Julia Data Science Docker with data science packages compiled for instant loading!
Time-to-first-plot (TTFP) is often regarded as one of Julia's main pain points. The PackageCompiler.jl package can compile these package and alleviate the pain. It works by pre-"compiling" the packages and baking them into the julia sysimage so that using Pkg1 will be fast just like base packages.
This is an experimental first attempt at making data science packages used by me into a docker image with pre-compiled data science packages.
Firstly, install Docker. If you are running Windows, I recommend installing git so you have access to git bash.
On Windows you IP can be found using ipconfig and on Linux with ifconfig. This is needed if you wish to do plotting from the docker image.
Basic: Windows
docker run --rm \
-e DISPLAY=YOUR_IP:0.0 \
-e JUPYTER_ENABLE_LAB=yes \
-v "$PWD":/home/jovyan/work\
-it -p 8888:8888 \
xiaodaidocker2019/julia-data-science-baseOften one may wish to save the data to somewhere on the hard drive, you may do this by attaching a local folder to the directory somedir.
The below packages are compiled using PackageCompiler.jl into the image
| Package | Type | Notes |
|---|---|---|
| CategoricalArrays | Foundation | |
| Clustering | Unsupervised learning | |
| CSV | Data IO | |
| DataConvenience | Data Manipulation/Convenience | |
| DataFrames | Data Manipulation | |
| DataFramesMeta | Data Manipulation | |
| DecisionTree | Supervised learning | |
| FastGroupBy | Data Manipulation/Convenience | |
| Feather | Data IO | |
| FreqTables | Foundation/Statistics | |
| GLM | Supervised learning | |
| JDF | Data IO | For reading/writing JDF files |
| JLBoost | Supervised learning | |
| Lazy | Data Manipulation/Convenience | |
| Missings | Foundation | |
| Parquet | Data IO | ParquetFiles is quite broken at the moment |
| Plots | Plotting | |
| RDatasets | Data | |
| SortingLab | Data Manipulation/Convenience | |
| StatsBase | Foundation/Statistics | |
| StatsPlots | Plotting | |
| Tables | Data Manipulation/Convenience | |
| TableView | Data Viewing | |
| XGBoost | Supervised learning |
The below packages are included but not compiled
| Package | Type | Notes |
|---|---|---|
| Pipe | Data Manipulation/Convenience | If compiled into base then there is warning message with Pipe |
| TableView | Data Viewing | If compiled then doesn't work with JupyterLab |