String) at .\loading.jl:576 [5] include(::String) at .\sysimg.jl:14 [6] anonymous at .\<missing>:2 while loading C:\Users\mary\.julia\v0.6\RCall\src\RCall.jl, in expression starti ng on line 8 ERROR: LoadError: Failed to precompile RCall to C:\Users\mary\.julia\lib\v0.6\RC all.ji. Stacktrace: [1] compilecache(::String) at .\loading.jl:710 [2] _require(::Symbol) at .\loading.jl:497 [3] require(::Symbol) at .\loading.jl:405 [4] include_from_node1(::String) at .\loading.jl:576 [5] include(::String) at .\sysimg.jl:14 while loading C:\Users\mary\Documents\R\win-library\3.4\JuliaCall\julia\setup.jl , in expression starting on line 19 julia> julia> Pkg.build("CodecZlib") INFO: Building CodecZlib Info: Downloading https://github.com/bicycle1885/ZlibBuilder/releases/download/v 1.0.0/Zlib.x86_64-w64-mingw32.tar.gz to C:\Users\mary\.julia\v0.6\CodecZlib\deps \usr\downloads\Zlib.x86_64-w64-mingw32.tar.gz... Exception setting "SecurityProtocol": "Cannot convert null to type "System.Net. SecurityProtocolType" due to invalid enumeration values. Specify one of the fol lowing enumeration values and try again. The possible enumeration values are "S sl3, Tls"." At line:1 char:35 + [System.Net.ServicePointManager]:: <<<< SecurityProtocol = + CategoryInfo : InvalidOperation: (:) [], RuntimeException + FullyQualifiedErrorId : PropertyAssignmentException Exception calling "DownloadFile" with "2" argument(s): "The underlying connecti on was closed: An unexpected error occurred on a send." At line:4 char:24 + $webclient.DownloadFile <<<< ("https://github.com/bicycle1885/ZlibBuilder/rel eases/download/v1.0.0/Zlib.x86_64-w64-mingw32.tar.gz", "C:\Users\mary\.julia\v0 .6\CodecZlib\deps\usr\downloads\Zlib.x86_64-w64-mingw32.tar.gz") + CategoryInfo : NotSpecified: (:) [], MethodInvocationException + FullyQualifiedErrorId : DotNetMethodException ==============================[ ERROR: CodecZlib ]============================== LoadError: Could not download https://github.com/bicycle1885/ZlibBuilder/release s/download/v1.0.0/Zlib.x86_64-w64-mingw32.tar.gz to C:\Users\mary\.julia\v0.6\Co decZlib\deps\usr\downloads\Zlib.x86_64-w64-mingw32.tar.gz while loading C:\Users\mary\.julia\v0.6\CodecZlib\deps\build.jl, in expression s tarting on line 92 ================================================================================ ================================[ BUILD ERRORS ]================================ WARNING: CodecZlib had build errors. - packages with build errors remain installed in C:\Users\mary\.julia\v0.6 - build the package(s) and all dependencies with `Pkg.build("CodecZlib")` - build a single package by running its `deps/build.jl` script ================================================================================
Windows 7 - 32 bit
Tried on VM, but did not get satisfactory outcome.
> ad_setup()
Julia version 0.6.4 at location C:\Users\john\AppData\Local\JULIA-~1.4\bin will be used.
Loading setup script for JuliaCall...
Error in .julia$cmd(paste0("include(\"", system.file("julia/setup.jl", :
Error happens when you try to execute command include("C:/Users/john/Documents/R/win-library/3.5/JuliaCall/julia/setup.jl") in Julia.
To have more helpful error messages,
you could considering running the command in Julia directly
In addition: Warning message:
In system2(file.path(.julia$bin_dir, "julia"), shQuote(command), :
running command '"C:\Users\john\AppData\Local\JULIA-~1.4\bin/julia" "C:/Users/john/Documents/R/win-library/3.5/JuliaCall/julia/install_dependency.jl" "C:/PROGRA~1/R/R-35~1.1"' had status 1
In Command window:
Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.
C:\Users\john>cd Documents
C:\Users\john\Documents>cd R
C:\Users\john\Documents\R>cd win-library
C:\Users\john\Documents\R\win-library>cd 3.5
C:\Users\john\Documents\R\win-library\3.5>cd JuliaCall
C:\Users\john\Documents\R\win-library\3.5\JuliaCall>cd julia
C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia>ls
JuliaObject.jl convert.jl incomplete_console.jl libjulia.jl
REPLhook.jl dispatch.jl install_dependency.jl setup.jl
asR.jl display interface1.jl
C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia>julia
_
_ _ _(_)_ | A fresh approach to technical computing
(_) | (_) (_) | Documentation: https://docs.julialang.org
_ _ _| |_ __ _ | Type "?help" for help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 0.6.4 (2018-07-09 19:09 UTC)
_/ |\__'_|_|_|\__'_| | Official http://julialang.org/ release
|__/ | i686-w64-mingw32
julia> include("setup.jl")
INFO: Precompiling module RCall.
ERROR: LoadError: RCall not properly installed. Please run Pkg.build("RCall")
Stacktrace:
[1] error at .\error.jl:21
[2] include_from_node1 at .\loading.jl:576
[3] include at .\sysimg.jl:14
while loading C:\Users\john\.julia\v0.6\RCall\src\RCall.jl, in expression starti
ng on line 40
ERROR: LoadError: Failed to precompile RCall to C:\Users\john\.julia\lib\v0.6\RC
all.ji.
