Initial attempt at hasvalue(vals, vn, dist)

Author: penelopeysm

src/sampler.jl

@@ -88,14 +88,6 @@ Default type of the chain of posterior samples from `sampler`.
 """
 default_chain_type(sampler::Sampler) = Any
 
-"""
-    init_strategy(sampler)
-
-Define the initialisation strategy used for generating initial values when
-sampling with `sampler`. Defaults to `PriorInit()`, but can be overridden.
-"""
-init_strategy(::Sampler) = PriorInit()
-
 """
     initialstep(rng, model, sampler, varinfo; kwargs...)
 

src/utils.jl

@@ -845,7 +845,7 @@ end
 
 # For `dictlike` we need to check wether `vn` is "immediately" present, or
 # if some ancestor of `vn` is present in `dictlike`.
-function hasvalue(vals::AbstractDict, vn::VarName)
+function hasvalue(vals::AbstractDict{<:VarName}, vn::VarName)
     # First we check if `vn` is present as is.
     haskey(vals, vn) && return true
 
@@ -867,6 +867,39 @@ function hasvalue(vals::AbstractDict, vn::VarName)
 
     return canview(child, value)
 end
+# TODO(penelopeysm): Figure out tuple / namedtuple distributions, and LKJCholesky (grr)
+function hasvalue(vals::AbstractDict, vn::VarName, dist::Distribution)
+    @warn "`hasvalue(vals, vn, dist)` is not implemented for $(typeof(dist)); falling back to `hasvalue(vals, vn)`."
+    return hasvalue(vals, vn)
+end
+hasvalue(vals::AbstractDict, vn::VarName, ::UnivariateDistribution) = hasvalue(vals, vn)
+function hasvalue(
+    vals::AbstractDict{<:VarName},
+    vn::VarName{sym},
+    dist::Union{MultivariateDistribution,MatrixDistribution},
+) where {sym}
+    # If `vn` is present as-is, then we are good
+    hasvalue(vals, vn) && return true
+    # If not, then we need to check inside `vals` to see if a subset of
+    # `vals` is enough to reconstruct `vn`. For example, if `vals` contains
+    # `x[1]` and `x[2]`, and `dist` is `MvNormal(zeros(2), I)`, then we
+    # can reconstruct `x`. If `dist` is `MvNormal(zeros(3), I)`, then we
+    # can't.
+    # To do this, we get the size of the distribution and iterate over all
+    # possible indices. If every index can be found in `subsumed_keys`, then we
+    # can return true.
+    sz = size(dist)
+    for idx in Iterators.product(map(Base.OneTo, sz)...)
+        new_optic = if getoptic(vn) === identity
+            Accessors.IndexLens(idx)
+        else
+            Accessors.IndexLens(idx) ∘ getoptic(vn)
+        end
+        new_vn = VarName{sym}(new_optic)
+        hasvalue(vals, new_vn) || return false
+    end
+    return true
+end
 
 """
     nested_getindex(values::AbstractDict, vn::VarName)