Re: [[nodiscard]] and KDE coding policy

> > > Yeah, I don't think getters need `[[nodiscard]]`, for the same reason
> > > that
> > > `std::min` and `std::move` don't have it:

Do not assume that because most std:: functions don't have it, that
users should be discouraged to use it.

There is a nice comment by Jonathan Wakely (ISO C++ committee member,
maintainer of libstdc++) of the current status of [[nodiscard]] in the=20
standard:

https://github.com/isocpp/CppCoreGuidelines/issues/2109

=2D-
That's not because the standard thinks it shouldn't be used elsewhere. Ther=
e=20
have been several proposals to add it to loads of functions in the standard=
=20
library, but I (and others) argued strongly against that, because:

=2D Specifying it in the standard has no normative effect, implementations =
are=20
allowed to ignore the attribute, and even if they don't ignore it, they're =
not=20
required to issue a diagnostic.
=2D Specifying it in a few places in the standard could be interpreted as s=
aying=20
it shouldn't be used elsewhere. That could result in users reporting bogus=
=20
bugs about nodiscard warnings from other functions.
=2D Specifying it in the standard would cause churn for implementation vend=
ors=20
who do want to add it where the standard "requires", to avoid complaints ab=
out=20
being non-conforming.
=2D Reviewing and approving those proposals would have eaten up hundreds of=
=20
person hours in the committee, which should be spent on far more productive=
=20
work.
=2D It can (and IMHO should) be left to implementation vendors to just Do T=
he=20
Right Thing.

And the Right Thing is to use it widely, everywhere that it makes sense.

IMHO compilers should warn about discarded values for all comparison=20
operators, all begin and end accessors, nearly all const member functions,=
=20
etc. etc. because the chances that a =3D=3D b; as an entire statement is a =
typo=20
for =3D is much higher than the chances that the developer really wanted to=
=20
compare two things just for the thrill of it. Compilers already warn about=
=20
this for fundamental types, so it makes sense to do so for strings, contain=
ers=20
etc. too. But unfortunately they're not smart enough to do it automatically=
,=20
so it is useful to add [[nodiscard]] to them.* And std::lib implementations=
 do=20
that, without needing the standard to tell them to.

tl;dr do not assume from the absence of [[nodiscard]] attributes in the=20
standard specification that the committee thinks it should be used sparingl=
y.=20
The opposite is true.
=2D-

Cheers,
Ivan

=2D-=20
Dr Ivan =C4=8Cuki=C4=87
[email protected], https://cukic.co/
gpg key fingerprint: 8FE4 D32F 7061 EA9C 8232 07AE 01C6 CE2B FF04 1C12