More generic safe similar (#550)

* More generic safe similar

* Fix docs failure

* Fix typo in docs

* Restrict type of T

---------

Co-authored-by: Christopher Rackauckas <[email protected]>

Author: ErikQQY

docs/src/basics/diagnostics_api.md

@@ -20,7 +20,7 @@ solve process. This is controlled by 3 keyword arguments to `solve`:
 
 All the native NonlinearSolve.jl algorithms come with in-built
 [TimerOutputs.jl](https://github.com/KristofferC/TimerOutputs.jl) support. However, this
-is disabled by default and can be enabled via [`NonlinearSolve.enable_timer_outputs`](@ref).
+is disabled by default and can be enabled via [`NonlinearSolveBase.enable_timer_outputs`](@ref).
 
 Note that you will have to restart Julia to disable the timer outputs once enabled.
 

docs/src/basics/faq.md

@@ -71,29 +71,29 @@ differentiate the function based on the input types. However, this function has
 `xx = [1.0, 2.0, 3.0, 4.0]` followed by a `xx[1] = var[1] - v_true[1]` where `var` might
 be a Dual number. This causes the error. To fix it:
 
-1. Specify the `autodiff` to be `AutoFiniteDiff`
-
-   ```@example dual_error_faq
-   sol = solve(prob_oop, LevenbergMarquardt(; autodiff = AutoFiniteDiff());
-       maxiters = 10000, abstol = 1e-8)
-   ```
-
-   This worked but, Finite Differencing is not the recommended approach in any scenario.
-
-2. Rewrite the function to use
-   [PreallocationTools.jl](https://github.com/SciML/PreallocationTools.jl) or write it as
-
-   ```@example dual_error_faq
-   function fff_correct(var, p)
-       v_true = [1.0, 0.1, 2.0, 0.5]
-       xx = eltype(var)[1.0, 2.0, 3.0, 4.0]
-       xx[1] = var[1] - v_true[1]
-       return xx - v_true
-   end
-
-   prob_oop = NonlinearLeastSquaresProblem{false}(fff_correct, v_init)
-   sol = solve(prob_oop, LevenbergMarquardt(); maxiters = 10000, abstol = 1e-8)
-   ```
+ 1. Specify the `autodiff` to be `AutoFiniteDiff`
+    
+    ```@example dual_error_faq
+    sol = solve(prob_oop, LevenbergMarquardt(; autodiff = AutoFiniteDiff());
+        maxiters = 10000, abstol = 1e-8)
+    ```
+    
+    This worked but, Finite Differencing is not the recommended approach in any scenario.
+
+ 2. Rewrite the function to use
+    [PreallocationTools.jl](https://github.com/SciML/PreallocationTools.jl) or write it as
+    
+    ```@example dual_error_faq
+    function fff_correct(var, p)
+        v_true = [1.0, 0.1, 2.0, 0.5]
+        xx = eltype(var)[1.0, 2.0, 3.0, 4.0]
+        xx[1] = var[1] - v_true[1]
+        return xx - v_true
+    end
+    
+    prob_oop = NonlinearLeastSquaresProblem{false}(fff_correct, v_init)
+    sol = solve(prob_oop, LevenbergMarquardt(); maxiters = 10000, abstol = 1e-8)
+    ```
 
 ## I thought NonlinearSolve.jl was type-stable and fast. But it isn't, why?
 

docs/src/tutorials/iterator_interface.md

@@ -20,7 +20,7 @@ to iterate the solver.
 The iterator interface supports:
 
 ```@docs
-step!(nlcache::NonlinearSolve.AbstractNonlinearSolveCache, args...; kwargs...)
+step!(nlcache::NonlinearSolveBase.AbstractNonlinearSolveCache, args...; kwargs...)
 ```
 
