1039521 - nsTArray: Add an AppendElement take taking a r-value reference

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Jeff Muizelaar [:jrmuizel]]

Attachment: Working implementation

Here's a patch that implements this. I wasn't able to get
nsTArrayElementTraits<...>::Construct(iter, mozilla::Move(aItem));
to work so I just manually construct it. What's the correct way to do this?

Comment 1

[Trevor Saunders (:tbsaunde)]

khuey wrote a patch for this in bug 982212, but at least part of it foundered on the rocks of gcc 4.4.  Does your patch do better there?

Comment 2

[Jeff Muizelaar [:jrmuizel]]

https://tbpl.mozilla.org/?tree=Try&rev=0ac136ca5cc5

Comment 3

[Jeff Muizelaar [:jrmuizel]]

https://tbpl.mozilla.org/?tree=Try&rev=559cc40cf9d9

Comment 4

[Jeff Muizelaar [:jrmuizel]]

Attachment: nsTArray: Add an AppendElement take taking a r-value reference

This builds on all platforms on try.

Comment 5

[Nathan Froyd [:froydnj]]

Comment on attachment 8459698 [details] [diff] [review]
nsTArray: Add an AppendElement take taking a r-value reference

Review of attachment 8459698 [details] [diff] [review]:
-----------------------------------------------------------------

r=me with the change below.

::: xpcom/glue/nsTArray.h
@@ +1255,5 @@
>      return AppendElements(&aItem, 1);
>    }
>  
> +  // A variation on the AppendElements method defined above.
> +  elem_type* AppendElement(elem_type&& aItem)

Seems like this should be templated over |class Item| like all the other AppendElement methods.

@@ +1262,5 @@
> +                                                sizeof(elem_type))))
> +      return nullptr;
> +    index_type len = Length();
> +    elem_type* iter = Elements() + len;
> +    nsTArrayElementTraits<elem_type>::Construct(iter, mozilla::Forward<elem_type>(aItem));

...and we'd Forward<Item> here.

Comment 6

[Nathan Froyd [:froydnj]]

Comment on attachment 8459698 [details] [diff] [review]
nsTArray: Add an AppendElement take taking a r-value reference

Review of attachment 8459698 [details] [diff] [review]:
-----------------------------------------------------------------

::: xpcom/glue/nsTArray.h
@@ +1260,5 @@
> +  {
> +    if (!Alloc::Successful(this->EnsureCapacity(Length() + 1,
> +                                                sizeof(elem_type))))
> +      return nullptr;
> +    index_type len = Length();

Seems a little odd to have this named here for a single use, but calling Length() in the EnsureCapacity call earlier.

Comment 7

[Jeff Muizelaar [:jrmuizel]]

(In reply to Nathan Froyd (:froydnj) from comment #5)
> Comment on attachment 8459698 [details] [diff] [review]
> nsTArray: Add an AppendElement take taking a r-value reference
> 
> Review of attachment 8459698 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> r=me with the change below.
> 
> ::: xpcom/glue/nsTArray.h
> @@ +1255,5 @@
> >      return AppendElements(&aItem, 1);
> >    }
> >  
> > +  // A variation on the AppendElements method defined above.
> > +  elem_type* AppendElement(elem_type&& aItem)
> 
> Seems like this should be templated over |class Item| like all the other
> AppendElement methods.

It seems like the templating over class Item is to allow implicit conversions from things that convert to elem_type. Adding that templating causes the build to fail on some compilers like msvc and older gcc. I'm also not sure it makes sense to accept an r-value reference of something that we'd need to convert.

Comment 8

[Jeff Muizelaar [:jrmuizel]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #7)
> 
> It seems like the templating over class Item is to allow implicit
> conversions from things that convert to elem_type. Adding that templating
> causes the build to fail on some compilers like msvc and older gcc. I'm also
> not sure it makes sense to accept an r-value reference of something that
> we'd need to convert.

You can see the compiler problems here with an older version:
https://tbpl.mozilla.org/?tree=Try&rev=0ac136ca5cc5

Comment 9

[Nathan Froyd [:froydnj]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #7)
> It seems like the templating over class Item is to allow implicit
> conversions from things that convert to elem_type. Adding that templating
> causes the build to fail on some compilers like msvc and older gcc. I'm also
> not sure it makes sense to accept an r-value reference of something that
> we'd need to convert.

OK, then.  Patch without templating is fine.

Comment 10

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/db1ceb8f8d92