P1209R0 erase_if(), erase()

[StephanTLavavej]

P1209R0 erase_if(), erase()

[utilForever]

I wanna implement it! 😆

[StephanTLavavej]

Cool 😸 This one is a fairly straightforward conversion of the experimental functions, but we need to decide whether to:

• Outright remove the experimental feature (perhaps only in C++20 mode)
• Deprecate the experimental feature (always with an escape hatch; perhaps only in C++20 mode)
• Allow both to coexist

As the experimental feature isn’t problematic, I think that gentle migration may be best: deprecate only in C++20 mode (where the Standard version is available), with an escape hatch macro.

As this will require some test changes, I would prefer for test/CI to be ready (and the contribution guidelines too) before reviewing this PR. Now that we’re returning from CppCon 2019, this will be our major focus for the next month or two.

[utilForever]

Stephan T. Lavavej (@StephanTLavavej) Is there an example of a code that I can use as a reference when I'm working on this? :)

[StephanTLavavej]

Look at

STL/stl/inc/experimental/vector

Lines 26 to 36 in 957fe99

	// FUNCTION TEMPLATE erase_if
	template <class _Ty, class _Alloc, class _Pr>
	void erase_if(vector<_Ty, _Alloc>& _Cont, _Pr _Pred) { // erase each element satisfying _Pred
	_Erase_remove_if(_Cont, _Pass_fn(_Pred));
	}
	// FUNCTION TEMPLATE erase
	template <class _Ty, class _Alloc, class _Uty>
	void erase(vector<_Ty, _Alloc>& _Cont, const _Uty& _Val) { // erase each element matching _Val
	_Erase_remove(_Cont, _Val);
	}

and the other headers in the experimental directory.

Also look at

STL/stl/inc/yvals_core.h

Line 1 in 957fe99

// yvals_core.h internal header (core)

; this needs to be mentioned as a feature, with a feature-test macro, and deprecation machinery needs to be added.

Finally, test changes will be necessary: for the feature-test macro, to test the new feature alongside the experimental one, and to silence the deprecation for the existing test.

[NathanSWard]

Chris Ohk (@utilForever) This is just a follow up to see if you're still good with implementing this feature? If not, I'd gladly take the torch :)