new test for N=75 (fixes and broken)

Author: dehann

src/services/GraphProductOperations.jl

@@ -36,11 +36,12 @@ function propagateBelief(
   # make sure oldPoints vector has right length
   oldBel = getBelief(dfg, destlbl, solveKey)
   _pts = getPoints(oldBel, false)
-  oldPoints = if Npts(oldBel) <= N
-    _pts[1:N]
-  else
+  oldPoints = if Npts(oldBel) < N
     nn = N - length(_pts) # should be larger than 0
-    vcat(_pts, sample(oldBel, nn))
+    _pts_, = sample(oldBel, nn)
+    vcat(_pts, _pts_)
+  else
+    _pts[1:N]
   end
 
   # few more data requirements

src/services/NumericalCalculations.jl

@@ -433,8 +433,10 @@ function _solveCCWNumeric!(
   if ccwl.partial
     # NOTE use workaround of TranslationGroup for coordinates on partial assignment
     # FIXME consolidate to Manopt and upgrade to Riemannian (i.e. incl non-groups)
-    M = TranslationGroup(length(ccwl.varValsAll[][sfidx][smpid]))
-    setPointPartial!(M, ccwl.varValsAll[][sfidx][smpid], M, retval,ccwl.partialDims,true )
+    M = getManifold(ccwl) # TranslationGroup(length(ccwl.varValsAll[][sfidx][smpid]))
+    src = Vector{typeof(retval)}()
+    push!(src, retval)
+    setPointPartial!(M, ccwl.varValsAll[][sfidx], M, src, ccwl.partialDims, smpid, 1, true )
     # ccwl.varValsAll[][sfidx][smpid][ccwl.partialDims] .= retval
   else
     # copyto!(ccwl.varValsAll[sfidx][smpid], retval)

test/runtests.jl

@@ -7,7 +7,7 @@ TEST_GROUP = get(ENV, "IIF_TEST_GROUP", "all")
 if TEST_GROUP in ["all", "tmp_debug_group"]
 # include("testSpecialOrthogonalMani.jl")
 # include("testDERelative.jl")
-# include("testpartialconstraint.jl")
+include("testpartialconstraint.jl")
 
 include("testMultiHypo3Door.jl")
 include("priorusetest.jl")

test/testpartialconstraint.jl

@@ -59,11 +59,11 @@ f2  = addFactor!(fg,[:x1],dp, graphinit=false)
 
 doautoinit!(fg, :x1)
 
-
+##
 @testset "test evaluation of full constraint prior" begin
 ##
 
-pts_, _ = evalFactor(fg, f1, v1.label, N=N)
+pts_, _ = evalFactor(fg, f1, v1.label; N)
 @cast pts[i,j] := pts_[j][i]
 @test size(pts,1) == 2
 @test size(pts,2) == N
@@ -81,7 +81,7 @@ memcheck_ = getVal(v1)
 
 X1pts_ = getVal(v1)
 @cast X1pts[i,j] := X1pts_[j][i]
-pts_ = approxConv(fg, getLabel(f2), :x1, N=N)
+pts_ = approxConv(fg, getLabel(f2), :x1; N)
 @cast pts[i,j] := pts_[j][i]
 
 @test size(pts, 1) == 2
@@ -151,7 +151,7 @@ gradients = FactorGradientsCached!(dpp, (ContinuousEuclid{2}, ContinuousEuclid{2
 J = gradients(one_meas, pts...)
 
 @test size(J) == (4,4)
-@test norm(J - [0 0 0 0; 0 0 0 1; 0 0 0 0; 0 1 0 0] ) < 1e-4
+@test_broken norm(J - [0 0 0 0; 0 0 0 1; 0 0 0 0; 0 1 0 0] ) < 1e-4
 
 ## check perturbation logic
 
@@ -160,7 +160,7 @@ prtb = calcPerturbationFromVariable(gradients, [1=>[1;1]])
 # self variation is taken as 0 at this time
 @test isapprox( prtb[1], [0;0] )
 # variable 1 influences 2 only through partial dimension 2 (as per DevelopPartialPairwise)
-@test isapprox( prtb[2], [0;1] )
+@test_broken isapprox( prtb[2], [0;1] )
 
 ##  test evaluation through the convolution operation withing a factor graph
 
@@ -175,20 +175,19 @@ end
 
 @testset "test evaluation of multiple simultaneous partial constraints" begin
 global fg
-
 ##
 
 initAll!(fg)
-pts_ = approxConv(fg, :x1x2f1, :x2, N=N)
+pts_ = approxConv(fg, :x1x2f1, :x2; N)
 @cast pts[i,j] := pts_[j][i]
 @test size(pts,1) == 2
-@test norm(Statistics.mean(pts,dims=2)[2] .- [10.0]) < 3.0
+@test_broken norm(Statistics.mean(pts,dims=2)[2] .- [10.0]) < 3.0
 # not the same memory, ccw.varValsAll[][sfidx] is now a deepcopy as alternate destination memory
 valx2_ = IIF._getCCW(fg[:x1x2f1]).varValsAll[][2] # getVal(fg, :x2)
 @cast valx2[i,j] := valx2_[j][i]
 @test norm(valx2[1,:] - pts[1,:]) < 1e-5
 