Stacktrace:
[1] compilecache(::String) at .\loading.jl:710
[2] _require(::Symbol) at .\loading.jl:497
[3] require(::Symbol) at .\loading.jl:405
[4] include_from_node1 at .\loading.jl:576
[5] include at .\sysimg.jl:14
while loading C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\setup.jl
, in expression starting on line 19
julia> Pkg.build("RCall")
INFO: Building CodecZlib
INFO: Building SpecialFunctions
INFO: Building Conda
INFO: Building RCall
INFO: Using R at c:\Program Files\R\R-3.5.1 and libR at c:\Program Files\R\R-3.5
.1\bin\i386\R.dll.
julia> include("setup.jl")
WARNING: replacing module JuliaCall
INFO: Precompiling module RCall.
JuliaCall
julia>
-------------------------------
But running again in julia, after ad_setup() in R gives
julia> include("setup.jl")
WARNING: replacing module JuliaCall
WARNING: Method definition sexp(Tuple) in module JuliaCall at C:\Users\john\Docu
ments\R\win-library\3.5\JuliaCall\julia\convert.jl:8 overwritten in module Julia
Call at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\convert.jl:8.
WARNING: Method definition rcopy(Type{Tuple}, Ptr{RCall.VecSxp}) in module Julia
Call at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\convert.jl:14
overwritten in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\Jul
iaCall\julia\convert.jl:14.
WARNING: Method definition rcopytype(Type{RCall.RClass{:JuliaTuple}}, Ptr{RCall.
VecSxp}) in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\JuliaC
all\julia\convert.jl:17 overwritten in module JuliaCall at C:\Users\john\Documen
ts\R\win-library\3.5\JuliaCall\julia\convert.jl:17.
WARNING: Method definition rcopytype(Type{RCall.RClass{:JuliaObject}}, Ptr{RCall
.EnvSxp}) in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\Julia
Call\julia\JuliaObject.jl:92 overwritten in module JuliaCall at C:\Users\john\Do
cuments\R\win-library\3.5\JuliaCall\julia\JuliaObject.jl:92.
WARNING: Method definition sexp(Any) in module JuliaCall at C:\Users\john\Docume
nts\R\win-library\3.5\JuliaCall\julia\JuliaObject.jl:96 overwritten in module Ju
liaCall at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\JuliaObject
.jl:96.
WARNING: Method definition sexp(AbstractArray{Any, N} where N) in module JuliaCa
ll at C:\Users\john\Documents\R\win-library\3.5\JuliaCall\julia\JuliaObject.jl:1
05 overwritten in module JuliaCall at C:\Users\john\Documents\R\win-library\3.5\
JuliaCall\julia\JuliaObject.jl:105.
JuliaCall
julia>
Running in a fresh instance of julia gives the same "RCall not properly installed" error from before.
Testing the autodiffr installation
And we can use package numDeriv to compare with autodiffr.
require(numDeriv)
extra material
Note that later we will probably need to adjust some or other PATH variables, and possibly set up links to ensure R can find Julia.
To get Julia to start in R, C:\Users\john\AppData\Local\Julia-0.6.4\bin needs
to be in the PATH. Then a CMD box will start Julia with the command julia.
However, after require(autodiffr), the ad_setup() call gives an error
Julia version 0.6.4 at location C:\Users\john\AppData\Local\JULIA-~1.4\bin will be used. Error in juliacall_initialize(.julia$dll_file) : C:\Users\john\AppData\Local\JULIA-~1.4\bin\libjulia.dll - The specified procedure could not be found.
Thus for Windows, there must be another setting somewhere needed.
May be due to R 3.4.4??
then download Helium 64 and install, Guest Addiaions download and try to install, todos, Allowed upgrade, added java support. Backports from testing, leafpad, build-essential sudo apt-get install linux-headers-$(uname -r) ## to get right headers "Insert" VBoxGuestAdditions, cd /media/cdrom sudo sh ./VBoxLinuxAdditions.run Reboot. Could not get larger. xrandr shows 1024 is largest screen. Also had difficulty getting R 3.5.1 -- some key problem. Set up Debian Stretch. installed build-essential, checked uname -a, then installed linux-headers for correct version. Then su - and install VBoxLinuxAdditions.run. Had to do 'xrandr --output VGA-1 --mode "1440x900" to get large display. Installed leafpad. Added ... /debian stretch-cran35/ to sources.list Note: needed to install dirmngr. But stretch-cran35 does not have a Release file. Turned out it was "cdrom" that was problem. -- just comment it out in sources.list Downloaded julia 0.6.4 from julialang.org, unpacked in ~, renamed julia-xxxxx to j As root set symlink in /usr/bin to j/bin/julia. Tested it works. start R. install("devtools"), but needed git2r, httr (which needed system libssl-dev) and libcurl-openssl-dev. devtools in. require(devtools) / install_github("Non-contradiction/JuliaCall") install_github("Non-contradiction/autodiffr") require(autodiffr) / ad_setup() ## takes a very long time and disk is flashing ## Is it trying to locate a library file?? Eventually finished (20 mins?) Then test <- function(x) { sum(x*x) } testg <- makeGradFunc(test) x0 <- c(1,3,5) test(x0) ## 35 testg(x0) ## 2 6 10
Later re-installed mintupdate and virtualbox-guest-dkms, ...x11 and one other removed, then screen sized to max automatically.
?? need to check path
?? symlink
Julia.exe is in C:\Users\john\AppData\Local\Julia-0.6.4 after installation
( 32 bit -- the extra steps were not necessary as far as I can tell)
C:\Users\john\AppData\Local\Julia-0.6.4\bin was added to the PATH.
CMD box then runs julia fine.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.