 We can perform 10 steps of the Newton-Raphson solver with the following:

lib/NonlinearSolveBase/src/NonlinearSolveBase.jl

@@ -19,7 +19,8 @@ using RecursiveArrayTools: AbstractVectorOfArray, ArrayPartition
 using SciMLBase: SciMLBase, ReturnCode, AbstractODEIntegrator, AbstractNonlinearProblem,
                  AbstractNonlinearAlgorithm, AbstractNonlinearFunction,
                  NonlinearProblem, NonlinearLeastSquaresProblem, StandardNonlinearProblem,
-                 NonlinearFunction, NullParameters, NLStats, LinearProblem, LinearAliasSpecifier
+                 NonlinearFunction, NullParameters, NLStats, LinearProblem,
+                 LinearAliasSpecifier
 using SciMLJacobianOperators: JacobianOperator, StatefulJacobianOperator
 using SciMLOperators: AbstractSciMLOperator, IdentityOperator
 using SymbolicIndexingInterface: SymbolicIndexingInterface

lib/NonlinearSolveBase/src/jacobian.jl

@@ -41,7 +41,7 @@ function construct_jacobian_cache(
                           (linsolve === nothing || needs_concrete_A(linsolve)))
     needs_jac = linsolve_needs_jac || concrete_jac(alg)
 
-    @bb fu_cache = similar(fu)
+    fu_cache = Utils.safe_similar(fu)
 
     if !has_analytic_jac && needs_jac
         if autodiff === nothing
@@ -80,9 +80,9 @@ function construct_jacobian_cache(
             end
         else
             if eltype(f.jac_prototype) <: Bool
-                similar(f.jac_prototype, promote_type(eltype(fu), eltype(u)))
+                Utils.safe_similar(f.jac_prototype, promote_type(eltype(fu), eltype(u)))
             else
-                similar(f.jac_prototype)
+                Utils.safe_similar(f.jac_prototype)
             end
         end
     end

lib/NonlinearSolveBase/src/linear_solve.jl

@@ -74,7 +74,8 @@ function construct_linear_solver(alg, linsolve, A, b, u; stats, kwargs...)
     linprob = LinearProblem(A, b; u0 = u_cache, kwargs...)
 
     # unlias here, we will later use these as caches
-    lincache = init(linprob, linsolve; alias = LinearAliasSpecifier(alias_A = false, alias_b = false))
+    lincache = init(
+        linprob, linsolve; alias = LinearAliasSpecifier(alias_A = false, alias_b = false))
     return LinearSolveJLCache(lincache, linsolve, nothing, stats)
 end
 

lib/NonlinearSolveBase/src/utils.jl

@@ -101,13 +101,13 @@ restructure(y, x) = ArrayInterface.restructure(y, x)
 
 function safe_similar(x, args...; kwargs...)
     y = similar(x, args...; kwargs...)
-    return init_bigfloat_array!!(y)
+    return init_similar_array!!(y)
 end
 
-init_bigfloat_array!!(x) = x
+init_similar_array!!(x) = x
 
-function init_bigfloat_array!!(x::AbstractArray{<:BigFloat})
-    ArrayInterface.can_setindex(x) && fill!(x, BigFloat(0))
+function init_similar_array!!(x::AbstractArray{<:T}) where {T <: Number}
+    ArrayInterface.can_setindex(x) && fill!(x, T(0))
     return x
 end
 

lib/SimpleNonlinearSolve/src/utils.jl

@@ -160,7 +160,7 @@ end
 
 function compute_hvvp(prob, autodiff, _, x::Number, dir::Number)
     H = DI.second_derivative(prob.f, autodiff, x, Constant(prob.p))
-    return H*dir
+    return H * dir
 end
 function compute_hvvp(prob, autodiff, fx, x, dir)
     jvp_fn = if SciMLBase.isinplace(prob)

test/qa_tests.jl

@@ -1,7 +1,8 @@
 @testitem "Aqua" tags=[:misc] begin
     using NonlinearSolve, SimpleNonlinearSolve, Aqua
 
-    Aqua.test_all(NonlinearSolve; ambiguities = false, piracies = false, stale_deps = false, deps_compat = false)
+    Aqua.test_all(NonlinearSolve; ambiguities = false, piracies = false,
+        stale_deps = false, deps_compat = false)
     Aqua.test_ambiguities(NonlinearSolve; recursive = false)
     Aqua.test_stale_deps(SimpleNonlinearSolve; ignore = [:SciMLJacobianOperators])
     Aqua.test_deps_compat(SimpleNonlinearSolve; ignore = [:SciMLJacobianOperators])