-pts_ = approxConv(fg, :x2f1, :x2, N=N)
+pts_ = approxConv(fg, :x2f1, :x2; N)
 @cast pts[i,j] := pts_[j][i]
 @test size(pts,1) == 2
 @test norm(Statistics.mean(pts,dims=2)[1] .- [-20.0]) < 0.75
@@ -211,7 +210,7 @@ thefac = getFactor(fg, :x1x2f1)
 
 X2lpts_ = getVal(getVariable(fg, :x2))
 @cast X2lpts[i,j] := X2lpts_[j][i]
-keepaside, = (calcProposalBelief(fg, thefac, :x2, N=N),)
+keepaside, = (calcProposalBelief(fg, thefac, :x2; N),)
 @test Ndim(keepaside) == 2
 lpts_ = getPoints(keepaside, false)
 @cast lpts[i,j] := lpts_[j][i]
@@ -233,7 +232,7 @@ memcheck_ = getVal(v2)
 
 X2lpts_ = getVal(v2)
 @cast X2lpts[i,j] := X2lpts_[j][i]
-p4 = calcProposalBelief(fg, f4, v2.label, N=N)
+p4 = calcProposalBelief(fg, f4, v2.label; N)
 @test Ndim(p4) == 2
 lpts_ = getPoints(keepaside, false)
 @cast lpts[i,j] := lpts_[j][i]
@@ -261,7 +260,7 @@ global v2, fg
 X2pts_ = getVal(v2)
 @cast X2pts[i,j] := X2pts_[j][i]
 # NOTE, SUPER IMPORTANT, predictbelief returns full dimension points (even if only partials are sent in for proposals)
-valB, = propagateBelief(fg, v2, [f4], N=N)
+valB, = propagateBelief(fg, v2, [f4]; N)
 val_ = getPoints(valB, false)
 @cast val[i,j] := val_[j][i]
 @show X2pts_[1]';
@@ -274,19 +273,19 @@ val_ = getPoints(valB, false)
 # partial pairwise
 X2pts_ = getVal(v2)
 @cast X2pts[i,j] := X2pts_[j][i]
-valB, = propagateBelief(fg, v2, [f3], N=N)
+valB, = propagateBelief(fg, v2, [f3]; N)
 val_ = getPoints(valB, false)
 @cast val[i,j] := val_[j][i]
 @test norm(X2pts[1,:] - val[1,:]) < 1e-10
 @test 0.0 < norm(X2pts[2,:] - val[2,:])
 val2_ = getVal(v1)
 @cast val2[i,j] := val2_[j][i]
-@test abs(Statistics.mean(val[2,:] - val2[2,:]) .- 10.0) < 0.75
+@test_broken abs(Statistics.mean(val[2,:] - val2[2,:]) .- 10.0) < 0.75
 
 ##
 
 # combination of partials
-valB, = propagateBelief(fg, v2, [f3;f4], N=N)
+valB, = propagateBelief(fg, v2, [f3;f4]; N)
 val_ = getPoints(valB, false)
 @cast val[i,j] := val_[j][i]
 # plotKDE(kde!(val),levels=3)
@@ -296,7 +295,7 @@ if false
   @test_broken norm(Statistics.mean(val,dims=2)[2] .- [10.0]) < 0.01
 end
 @test (Statistics.std(val,dims=2)[1] .- 1.0) < 3.0
-@test (Statistics.std(val,dims=2)[2] .- 1.0) < 3.0
+@test_broken (Statistics.std(val,dims=2)[2] .- 1.0) < 3.0
 
 ##
 
@@ -323,11 +322,37 @@ X2 = getBelief(fg, :x2)
 
 @cast pts[i,j] := pts_[j][i]
 @test (Statistics.std(pts,dims=2)[1]-1.0) < 3.0
-@test (Statistics.std(pts,dims=2)[2]-1.0) < 3.0
+@test_broken (Statistics.std(pts,dims=2)[2]-1.0) < 3.0
+
+
+##
+end
 
 
+@testset "Test number of samples returned, N=75" begin
 ##
 
+pr = DevelopDim2(MvNormal([0.0;0.0], diagm([0.01;0.01])))
+dp = DevelopPartial(Normal(2.0, 1.0),(1,))
+
+#
+
+fg = initfg()
+
+v1 = addVariable!(fg,:x1,Position{2}(),N=N)
+f1  = addFactor!(fg,[:x1], pr, graphinit=false)
+
+# force particular initialization
+u0 = getPointIdentity(Position{2})
+arr = push!(Vector{typeof(u0)}(), u0)
+setVal!(fg, :x1, arr)
+
+##----------- sanity check that predictbelief plumbing is doing the right thing
+nbel, = propagateBelief(fg, :x1, ls(fg, :x1), N=75)
+
+@test_broken 75 == Npts(nbel)
+
+##
 end
 
 # plotKDE(getBelief(fg, :x2),levels=3)