Inconsistent behaviour of `hasproperty` with Modules

[tecosaur]

Julia 1.8.2, Linux

I feel like the behavior of {has,get}{property,field} is a bit suspect when it comes to Modules. The following example should be sufficiently illustrative.

julia> hasproperty(Main, :DataFrames)
false

julia> getproperty(Main, :DataFrames)
DataFrames

julia> hasfield(Main, :DataFrames)
ERROR: MethodError: no method matching hasfield(::Module, ::Symbol)
Closest candidates are:
  hasfield(::Type, ::Symbol) at reflection.jl:215
Stacktrace:
 [1] top-level scope
   @ REPL[26]:1

julia> getfield(Main, :DataFrames)
DataFrames

[jishnub]

Related: https://github.com/JuliaLang/julia/issues/46857

[KristofferC]

For modules, I think it makes more sense to use isdefined.

[oxinabox]

I feel liked getfield has always been a bit of a pun when used on modules. But while we support it, it kins of seems like we should complete the metaphor with the others from {has,get}{property,field} Or deprecate it and provide an alternative

[simeonschaub]

Or deprecate it and provide an alternative

That's basically what #44231 did and why we now have getglobal. We don't really want to add a loud deprecation warning though since it's too commonly used and the benefits of the change aren't worth the annoyance of a ton of warnings

[oxinabox]

Makes sense. We really need to sort out deprecating warnings that warn only for deprecations that are directly your fault.

IMO we can close this as resolved

[ericphanson]

I don't think this is resolved since getglobal only replaces getfield. The docstring for getglobal recommends getproperty, so getproperty definitely doesn't seem to be deprecated, but getproperty is still in contradiction to hasproperty for modules, so there is still an issue.

[ericphanson]

it is valid to add hasproperty(mod::Module, x::Symbol) = isdefined(mod, x) ? That could at least bring hasproperty and getproperty into alignment.

Or should getproperty be deprecated for modules also? (e.g. at least remove it from the getglobal docstring)

[vtjnash]

Seems valid, though I think it may be isresolved and ispublic?