From 6a36f7f9f87fd7206346a0a63109d2579d67aa0e Mon Sep 17 00:00:00 2001
From: dehann <dehann.fourie@gmail.com>
Date: Sat, 18 Sep 2021 14:20:59 -0400
Subject: [PATCH] more kwargs on prod, bump v0.5.8

---
 Project.toml     |  2 +-
 src/MSGibbs01.jl | 31 ++++++++++++++-----------------
 2 files changed, 15 insertions(+), 18 deletions(-)

diff --git a/Project.toml b/Project.toml
index 0884b11..cf80c24 100644
--- a/Project.toml
+++ b/Project.toml
@@ -2,7 +2,7 @@ name = "KernelDensityEstimate"
 uuid = "2472808a-b354-52ea-a80e-1658a3c6056d"
 keywords = ["NBP", "nonparametric", "kernel density", "functional products", "multiscale", "Gibbs sampling", "KDE"]
 desc = "A modified version of the multiscale Gibbs product and KDE"
-version = "0.5.7"
+version = "0.5.8"
 
 [deps]
 DelimitedFiles = "8bb1440f-4735-579b-a4ab-409b98df4dab"
diff --git a/src/MSGibbs01.jl b/src/MSGibbs01.jl
index c94bb52..c5c65f0 100644
--- a/src/MSGibbs01.jl
+++ b/src/MSGibbs01.jl
@@ -633,17 +633,20 @@ function prodAppxMSGibbsS(npd0::BallTreeDensity,
                           getLambda::T4=(getEuclidLambda,),
                           glbs = makeEmptyGbGlb(),
                           addEntropy::Bool=true,
-                          ndims::Int=maximum(Ndim.(trees)),
-                          partialDimMask::AbstractVector{BitVector} = [ones(Int,ndims) .== 1 for i in 1:length(trees)]  
+                          ndims::Integer=maximum(Ndim.(trees)),
+                          Ndens = length(trees),              # of densities
+                          Np    = Npts(npd0),                 # of points to sample
+                          maxNp = maximum([Np; Npts.(trees)]),
+                          Nlevels = floor(Int,(log(Float64(maxNp))/log(2.0))+1.0), # how many levels to a balanced binary tree?
+                          randU = rand(Int(Np*Ndens*(Niter+2)*Nlevels)),
+                          randN = randn(Int(ndims*Np*(Nlevels+1))),                
+                          partialDimMask::AbstractVector{<:BitVector} = [ones(Int,ndims) .== 1 for i in 1:length(trees)],
                         ) where {T1<:Tuple,T2<:Tuple,T3<:Tuple,T4<:Tuple}
   #
   #
   
   # See  Ihler,Sudderth,Freeman,&Willsky, "Efficient multiscale sampling from products
   #         of Gaussian mixtures", in Proc. Neural Information Processing Systems 2003
-  Ndens = length(trees)              # of densities
-  # ndims  = trees[1].bt.dims           # of dimensions
-  Np    = Npts(npd0)                # of points to sample
 
   # prepare stack manifold add and diff operations functions (manifolds must match dimension)
   addopT = length(addop)!=ndims ? ([ (addop[1]) for i in 1:ndims]...,) : addop
@@ -657,21 +660,15 @@ function prodAppxMSGibbsS(npd0::BallTreeDensity,
   points = zeros(ndims*Np)
   #??plhs[1] = mxCreateNumericMatrix(Ndens, Np, mxUINT32_CLASS, mxREAL);
   indices=ones(Int,Ndens, Np)
-  maxNp = Np
-  for tree in trees
-    if (maxNp < Npts(tree))
-      maxNp = Npts(tree)
-    end
-  end
+  # for tree in trees
+  #   if (maxNp < Npts(tree))
+  #     maxNp = Npts(tree)
+  #   end
+  # end
 
-  # how many levels to a balanced binary tree?
-  Nlevels = floor(Int,(log(Float64(maxNp))/log(2.0))+1.0)
 
   # Generate enough random numbers to get us through the rest of this
-  if true
-    randU = rand(Int(Np*Ndens*(Niter+2)*Nlevels))
-    randN = randn(Int(ndims*Np*(Nlevels+1)))
-  else
+  if false
     # FIXME using DelimitedFiles
     randU = vec(readdlm("randU.csv"))
     randN = vec(readdlm("randN.csv"))