RWebPPL

Installation

System requirements

Mac or Linux OS [Windows currently not supported]
R v3.3 (in RStudio, type version)
node v4.4.5 or higher (in Terminal, type node --version; close and re-open terminal after install)
npm v3.6 or higher (in Terminal, type npm --version; if it's not >= v3.6; try sudo npm install npm -g, close and reopen terminal and check version)
devtools R library: in R: install.packages('devtools')

N.B.: If you plan to use command-line WebPPL, in addition to RWebPPL, we advise you first npm install -g webppl and then install RWebPPL.

N.B.: If your node set-up depends on your environment variables, you may need to start rstudio from the terminal (rstudio does not otherwise load .bashrc files, for example).

devtools::install_github("mhtess/rwebppl")

If you already have installed WebPPL using npm install -g webppl, then RWebPPL will link to your installed version.

If you do not have WebPPL installed already, it will install it for you (into the R package directory).

Usage

Current list of function arguments and supported functionality

program_code: A string of a webppl program
program_file: A file containing a webppl program
data: A data frame (or other serializable object) to be passed from R to the webppl program
data_var: A name by which the data can be referenced in the webppl program
packages: A character vector of names of external webppl package to use
model_var: When using inference opts, the name by which the model be referenced in the program.
inference_opts: A list with options for inference of a particular model in the program. (see http://webppl.readthedocs.io/en/master/inference.html) [N.B.: requires using model_var]
output_format: An optional string indicating posterior output format: "webppl" probability table (default), "samples" for just the samples, "ggmcmc" for use with ggmcmc package. [N.B.: requires using inference_opts and model_var]
chains: Number of times to run program (defaults to 1).
cores: Number of cores to use when running multiple chains (defaults to 1).

Examples

Write a model as a string in R:

my_model <- "
var model = function () {
 var a = flip(0.3)
 var b = flip(0.6)
 return a + b
}
model()
"
webppl(my_model)

Or write a model in an external file:

webppl(program_file = "path/to/model/model.wppl")

WebPPL packages can be used in more complex models:

webppl(program_file = "path/to/model/model.wppl",
       packages = c("linked-list", "timeit"))

List of useful packages.

NPM packages that ares used inside of WebPPL packages can be installed directly from RWebPPL e.g. install_webppl_package("babyparse"). They can also be uninstalled in the same way: uninstall_webppl_package("babyparse")

Passing data from R to WebPPL

Data can be passed directly from R to WebPPL as in:

my_model <- "
var model = function () {
 var a = flip(0.3)
 var b = flip(0.6)
 var scores = map( function(d) {
 	return a + b - d
 }, myDF)
 return scores
}
model()
"

webppl(my_model,
	   data = df,
	   data_var = "myDF")

In this example, myDF is not defined inside the WebPPL program, but is passed into it from R, using data = df. The argument data_var tells WebPPL what the data should be called.

Structure of data when passing

If myDF looks like this in R:

Participant	Condition	Response
1	A	0.4
1	B	0.8
2	A	0.2

It will exist in WebPPL as a list of js objects e.g.

[
  {
    participant: 0,
    condition: "A",
    response: 0.4
  },
  {
    participant: 0,
    condition: "B",
    response: 0.8
  },
  {
    participant: 1,
    condition: "A",
    response: 0.2
  },
  ...
]

erindb/rwebppl