A small confusing implementation on assignments in Algorithm 2.1 and Example 3.1.

[cocomoff]

I tried to run Example 3.1 based on codes in Sec. 2 and Sec.3 as follows:

X = Variable(:x, 2); Y = Variable(:y, 2); Z = Variable(:z, 2)

# Table of X, Y (above) and Y, Z (below) at p.45
φ1 = Factor([X, Y], FactorTable((x=0, y=0) => 0.3, (x=0, y=1) => 0.4, (x=1, y=0) => 0.2, (x=1, y=1) => 0.1))
φ2 = Factor([Y, Z], FactorTable((y=0, z=0) => 0.2, (y=0, z=1) => 0.0, (y=1, z=0) => 0.3, (y=1, z=1) => 0.5))

and I got the below result after applying Base.:* function:

using PrettyPrint
φ12 = φ1 * φ2;
pprint(φ12.table)

# output is
http://localhost:8888/lab?{
  {:y : 0, :z : 1, :x : 0} : 0.0,
  {:y : 0, :z : 2, :x : 0} : 0.0,
  {:y : 1, :z : 1, :x : 1} : 0.05,
  {:y : 1, :z : 2, :x : 0} : 0.0,
  {:y : 1, :z : 2, :x : 1} : 0.0,
  {:y : 0, :z : 1, :x : 1} : 0.0,
  {:y : 0, :z : 2, :x : 1} : 0.0,
  {:y : 1, :z : 1, :x : 0} : 0.2,
}

This result did not equal to what I expected; The result table at p.45 of X, Y, Z can be obtained by replacing 0 -> 1, 1-> 2, and 2->3 in our definition. After checking the code of assignment in Algorithm 2.1, I'm thinking that this result is from 1:v.m of

function assignments(vars::AbstractVector{Variable})
    names = [var.name for var in vars]
    return vec([Assignment(n=>v for (n,v) in zip(names, values)) for values in product((1:v.m for v in vars)...)])
end

So, some parts of examples are a bit confusing for me. I know that the value in Julia is often assumed as 1-origin, but for probabilities we often uses {0,1} and I'm sorry I have no answer about this point. However, some remarks or comments are useful to learn deeply from the textbook with Julia.

[mykelk]

Thanks for filing this issue! To make the algorithms simpler, we assume all variables can take on values 1:n. If you want something else, then you'd have to wrap the code to do the appropriate mapping. See notes in Example 2.5.

[cocomoff]

Thanks @mykelk for your great work! I have overlooked some notes in Sec.2, but now I can confirm that the code works (maybe comments in Algorithm 2.1 rather than Example 2.5?).