Add a R package to Arbor

[Caetanods]

Hello Arbor team,

This is more like a series of questions than a 'bug-report'.

I have a new R package and I want to implement it on Arbor. I was in search for documentation and examples of how to do this. As a developer, I have some questions about current Arbor functionality:

1. Where can I find documentation to implement a new package as a collection of functions in Arbor? I found documentation how to apply an analysis and how to create a workflow with existing functions on Arbor. But I do not know how to create a new set of tools.
2. My package depends on writing the output to files while the analysis is running. Right now a R function will create a new file and append to it. Would this kind of function be compatible with Arbor?
3. Can I make a package available from github also available on Arbor?
4. Can I push the Arbor implementation of my package so people will have access to it? Or I need to host the Arbor implementation in my own repository?

Thanks for the help and sorry for the long list of questions.

[jeffbaumes]

These should all be possible.

1. Deploying a new set of analyses to a particular Arbor instance requires admin access to that instance, where you can create and edit analyses from the UI. Do you have your own instance or are you looking to add to an existing instance?

2. Writing to files is fine during execution, but Arbor currently requires that outputs get read in and stored in a variable at the end of execution in order to be sent to the UI or later workflow stages.

3. Any R package used by Arbor analyses are normally installed by an admin on the worker server. This can include github installable R packages.

4. If you are looking to utilize an existing Arbor installation, I would ask @curtislisle about possibly hosting the analyses. I would recommend a local install to set up and test your new analyses.

[Caetanods]

Thanks for your answer!

I got in touch with Luke and Curtis, so I have access to an instance with admin powers now.

About topics 1. and 4. , it was my impression that Arbor would make it easy to make workflows public to users everywhere. I guess that these are future aims?

Just comparing Arbor with base R. In R I can create a package and push to github or RCran. Then I tell the user to install it with a single line of code. After that, the user can follow a tutorial in a website or follow the example in the functions help pages.

My objective is to do the same with Arbor. I see I can set up my own workflow in an instance that I have admin access to. I guess my question is: What would be the steps so that the workflow (and Arbor functions) I created based on my package can be used by someone else with their own instance (possibly without admin powers)?

[jeffbaumes]

Thanks for flexing what Arbor can do. Analyses and workflows can be downloaded as JSON and exchanged between instances. This is the nominal way to share analyses between instances.

Whether a user needs to be admin (really needs a special whiltelisted privilege on their instance) depends on the scenario. If there is a central Arbor instance that they can log into where you've put your analysis, Arbor can be configured so whole sets of algorithms can be run by any user. If they have their own instance that they control, it is reasonable to expect that they have the privileges needed and can load your analysis JSON files to use your analyses.

Of course there are other ways that analyses could be shared and used, but these two modes (sharing an instance or download/upload JSON files) is what is possible in Arbor today.