switch upInitMsg to pull model

[dehann]

Related to #674 decision to use only a message pull model. This PR is part of #459 refactoring. This PR is specifically focused on up-init-msg refactoring steps.

[dehann]

keep temporary debugging code:

using IncrementalInference

## small canonical factor graph, without graphinit
fg = generateCanonicalFG_CaesarRing1D(graphinit=false)
getSolverParams(fg).graphinit = false
getSolverParams(fg).dbg = true

# do all init on tree as part of solve
# getSolverParams(fg).drawtree = true
# tree, smt, hist = solveTree!(fg)


tree = wipeBuildNewTree!(fg)


# getSolverParams(fg).limititers = 50

TS = Vector{Task}(undef,6)

iters = 100*ones(Int, 6)

# good case
# iters = [47;53;25;34;34;34]
# iters = [59;46;26;35;35;35] # adding CSM 8e


# iters = [3;5;14;14;15;18]
# iters = [3;5;14;14;18;18] cannot find key 2
# iters = [10;16;14;18;18;15]
# iters = [6;16;18;18;18;15]
# iters = [10;16;18;18;18;15]

# all issues on attemptCliqInitDown
# iters = [10;16;19;18;18;15] # breaks on 3.19, key 2 not found
# iters = [10;16;18;19;18;15] # breaks on 4.19, key 2 not found
# iters = [10;17;18;18;18;15] # blocks on 2.17 ??
# iters = [10;16;18;18;19;15] # blocks on 5.19 ??

# iters = [10;16;18;18;24;15]

# w len == 1
# iters = [18;54;22;40;50;50]
# iters = [18;44;22;31;34;34]
# iters = [18;39;22;31;34;34]

# csm add 8e
# iters = [18;39;22;32;35;35]


SLEEP = 10

getSolverParams(fg).limititers = iters[3]
TS[3] = solveCliq!(fg, tree, :l1, recordcliq=true, async=true)
sleep(SLEEP)
getSolverParams(fg).limititers = iters[4]
TS[4] = solveCliq!(fg, tree, :x1, recordcliq=true, async=true)
sleep(SLEEP)
getSolverParams(fg).limititers = iters[2]
TS[2] = solveCliq!(fg, tree, :x4, recordcliq=true, async=true)
sleep(SLEEP)
getSolverParams(fg).limititers = iters[5]
TS[5] = solveCliq!(fg, tree, :x5, recordcliq=true, async=true)
sleep(SLEEP)
getSolverParams(fg).limititers = iters[1]
TS[1] = solveCliq!(fg, tree, :x0, recordcliq=true, async=true)
sleep(SLEEP)
getSolverParams(fg).limititers = iters[6]
TS[6] = solveCliq!(fg, tree, :x3, recordcliq=true, async=true)


sleep(SLEEP)
TS

##




hists = fetchCliqHistoryAll!(TS)
printCliqHistorySequential(hists)
printCliqHistorySequential(hists, joinLogPath(fg,"HistoryCSMAll.txt"))

0


##


using Logging


CLIQ = 5
STEP = 29

hist_ = fetch(TS[CLIQ])

fnct__ = deepcopy(hist_[STEP][3])
csmc__ = deepcopy(hist_[STEP][4])

# attachCSM!(csmc__, fg, tree)

csmc__.logger = SimpleLogger(stdout)

# getCliqueData(csmc__.cliq).solveCondition = Condition()
#
# getCliqueData(csmc__.cliq).initUpChannel
# tmpmsg = getCliqueData(csmc__.cliq).initUpChannel |> fetch
# getCliqueData(csmc__.cliq).initUpChannel = Channel{LikelihoodMessage}(1)
# put!(getCliqueData(csmc__.cliq).initUpChannel, tmpmsg)

# Juno.@enter fnct__(csmc__)
fnct__
fnct__ = fnct__(csmc__)
# Juno.@enter fnct__(csmc__)
# t5_19 = @async fnct__(csmc__)


# t4_19 = @async fnct__(csmc__)
# t3_19 = @async fnct__(csmc__)
# t5_25 = @async fnct__(csmc__)


istaskdone(t4_19)
istaskdone(t3_19)
istaskdone(t5_25)

fnct__ = fetch(t5_24)
0


## The error

# cliq 2, 3, or 4
csmc = csmc__
prnt = getParent(csmc.tree, csmc.cliq)[1]
dwinmsgs = prepCliqInitMsgsDown!(csmc.dfg, csmc.tree, prnt, csmc.cliq, logger=SimpleLogger(stdout), dbgnew=true)


0

## Other sequences that induce the same failures in either clique 2, 3, 4



# # makes cliq2 error with key 2 not found
# TS[1] = solveCliq!(fg, tree, :x0, recordcliq=true, async=true)
# sleep(5)
# TS[2] = solveCliq!(fg, tree, :x4, recordcliq=true, async=true)
# sleep(2)
# TS[5] = solveCliq!(fg, tree, :x5, recordcliq=true, async=true)
# sleep(2)
# TS[6] = solveCliq!(fg, tree, :x3, recordcliq=true, async=true)
# sleep(2)
# TS[3] = solveCliq!(fg, tree, :l1, recordcliq=true, async=true)
# sleep(2)
# TS[4] = solveCliq!(fg, tree, :x1, recordcliq=true, async=true)


# usually breaks on 4
# TS[4] = solveCliq!(fg, tree, :x1, recordcliq=true, async=true)
# TS[3] = solveCliq!(fg, tree, :l1, recordcliq=true, async=true)
# TS[1] = solveCliq!(fg, tree, :x0, recordcliq=true, async=true)
# TS[2] = solveCliq!(fg, tree, :x4, recordcliq=true, async=true)
# TS[5] = solveCliq!(fg, tree, :x5, recordcliq=true, async=true)
# TS[6] = solveCliq!(fg, tree, :x3, recordcliq=true, async=true)


#

[codecov-commenter]

Codecov Report

Merging #761 into master will decrease coverage by 1.37%.
The diff coverage is 36.89%.

@@            Coverage Diff             @@
##           master     #761      +/-   ##
==========================================
- Coverage   62.12%   60.75%   -1.38%     
==========================================
  Files          37       37              
  Lines        4520     4688     +168     
==========================================
+ Hits         2808     2848      +40     
- Misses       1712     1840     +128     

Impacted Files	Coverage Δ
src/BeliefTypes.jl	25.80% <0.00%> (-27.53%)	⬇️
src/CliqStateMachineUtils.jl	53.18% <0.00%> (-6.26%)	⬇️
src/CliqueTypes.jl	47.05% <0.00%> (-41.84%)	⬇️
src/Deprecated.jl	20.00% <0.00%> (-2.00%)	⬇️
src/IncrementalInference.jl	75.00% <ø> (ø)
src/JunctionTree.jl	73.80% <0.00%> (ø)
src/JunctionTreeTypes.jl	17.68% <1.33%> (-13.84%)	⬇️
src/TreeMessageUtils.jl	54.76% <50.00%> (ø)
src/SolverAPI.jl	67.90% <66.66%> (+0.36%)	⬆️
src/TreeBasedInitialization.jl	85.38% <90.00%> (+1.40%)	⬆️
... and 9 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cbddb39...06c00b0. Read the comment docs.