Make Adapt and InverseFunctions hard dependencies again

Project.toml

@@ -1,33 +1,31 @@
 name = "ValueShapes"
 uuid = "136a8f8c-c49b-4edb-8b98-f3d64d48be8f"
-version = "0.11.5"
+version = "0.11.6"
 
 [deps]
+Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
 ArgCheck = "dce04be8-c92d-5529-be00-80e4d2c0e197"
 ArraysOfArrays = "65a8f2f4-9b39-5baf-92e2-a9cc46fdf018"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 ElasticArrays = "fdbdab4c-e67f-52f5-8c3f-e7b388dad3d4"
 FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b"
 IntervalSets = "8197267c-284f-5f27-9208-e0e47529a953"
+InverseFunctions = "3587e190-3f89-42d0-90ee-14403ec27112"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 TypedTables = "9d95f2ec-7b3d-5a63-8d20-e2491e220bb9"
 
 [weakdeps]
-Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 ChangesOfVariables = "9e997f8a-9a97-42d5-a9f1-ce6bfc15e2c0"
-InverseFunctions = "3587e190-3f89-42d0-90ee-14403ec27112"
 Mooncake = "da2b9cff-9c12-43a0-ae48-6db2b0edb7d6"
 ZygoteRules = "700de1a5-db45-46bc-99cf-38207098b444"
 
 [extensions]
-ValueShapesAdaptExt = "Adapt"
 ValueShapesChainRulesCoreExt = "ChainRulesCore"
 ValueShapesChangesOfVariablesExt = "ChangesOfVariables"
-ValueShapesInverseFunctionsExt = "InverseFunctions"
 ValueShapesMooncakeExt = "Mooncake"
 ValueShapesZygoteRulesExt = "ZygoteRules"
 

src/ValueShapes.jl

@@ -18,10 +18,12 @@ using ArraysOfArrays
 using Distributions
 using ElasticArrays
 using FillArrays
+using InverseFunctions
 using Random
 using Statistics
 using StatsBase
 
+import Adapt
 import IntervalSets
 import Tables
 import TypedTables

src/named_tuple_shape.jl

@@ -389,6 +389,11 @@ Base.isapprox(A::ShapedAsNT, B::ShapedAsNT; kwargs...) = isapprox(_data(A), _dat
 Base.copy(A::ShapedAsNT) = ShapedAsNT(copy(_data(A)),_valshape(A))
 
 
+function Adapt.adapt_structure(to, x::ShapedAsNT)
+    ShapedAsNT(Adapt.adapt(to, _data(x)), _valshape(x))
+end
+
+
 # Required for accumulation during automatic differentiation:
 function Base.:(+)(A::ShapedAsNT{names}, B::ShapedAsNT{names}) where names
     @argcheck _valshape(A) == _valshape(B)
@@ -638,6 +643,11 @@ end
 @inline Tables.rows(A::ShapedAsNTArray) = A
 
 
+function Adapt.adapt_structure(to, x::ShapedAsNTArray)
+    ShapedAsNTArray(Adapt.adapt(to, _data(x)), _elshape(x))
+end
+
+
 const _AnySNTArray{names} = ShapedAsNTArray{<:Union{NamedTuple{names},ShapedAsNT{names}}}
 
 

src/value_shape.jl

@@ -270,6 +270,10 @@ const _InvValueShape = Base.Fix2{typeof(unshaped),<:AbstractValueShape}
     Base.Broadcast.broadcasted(unshaped, xs, Ref(inv_vs.x))
 end
 
+
+InverseFunctions.inverse(vs::AbstractValueShape) = Base.Fix2(unshaped, vs)
+InverseFunctions.inverse(inv_vs::_InvValueShape) = inv_vs.x
+
 const _BroadcastValueShape = Base.Fix1{typeof(broadcast),<:AbstractValueShape}
 const _BroadcastInvValueShape = Base.Fix1{typeof(broadcast),<:_InvValueShape}
 const _BroadcastUnshaped = Base.Fix1{typeof(broadcast),typeof(unshaped)}