Thank you for your help. I tried experimenting with LOAD_PATH a bit before
I made this post and I couldn't get it to work. Here's an example - I have
previously using Pkg.add() to add PyPlot to ~/.julia/gcc. However, if I try
adding that directory to LOAD_PATH (via environment variables) then I am
not able to use PyPlot:
[username@nrlogin1 01/15 09:38:08 ~]$ export JULIA_LOAD_PATH="/home/username
/.julia/gcc"
[username@nrlogin1 01/15 09:38:13 ~]$ ~/apps/gcc/5.2.0/julia/0.4.2/bin/julia
_
_ _ _(_)_ | A fresh approach to technical computing
(_) | (_) (_) | Documentation: http://docs.julialang.org
_ _ _| |_ __ _ | Type "?help" for help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 0.4.2 (2015-12-06 21:47 UTC)
_/ |\__'_|_|_|\__'_| |
|__/ | x86_64-unknown-linux-gnu
julia> LOAD_PATH
3-element Array{ByteString,1}:
"/home/username/.julia/gcc"
"/home/username/apps/gcc/5.2.0/julia/0.4.2/local/share/julia/site/v0.4"
"/home/username/apps/gcc/5.2.0/julia/0.4.2/share/julia/site/v0.4"
julia> using PyPlot
ERROR: ArgumentError: PyPlot not found in path
in require at ./loading.jl:233
julia> exit()
By contrast (mostly to show that PyPlot is indeed in that directory),
PyPlot can be loaded and used if I set JULIA_PKGDIR to point to this
directory:
[username@nrlogin1 01/15 09:38:22 ~]$ export JULIA_PKGDIR="/home/username
/.julia/gcc"
[username@nrlogin1 01/15 09:38:33 ~]$ ~/apps/gcc/5.2.0/julia/0.4.2/bin/julia
_
_ _ _(_)_ | A fresh approach to technical computing
(_) | (_) (_) | Documentation: http://docs.julialang.org
_ _ _| |_ __ _ | Type "?help" for help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 0.4.2 (2015-12-06 21:47 UTC)
_/ |\__'_|_|_|\__'_| |
|__/ | x86_64-unknown-linux-gnu
julia> using PyPlot
julia> plot(1:100,sort(rand(100,)))
1-element Array{Any,1}:
PyObject <matplotlib.lines.Line2D object at 0x7fa981029390>
julia> close("all")
julia> exit()
The latter obviously won't work for what I want to do since I want users to
be able to have their own PKGDIR where they can add packages that they
need. Am I doing something incorrectly?
Side note: If I do end up getting LOAD_PATH to work as you say, I suppose
installing the global packages to one of the directories prepopulated into
LOAD_PATH would be the "best practice"?
"/home/username/apps/gcc/5.2.0/julia/0.4.2/local/share/julia/site/v0.4"
"/home/username/apps/gcc/5.2.0/julia/0.4.2/share/julia/site/v0.4"
On Thursday, January 14, 2016 at 10:34:14 PM UTC-5, Yichao Yu wrote:
>
> On Thu, Jan 14, 2016 at 10:26 PM, Tony Kelman <[email protected]
> <javascript:>> wrote:
> > You can add system-wide packages to LOAD_PATH while keeping the local
> > Pkg.dir (maybe modify the JULIA_PKGDIR environment variable for the
> initial
> > installation of the system-wide packages). Not all packages work when
> used
> > from outside Pkg.dir, and Pkg operations often don't see them, but you
> can
> > try and see what works vs what doesn't.
>
> Most packages I've tested works pretty well. (Just don't expect
> `Pkg.*` to work). It's also usually not very hard to fix if it doesn't
> work.
>
> >
> >
> > On Thursday, January 14, 2016 at 9:11:20 AM UTC-8, [email protected]
> wrote:
> >>
> >> I work on a large compute cluster used by hundreds of university
> >> researchers. Our scientific applications stack is managed via RPM. I
> would
> >> like to provide users with a Julia build that:
> >> (1) Includes Julia base
> >> (2) Provides some commonly-used secondary packages - the main ones that
> I
> >> have in mind right now are the plotting packages Gadfly and PyPlot and
> >> perhaps also IJulia. But there may be others down the line. I want to
> >> provide these centrally because many, many users will want to use them
> and I
> >> want to lower the barrier to entry. I also would like to figure out the
> >> matplotlib dependencies in PyPlot for them so that they don't have to.
> >>
> >> The idea is to provide a base install that will be enough for many
> users
> >> (at least to get started); they can then use Pkg to add any other
> packages
> >> that they require.
> >>
> >> Is there a recommended method for centrally-installing packages like in
> >> (2) while also allowing users to maintain additional packages in their
> Home
> >> directories? I have not been able to track down documentation for this.
> >>
> >
>
