Bug 115102 – Use std::move in AtomicString

• Status: RESOLVED
• Resolution: FIXED
• Component: Web Template Framework

[Mikhail Pozdnyakov]

SSIA.

[Mikhail Pozdnyakov]

Created attachment 199418 [details]
WIP

Newly added file is not yet included into proj files.

[Anders Carlsson]

I like it, but I think it should just go into TypeTraits.h directly.

[Mikhail Pozdnyakov]

Created attachment 199454 [details]
patch

Put 'move()' to TypeTraits.h as requested by Anders.

[Alexey Proskuryakov]

What is the difference with std::move? I'm wondering if we can just use it directly.

[Benjamin Poulain]

(In reply to comment #4)
> What is the difference with std::move? I'm wondering if we can just use it directly.

Same question here.

[Anders Carlsson]

Good point. We could just state that any C++ compiles that supports rvalue references must also have an implementation of std::move.

[Benjamin Poulain]

Comment on attachment 199454 [details]
patch

Looks like we should use std::move in AtomicString instead of introducing our own.

[Mikhail Pozdnyakov]

(In reply to comment #5)
> (In reply to comment #4)
> > What is the difference with std::move? I'm wondering if we can just use it directly.
> 
> Same question here.

Ok, I was just thinking that one of the reasons for having WTF is that we do not want to use STL, this assumption seems to be wrong however.

[Benjamin Poulain]

> Ok, I was just thinking that one of the reasons for having WTF is that we do not want to use STL, this assumption seems to be wrong however.

The STL is pretty crappy when it come to ADT. Apart from that, I don't think there is any problem with using STL.

[Mikhail Pozdnyakov]

Created attachment 199631 [details]
patch v3

Use std::move.

[Mikhail Pozdnyakov]

(In reply to comment #9)
> > Ok, I was just thinking that one of the reasons for having WTF is that we do not want to use STL, this assumption seems to be wrong however.
> 
> The STL is pretty crappy when it come to ADT. Apart from that, I don't think there is any problem with using STL.

btw, why do we use WTF TypeTraits? could we use std::type_traits as well?

[Anders Carlsson]

Comment on attachment 199631 [details]
patch v3

View in context: https://bugs.webkit.org/attachment.cgi?id=199631&action=review

> Source/WTF/wtf/text/AtomicString.h:26
> +#if COMPILER_SUPPORTS(CXX_RVALUE_REFERENCES)
> +#include <utility>
> +#endif

I think we can just always include this.

[Mikhail Pozdnyakov]

Created attachment 199660 [details]
to be landed

[Anders Carlsson]

(In reply to comment #11)
> (In reply to comment #9)
> > > Ok, I was just thinking that one of the reasons for having WTF is that we do not want to use STL, this assumption seems to be wrong however.
> > 
> > The STL is pretty crappy when it come to ADT. Apart from that, I don't think there is any problem with using STL.

I think using STL for type traits is a great idea!

> 
> btw, why do we use WTF TypeTraits? could we use std::type_traits as well?

I believe that the STL type traits are C++11 only (but we could change the WTF type traits to emulate them when building with C++98).

[WebKit Commit Bot]

The commit-queue encountered the following flaky tests while processing attachment 199660 [details]:

svg/as-image/img-relative-height.html bug 114140 (author: [e-mail])
The commit-queue is continuing to process your patch.

[WebKit Commit Bot]

Comment on attachment 199660 [details]
to be landed

Clearing flags on attachment: 199660

Committed r149112: <http://trac.webkit.org/changeset/149112>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Alexey Proskuryakov]

> Ok, I was just thinking that one of the reasons for having WTF is that we do not want to use STL, this assumption seems to be wrong however.

The main reasons for reimplementing some containers and algorithms are:

1. We can fine tune performance to our needs.

2. We can disable exceptions and still have reasonable behavior.