o2r-project / dockerfiles

Dockerfiles for base images for executable research containers

dockerfiles

Dockerfiles for base images for executable research containers (ERC) and for use in the o2r microservice architecture.

geospatial-ctv

This image comes in three layers, each extending the previous one by installing a full CRAN Task View.

• o2rproject/geospatial-ctv:spatial has Spatial
• o2rproject/geospatial-ctv:spatiotemporal also has SpatioTemporal
• o2rproject/geospatial-ctv:spatiotemporal-timeseries also has TimeSeries

spatial

Based on the image rocker/geospatial.

spatiotemporal

Currently missing packages which could not be installed from the task view:

• rpanel ([tcl] can't find package BWidget.)
• lgcp (depends on rpanel)
• tkrplot (tcltkimg.c:2:16: fatal error: tk.h: No such file or directory)
• RPyGeo (package ‘RPyGeo’ is not available (for R version 3.4.0))
• rJava (error: Java interpreter '/usr/bin/java' does not work)
• adehabitat (depends on tkrplot)
• spcosa (depends on rJava)
• OpenStreetMap (depends on rJava)
• spatsurv (depends on OpenStreetMap)

spatiotemporal-timeseries

Currently missing packages which could not be installed from the task view:

• rpanel ([tcl] can't find package BWidget.)
• RGtk2 (configure: error: GTK version 2.8.0 required)
• cairoDevice (ERROR: gtk+2.0 not found by pkg-config.)
• SDD (depends on rpanel)
• x12GUI (depends on RGtk2, cairoDevice)

geospatial-containerit

Extending the image geospatial-ctv:spatiotemporal-timeseries with containerit to have a good image while running submitted workspaces in o2r-loader.

dockerfiles's Issues

Add Dockerfile *st-base*

Installed software

• LaTeX
• pandoc
• R
• gdal
• proj
• CRAN task views (and the required system libs)
  • spatial
  • spatiotemporal
  • install via ctv or install via Dockerfile? write small R script that creates a RUN statement with all packages in spatial und spatiotemporal.

See also https://github.com/o2r-project/o2r-bagtainers/blob/master/0010/data/container/Dockerfile

This should be based on rsysreqs, not on manually installing all system dependencies: https://github.com/r-hub/sysreqsdb

A source might be https://github.com/rstudio/shinyapps-package-dependencies/tree/master/packages

Add Dockerfiles with "all of CRAN"

• variants
  • install from CRAN
  • install from source using github.com/cran
  • install from source using CRAN
• compare rocker (Ubuntu) vs. alpine base image
  • https://hub.docker.com/r/sgerrand/alpine-r/
  • with all CRAN packages, is a large OS actually still overhead?
  • install R from source vs. install R from repos

How large are the images?
What packages are not easy to install?

Full source Dockerfile

Create a Dockerfile that either installs all source packages, or builds everything that is needed from source (i.e. R, the compilers, libraries for geo-related packages)