Make Base.ifelse a generic function (JuliaLang#37343)

Allow user code to directly extend `Base.ifelse` rather than needing a
special package for it.

base/boot.jl

@@ -192,8 +192,8 @@ export
  Expr, QuoteNode, LineNumberNode, GlobalRef,
  # object model functions
  fieldtype, getfield, setfield!, swapfield!, modifyfield!, replacefield!,
- nfields, throw, tuple, ===, isdefined, eval, ifelse,
- # sizeof # not exported, to avoid conflicting with Base.sizeof
+ nfields, throw, tuple, ===, isdefined, eval,
+ # ifelse, sizeof # not exported, to avoid conflicting with Base
  # type reflection
  <:, typeof, isa, typeassert,
  # method reflection

base/compiler/abstractinterpretation.jl

@@ -1082,7 +1082,7 @@ function abstract_call_builtin(interp::AbstractInterpreter, f::Builtin, (; fargs
  sv::InferenceState, max_methods::Int)
  @nospecialize f
  la = length(argtypes)
- if f === ifelse && fargs isa Vector{Any} && la == 4
+ if f === Core.ifelse && fargs isa Vector{Any} && la == 4
  cnd = argtypes[2]
  if isa(cnd, Conditional)
  newcnd = widenconditional(cnd)

base/compiler/optimize.jl

@@ -162,7 +162,7 @@ const _PURE_BUILTINS = Any[tuple, svec, ===, typeof, nfields]
 const _PURE_OR_ERROR_BUILTINS = [
  fieldtype, apply_type, isa, UnionAll,
  getfield, arrayref, const_arrayref, isdefined, Core.sizeof,
- Core.kwfunc, ifelse, Core._typevar, (<:)
+ Core.kwfunc, Core.ifelse, Core._typevar, (<:)
 ]
 
 const TOP_TUPLE = GlobalRef(Core, :tuple)

base/compiler/tfuncs.jl

@@ -231,7 +231,7 @@ function ifelse_tfunc(@nospecialize(cnd), @nospecialize(x), @nospecialize(y))
  end
  return tmerge(x, y)
 end
-add_tfunc(ifelse, 3, 3, ifelse_tfunc, 1)
+add_tfunc(Core.ifelse, 3, 3, ifelse_tfunc, 1)
 
 function egal_tfunc(@nospecialize(x), @nospecialize(y))
  xx = widenconditional(x)

base/essentials.jl

@@ -477,6 +477,22 @@ Stacktrace:
 """
 sizeof(x) = Core.sizeof(x)
 
+"""
+ ifelse(condition::Bool, x, y)
+
+Return `x` if `condition` is `true`, otherwise return `y`. This differs from `?` or `if` in
+that it is an ordinary function, so all the arguments are evaluated first. In some cases,
+using `ifelse` instead of an `if` statement can eliminate the branch in generated code and
+provide higher performance in tight loops.
+
+# Examples
+```jldoctest
+julia> ifelse(1 > 2, 1, 2)
+2
+```
+"""
+ifelse(condition::Bool, x, y) = Core.ifelse(condition, x, y)
+
 # simple Array{Any} operations needed for bootstrap
 @eval setindex!(A::Array{Any}, @nospecialize(x), i::Int) = arrayset($(Expr(:boundscheck)), A, x, i)
 

base/operators.jl

@@ -429,22 +429,6 @@ const ≥ = >=
 # which is more idiomatic:
 isless(x::Real, y::Real) = x<y
 
-"""
- ifelse(condition::Bool, x, y)
-
-Return `x` if `condition` is `true`, otherwise return `y`. This differs from `?` or `if` in
-that it is an ordinary function, so all the arguments are evaluated first. In some cases,
-using `ifelse` instead of an `if` statement can eliminate the branch in generated code and
-provide higher performance in tight loops.
-
-# Examples
-```jldoctest
-julia> ifelse(1 > 2, 1, 2)
-2
-```
-"""
-ifelse
-
 """
  cmp(x,y)