Reusing R code

When you write a lot of R code in a project, or across projects, you will often want to make reusable parts of code.

DSS provides several mechanisms for reusing R code:

• Packaging your code as functions or modules, and making them available in a specific project

• Importing code that has been made available from one project to another

• Packaging your code as functions or modules, and making them available in all projects

• Packaging your code as a reusable plugin, and making it available for coder and non-coder users alike

Sharing R code within a project

In each project, you can write R files in the Library editor. The project’s library editor is available by going into the “Code” universe (orange), and selecting the “Libraries” tab.

You can then write code under a “R source” folder, i.e. a folder that has the “R” icon associated to it.

Each time you want to use the functions defined in these R files, you need to use the dkuSourceLibR function, which uses the source feature of R to import the functions in the global namespace.

For example if you wrote a file “miscfuncs.R” that contains a do_awesome_stuff() function, you can use:

library(dataiku)

dkuSourceLibR("miscfuncs.R")

do_awesome_stuff()

Working with multiple source folders

By default, the Library editor contains two top-level folders, “python” and “R”, which are respectively the root folder for Python and R code.

You can create other folders that will also act as R source folders, i.e. that will be available using dkuSourceLibR to all R processes running in the project.

Working with multiple source folders is mainly useful when importing code from external Git repositories.

To manually edit the list and order of the R source folders, open the external-libraries.json file in the library editor and edit the rsrcPath list. All pathes must be relative to the root of the library editor.

Importing libraries from other projects

If you have created libraries in a project A, you can import them in project B. The libraries of project A will be added to the source path of all code running in project B.

• Go to the library editor of project B

• Open the external-libraries.json file

• Edit the importLibrariesFromProjects list and add the project key (which appears in the URL, i.e. not the project display name) to it

• Save the external-libraries.json file

You need to have “Read project content” permission on A and “write project content” on B.

Sharing R code globally

You can write custom R code files in the global R library editor. he global R library editor is available from the “Settings” menu.

The click on the “lib/R” tab

Files that are written here can be used by all R code in all projects.

Each time you want to use the functions defined in these R files, you need to use the dkuSourceLibR function, which uses the source feature of R to import the functions in the global namespace.

For example if you wrote a file “miscfuncs.R” that contains a do_awesome_stuff() function, you can use:

library(dataiku)

dkuSourceLibR("miscfuncs.R")

do_awesome_stuff()

Permissions

In order to use the global R library editor, you need to be granted the global (group-level) permission “Edit lib folders”

Manual editing of the R code library folder

Although not recommended, if you have shell access to the DSS machine, you can modify the library folder directly in DATA_DIR/lib/R

Packaging code as plugins

See Plugins

Packaging your R code as a package

The source mechanism is practical but does not enforce namespacing. To get namespacing, you need to build a complete R package, compile it, and install it (possibly in the code environment) using install.packages. This generally requires shell access on the machine running DSS.