Merge enforceEx with enforce

[mihails-strasuns]

Makes it possible to call enforce!MyException instead of enforceEx!MyException, eliminating
redundant notation. Old enforceEx is not deprecated but removed from documentation examples.

This fixes one of very common Phobos "WAT" moments for me - having separate name for semantically identical
function with a different argument set.

[burner]

The patch looks very boring, I like it.

One question though, would their be any benefit of making the file and line template arguments like so:

T enforce(E : Throwable = Exception, T, string file = __FILE__, size_t line = __LINE__)(T value, lazy const(char)[] msg = null)

and passing it like this all the way to bailOut. I would image the compiler could take some advantage of that.

[dnadlinger]

No. It would only lead to a gazillion of duplicate enforce functions being created, and with longer mangled names.

[burner]

thank you for the quick reply, I see my mistake

[dnadlinger]

I believe we had it like this at one point, though, so it's not a completely outlandish suggestion. ;)

[quickfur]

LGTM.

[dnadlinger]

Auto-merge toggled on

[dnadlinger]

Okay, let's go ahead with this. There might still be a small chance of breakage if somebody used enforce in an obscure way, but we'll find out during the next beta.

[MartinNowak]

Okay, let's go ahead with this. There might still be a small chance of breakage if somebody used enforce in an obscure way, but we'll find out during the next beta.

Anyone complained?
It would have been possible to add an overload for cases where someone explicitly named the return type.
Also why wasn't enforceEx deprecated?

[mihails-strasuns]

Also why wasn't enforceEx deprecated?

I remember lot of negative opinions about breaking changes in Phobos for the sake of just naming - wanted to increase the chances of this PR to be merged :) I think we can try deprecating it in next release if no one is against it.

[mihails-strasuns]

(I don't know of any complaints)

[MartinNowak]

Deprecation isn't breaking and we should still deprecate enforcers, starting by removing the documentation.

[MartinNowak]

It would have been possible to add an overload for cases where someone explicitly named the return type.

What about the overload, this change as is might break code without warning.

[CyberShadow]

This pull request introduced an accepts-invalid regression:
https://issues.dlang.org/show_bug.cgi?id=14685