Hi,

On 20/06/2022 17:39, Thiago Macieira wrote:
On Monday, 20 June 2022 07:46:57 PDT Giuseppe D'Angelo via Development wrote:
A fancy name for: "if a function/class is operating on a rvalue, should
it store a copy of it in order to keep it alive?". Consider
In other words, remove the rvalue reference and store a copy from the const-
lvalue reference overload you already have.

Well, not necessarily. Say the input is a lvalue std::u16string; you don't want the tokenizer to unconditionally copy it.
So it just takes a reference, and should continue to do so.

(Looking at std::views::all: this is the case where the input is a non-view lvalue, so you'll wrap it in ref_view -- i.e. merely hold a reference to it.)


std::u16string someLongString = u"...";
auto tokenizer = QStringTokenizer(someLongString, u"X");
For this reason QStringTokenizer moves and stores the input if it's an
rvalue, but only keeps a reference if it's an lvalue.
This means QStringTokenizer must have a QString member and a QString & member.
The simplest implementation removes one of them.

Not always, see above.


QString s = getString();
auto tok = QStringTokenizer(s, u"x");
would take a copy (given `s` is a view after [4]), while now it only
takes a reference.
Yup.

Opinions?
My only objection is to calling this by a fancy name, "rvalue pinning". Simply
call what it is: take all parameters by const-lvalue and never store a
reference.

It's also imprecise. We should use the C++20 range terminology which is more accurate, but I don't want to cause even more confusion (given we don't use that logic, and probably _can't_ use the logic just yet).


Thanks,
--
Giuseppe D'Angelo | [email protected] | Senior Software Engineer
KDAB (France) S.A.S., a KDAB Group company
Tel. France +33 (0)4 90 84 08 53, http://www.kdab.com
KDAB - The Qt, C++ and OpenGL Experts

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

_______________________________________________
Development mailing list
[email protected]
https://lists.qt-project.org/listinfo/development

Reply via email to