julia packages

jump to:

environments and adding packages
using packages, with data frames & plotting example
package RCall and the R mode
Pluto notebooks
other great packages

environments

Julia’s package manager is centered around environments. By default, we use the v1.5 environment, whose information is in ~/.julia/environments/v1.5/

shell> ls ~/.julia/environments/v1.5/
Manifest.toml	Project.toml
shell> cat ~/.julia/environments/v1.5/Project.toml   # what I already installed
shell> less ~/.julia/environments/v1.5/Manifest.toml # detailed version info

The file Project.toml lists which packages (or julia version) need to be used, Manifest.toml lists which exact versions. I will create a new environment for myself for this class, which can be useful if we want to develop a new package separately from an environment where I might want to use Julia as a “simple user”.

I created a new directory julia within my directory for the class. From this julia/, start julia, enter the “pkg” mode by typing ], then do this to install the MixedModels package:

(@v1.5) pkg> activate .
Activating new environment at `~/Documents/private/st679/julia/Project.toml`

(julia) pkg> status
Status `~/Documents/private/st679/julia/Project.toml` (empty project)

(julia) pkg> add MixedModels
   Updating registry at `~/.julia/registries/General`
   Updating git-repo `https://github.com/JuliaRegistries/General.git`
  Resolving package versions...
  ...
  Installed MixedModels ──────── v3.1.0
  ...
Updating `~/Documents/private/st679/julia/Project.toml`
  [ff71e718] + MixedModels v3.1.0
Updating `~/Documents/private/st679/julia/Manifest.toml`
  [4c555306] + ArrayLayouts v0.4.10
  ...
  [8dfed614] + Test
  [cf7118a7] + UUIDs
  [4ec0a83e] + Unicode
   Building TimeZones → `~/.julia/packages/TimeZones/fr1IP/deps/build.log`

(julia) pkg> status
Status `~/Documents/private/st679/julia/Project.toml`

(julia) pkg> status --manifest
Status `~/Documents/private/st679/julia/Manifest.toml`
  [4c555306] ArrayLayouts v0.4.10
  ...
  [38e38edf] GLM v1.3.11
  ...
  [ff71e718] MixedModels v3.1.0
  ...
  [2913bbd2] StatsBase v0.33.2
  [4c63d2b9] StatsFuns v0.9.5
  [3eaba693] StatsModels v0.6.15
  ...
  [4ec0a83e] Unicode

press backspace to go back to the julian mode.

Later, in a different session started from any other directory, we can use the environment we created earlier by providing the path to our folder julia/:

(v1.5) pkg> activate st679/julia
 Activating environment at `~/Documents/private/st679/julia/Project.toml`

(julia) pkg> status
Status `~/Documents/private/st679/julia/Project.toml`
  [ff71e718] MixedModels v3.1.0

julia> using MixedModels # first-time compilation can take a while!
[ Info: Precompiling MixedModels [ff71e718-51f3-5ec2-a782-8ffcbfa3c316]

help?> LinearMixedModel
search: LinearMixedModel GeneralizedLinearMixedModel

  LinearMixedModel

  Linear mixed-effects model representation

  Fields
  ========

    •    formula: the formula for the model
    ...

by the way: great tutorials on julia and mixed models here

using packages

install packages in pkg mode:

(julia) pkg> add RDatasets # to import data sets defined in R packages
(julia) pkg> add DataFrames
(julia) pkg> add CSV
(julia) pkg> add StatsPlots
(julia) pkg> add Plots
(julia) pkg> status # to check what package(s) you can use

then in julian mode:

using RDatasets
using DataFrames
# using Plots
using StatsPlots

iris = dataset("datasets", "iris")
head(iris)
ncol(iris)
nrow(iris)
describe(iris)
iris[:,5] # 5th column: copy -> no danger of modifying the iris data frame
iris[!,5] # 5th column: no copy -> can modify iris since vectors are mutable
iris[5]   # 5th column: no copy -- but deprecated
iris[:Species] #        no copy -- but deprecated
iris.Species
scatter(iris[:SepalLength], iris[:SepalWidth], group = iris[:Species])
@df iris scatter(:SepalLength, :SepalWidth, group=:Species,
   legend = :topleft,
   xlabel="sepal length", ylabel="sepal width",
   title="iris data")
@df iris marginalhist(:PetalLength, :PetalWidth)
@df iris marginalhist(:PetalLength, :PetalWidth, bins=30)
@df iris corrplot([:SepalLength :SepalWidth :PetalLength :PetalWidth],
    grid=false, bins=20)
@df iris corrplot([:SepalLength :SepalWidth :PetalLength :PetalWidth],
    grid=false, bins=20, group=:Species)

for plots, see many options from StatsPlots and Gadfly for an equivalent to ggplot2 in R.

The Interact package is super cool too to add interactive widgets to plots, e.g. with the @manipulate macro to make interactive graphs with sliders.

package RCall and the R mode

see Doug Bates’s video from his talk at JuliCon 2016, on “Julia and R”, about the RCall package

first install the package: do ] to switch to pkg mode

(julia) pkg> add RCall
(julia) pkg> status
Status `~/Documents/private/st679/julia/Project.toml`
  ...
  [6f49c342] RCall v0.13.9
  ...

to use this fantastic RCall package: switch back to julian mode (backspace)

julia> using RCall
julia> a = 1:10
R> $a  # switch to R mode by pressing $
R> b = (10:1)^2
julia> R"b"
julia> @rget b
julia> var"b"
julia> a + @rget(b)
R> $a + b
julia> R"$a+b"
julia> R"plot($a,b, col=1:10, pch=16, main='what a figure')"
julia> R"plot"(a, @rget(b), col=1:10, pch=16, main="figure 2")
R> install.packages("ggplot2") # do this if ggplot2 not installed already
julia> @rlibrary ggplot2

and now we can stick to julian mode only, but use the functions imported from ggplot, right from julia:

?ggplot
using RDatasets
using DataFrames
iris = dataset("datasets", "iris") # julia data set
ggplot(iris, aes(x=:SepalLength, y=:SepalWidth, color=:Species)) +
  geom_point()
ggsave("iris.pdf")
ggplot(iris, aes(x=:SepalLength, y=:SepalWidth, color=:PetalLength)) +
  geom_point(var"na.rm"=true) + facet_wrap(R"~Species")

note 2 things above:

the argument na.rm in an R function, but the dot notation is used for accessing an object’s fields (or module variable or function etc.) in julia. solution: use the var"" macro from RCall
y~x or ~x will produce an object of class formula in R, but not in julia. solution: use the R"" macro from RCall

Pluto notebooks

using Pluto
Pluto.run()

then follow instructions, like:

Opening http://localhost:1234/?secret=T3BDrOY9 in your default browser... ~ have fun!
Press Ctrl+C in this terminal to stop Pluto

other great packages

use JuliaHub to discover packages on a topic of interest, access package documentation, and more
PyCall, with @pyimport macro to import python modules and use them within julia
Flux for machine learning

previous & next