No error for custom continuous distribution

[erdal.k...@gmail.com]

Hi all,

For the following inference:

var posterior = Infer({
  method: "enumerate", model: function(){
  var x = gaussian({mu: 0, sigma: 1})
  return x
}})

I get this error message:

Gaussian({ mu: 0, sigma: 1 })

Enumerate can only be used with distributions that have finite support.

Which is telling me I cannot use a continuous distribution with enumerate.

However, when I simulate a continuous distribution using Math.rand(), there is no error message:

var posterior = Infer({
  method: "enumerate", model: function(){
  var x = Math.random()
  return x
}})

Should there be an error or warning message?

[null-a]

> Should there be an error or warning message?

Good question. Calling `Math.random` breaks one of the restrictions on calling out to JS functions:

http://docs.webppl.org/en/master/language.html#calling-javascript-functions

It would be nice if we had an error/warning for this, but doing this for the general case is probably quite hard. Alternatively, perhaps we could maintain a list of the names of often called functions that don't work, and use that to generate warnings. (Though I guess it's unlikely this will be something that gets added anytime soon.)

Cheers,

Paul