Prevent spurious error hints

[penelopeysm]

The error hint for MethodError (see #146) is indiscriminate in that it catches any MethodError. We can't do much about that, but we can try to restrict it to MethodErrors which only involve one of the arguments being a LogDensityModel.

AbstractMCMC.jl/src/AbstractMCMC.jl

Lines 123 to 135 in ececa17

	if isdefined(Base.Experimental, :register_error_hint)
	function __init__()
	Base.Experimental.register_error_hint(MethodError) do io, exc, argtypes, _
	if Base.parentmodule(exc.f) == LogDensityProblems &&
	any(a -> a <: LogDensityModel, argtypes)
	print(
	io,
	"\n`AbstractMCMC.LogDensityModel` is a wrapper and does not itself implement the LogDensityProblems.jl interface. To use LogDensityProblems.jl methods, access the inner type with (e.g.) `logdensity(model.logdensity, params)` instead of `logdensity(model, params)`.",
	)
	end
	end
	end
	end

Unfortunately, Base.parentmodule is not defined for callable structs, so if you called a callable struct with the wrong arguments and you happened to have AbstractMCMC in your environment, this error hint would try to take over and would itself error. I discovered this when calling (::DynamicPPL.Model)(args...) with the wrong combination of arguments.

This PR fixes it by (1) reversing the order of checks, so that it shortcircuits if none of the arguments are LogDensityModels; and also (2) explicitly checking that exc.f is a function, because the things we want to catch are LogDensityProblems.foo which are indeed functions. Both together might be a bit overkill, but anyway.

[github-actions]

AbstractMCMC.jl documentation for PR #164 is available at:
https://TuringLang.github.io/AbstractMCMC.jl/previews/PR164/

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 0.00%. Comparing base (ececa17) to head (20c0960).
Report is 1 commits behind head on main.

Additional details and impacted files

@@     Coverage Diff     @@
##   main   #164   +/-   ##
===========================
===========================

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.