Bug 56001 – <meter> should only support horizontal indicator

• Component: DOM
• Status: RESOLVED FIXED

[Hajime Morrita]

Current <meter> change its appearance based on its dimension.
But such kind of style-determined-by-layout behavior is tricky to implement 
and may (and did) cause some layout instability.
(See https://lists.webkit.org/pipermail/webkit-dev/2011-March/016140.html)

Instead of this dynamic appearance switching, WebKit should provide the way to
choose the visual direction by styles.

Because text-direction is used to switch its visual between rtl and ltr, 
writing-mode looks a good candidate for the knob.

Why don't provide two separate -webkit-appearance (like meter-vertical and meter-horizontal)?
Because WebKit allows user-styling for form controls via pseudo classes like -webkit-meter, 
and this user-style is enabled by specifying "-webkit-appearance: none". 
Here we already use -webkit-appearance, thus cannot give meter-vertical appearance for it.

[Dimitri Glazkov (Google)]

What do you think about adding a whole new CSS property? writing-mode still seems to say not quite what we mean.

[Simon Fraser (smfr)]

I don't think writing-mode is the right property to use.

[Hajime Morrita]

Created attachment 85292 [details]
Patch

[Hajime Morrita]

Hi Dimitri, could you take a look?
This is a preparation for making <meter> new-shadowish.

[Dimitri Glazkov (Google)]

(In reply to comment #4)
> Hi Dimitri, could you take a look?
> This is a preparation for making <meter> new-shadowish.

Given that Simon also thinks that writing-mode is the right approach, perhaps we should have a quick discussion first.

[Dimitri Glazkov (Google)]

(In reply to comment #5)
> (In reply to comment #4)
> > Hi Dimitri, could you take a look?
> > This is a preparation for making <meter> new-shadowish.
> 
> Given that Simon also thinks that writing-mode is NOT the right approach, perhaps we should have a quick discussion first.

[Dave Hyatt]

Yeah writing-mode is really not the correct approach here.

[Dave Hyatt]

Comment on attachment 85292 [details]
Patch

Can't use writing-mode like this.  That isn't what it is for at all.

[Dave Hyatt]

Meters orientation should just have to be explicitly specified IMO.  We need to just get that fixed in the standard.

[Dimitri Glazkov (Google)]

like <meter orientation="vertical">?

[Dimitri Glazkov (Google)]

(In reply to comment #9)
> Meters orientation should just have to be explicitly specified IMO.  We need to just get that fixed in the standard.

After talking with Hixie, he's suggesting adding "-webkit-direction" (or similar) CSS property instead. WDYT?

[Ian 'Hixie' Hickson]

The auto-orientation capability is similar to how Cocoa's NSSlider control automatically picks the best orientation (see its isVertical API). It increases the odds that people will make pretty pages.

I think -webkit-direction should have three values, 'auto' (the initial value, implementing the spec), and 'horizontal' and 'vertical' (which force a particular direction).

Maybe -webkit-control-direction to avoid confusion with 'direction'.

[Hajime Morrita]

Sorry for coming late and making the patch before reading simon's comment. (I just missed it...)

Okay, I'm going to introduce -webkit-control-direction.
The original intention to make direction explicit is that eliminate "auto"-like behavior
because it requires layout to determine direction, that causes complicated situation.
So I think it's OK to make the default value "horizontal" because 
there is no vertical meter indicator for native controls.

[Dimitri Glazkov (Google)]

(In reply to comment #13)
> Sorry for coming late and making the patch before reading simon's comment. (I just missed it...)
> 
> Okay, I'm going to introduce -webkit-control-direction.
> The original intention to make direction explicit is that eliminate "auto"-like behavior
> because it requires layout to determine direction, that causes complicated situation.
> So I think it's OK to make the default value "horizontal" because 
> there is no vertical meter indicator for native controls.

Hang on, hang on -- the discussion isn't over yet. Let Hyatt chime in.

[Simon Fraser (smfr)]

FYI I would prefer the name to contain 'orientation' rather than 'direction'. I'm still not sure that a normal property is the correct approach here. Maybe a pseudoclass?

[Dimitri Glazkov (Google)]

(In reply to comment #15)
> FYI I would prefer the name to contain 'orientation' rather than 'direction'. I'm still not sure that a normal property is the correct approach here. Maybe a pseudoclass?

[Dimitri Glazkov (Google)]

(In reply to comment #16)
> (In reply to comment #15)
> > FYI I would prefer the name to contain 'orientation' rather than 'direction'. I'm still not sure that a normal property is the correct approach here. Maybe a pseudoclass?

(In reply to comment #15)
> FYI I would prefer the name to contain 'orientation' rather than 'direction'. I'm still not sure that a normal property is the correct approach here. Maybe a pseudoclass?

Sorry for double-posting.  A pseudoclass wouldn't be enough, since it doesn't allow us to set the orientation by itself. Or perhaps I am misunderstanding... Can you pseudocode the pseudoclass idea for me?

[Hajime Morrita]

How about to ask whatwg@ for dropping vertical meter from the spec?
There is no vertical meter/indicator on any platform and 
Opera, which is the only browser that supports <meter> other than WebKit, also doesn't support it.

[Dimitri Glazkov (Google)]

(In reply to comment #18)
> How about to ask whatwg@ for dropping vertical meter from the spec?
> There is no vertical meter/indicator on any platform and 
> Opera, which is the only browser that supports <meter> other than WebKit, also doesn't support it.

Let's go with this notion for now. Implement horizontal only. We'll figure out what to do with spec later.

[Hajime Morrita]

Created attachment 87660 [details]
Patch

[Hajime Morrita]

Hi Dimitri, thank you for the suggestion!
> Let's go with this notion for now. Implement horizontal only. We'll figure out what to do with spec later.

So I removed vertical related code. It got simpler now.

[Dimitri Glazkov (Google)]

Comment on attachment 87660 [details]
Patch

this should become even simpler once you switch it to the shadow DOM.

[Hajime Morrita]

Created attachment 87746 [details]
Patch

[Hajime Morrita]

Comment on attachment 87746 [details]
Patch

Oops. I intended to land it...

[Hajime Morrita]

Committed r82589: <http://trac.webkit.org/changeset/82589>

[WebKit Review Bot]

http://trac.webkit.org/changeset/82589 might have broken Qt Linux Release
The following tests are not passing:
fast/dom/HTMLMeterElement/meter-boundary-values.html
fast/dom/HTMLMeterElement/meter-optimums.html
fast/dom/HTMLMeterElement/meter-styles-changing-pseudo.html
fast/dom/HTMLMeterElement/meter-styles.html

[Adam Barth]

I'm rolling out this patch because it turned many bots red and the author did not appear to be available to help repair them.

[Hajime Morrita]

Created attachment 87863 [details]
Added skip list entries that need expectation update.

[Dimitri Glazkov (Google)]

Comment on attachment 87863 [details]
Added skip list entries that need expectation update.

let's try again.

[Hajime Morrita]

Committed r82686: <http://trac.webkit.org/changeset/82686>

[WebKit Review Bot]

http://trac.webkit.org/changeset/82686 might have broken Windows 7 Release (Tests)
The following tests are not passing:
fast/dom/HTMLMeterElement/meter-appearances-capacity.html
fast/dom/HTMLMeterElement/meter-appearances-rating-relevancy.html