«

I just wasted an hour until I remembered that the dataset variables in
HTML5 have to be integers. Anyone know why there is this limitation?

I had ptr as a pointer to an <li> element, and couldn't figure out why:

if (ptr.dataset.myflag==true) return;

was not being executed even though myflag had been set to true earlier
(and I used console.info() to confirm it). All started to work as
expected when I changed to use 0 and 1 instead of false and true.

--
Tim

"That excessive bail ought not to be required, nor excessive fines imposed,
nor cruel and unusual punishments inflicted" -- Bill of Rights 1689

2012-06-14 22:31, Tim Streater wrote:

> I just wasted an hour until I remembered that the dataset variables in
> HTML5 have to be integers. Anyone know why there is this limitation?

There is no such limitation.

> if (ptr.dataset.myflag==true) return;
>
> was not being executed even though myflag had been set to true earlier

You cannot set it to the truth value true in HTML markup. If you have

<li data-myflag="true">

or, equivalently,

<li data-myflag=true>

then you are setting the value to the string "true". You can check this
in JavaScript by displaying typeof(ptr.dataset.myflag).

And in JavaScript, even the "==" operator, which tests for an
equality-like relation, does not treat the truth value true and the
string value "true" as identical.

So you would need to test ptr.dataset.myflag == "true" or
ptr.dataset.myflag === "true".

As a different issue, such a way of accessing data-* variables is not
supported by all browsers (see http://caniuse.com/dataset), so it would
be safer to test for

ptr.getAttribute('data-myflag') === "true"

--
Yucca, http://www.cs.tut.fi/~jkorpela/

In article <jrdgfr$pmp$>,
"Jukka K. Korpela" <> wrote:

> 2012-06-14 22:31, Tim Streater wrote:
>
> > I just wasted an hour until I remembered that the dataset variables in
> > HTML5 have to be integers. Anyone know why there is this limitation?
>
> There is no such limitation.

OK, I looked here:

<http://www.w3.org/TR/2009/WD-html5-20090423/dom.html>

(whether this is the correct place I know not) and there is no explicit
mention of types, although I guess, as you say below, that values have
to be settable via HTML, which would preclude boolean values.

> > if (ptr.dataset.myflag==true) return;
> >
> > was not being executed even though myflag had been set to true earlier
>
> You cannot set it to the truth value true in HTML markup. If you have
>
> <li data-myflag="true">
>
> or, equivalently,
>
> <li data-myflag=true>
>
> then you are setting the value to the string "true". You can check this
> in JavaScript by displaying typeof(ptr.dataset.myflag).
>
> And in JavaScript, even the "==" operator, which tests for an
> equality-like relation, does not treat the truth value true and the
> string value "true" as identical.
>
> So you would need to test ptr.dataset.myflag == "true" or
> ptr.dataset.myflag === "true".
>
> As a different issue, such a way of accessing data-* variables is not
> supported by all browsers (see http://caniuse.com/dataset), so it would
> be safer to test for
>
> ptr.getAttribute('data-myflag') === "true"

In fact I'm not setting it in HTML. I'm setting it in JavaScript, as in:

ptr.dataset.myflag = true;

(but now using integers instead, as I mentioned). So I wonder what that
actually sets it to.

I control which browser gets used (Safari, in this case).

--
Tim

"That excessive bail ought not to be required, nor excessive fines imposed,
nor cruel and unusual punishments inflicted" -- Bill of Rights 1689

2012-06-14 23:49, Tim Streater wrote:

> OK, I looked here:
>
> <http://www.w3.org/TR/2009/WD-html5-20090423/dom.html>
>
> (whether this is the correct place I know not)

HTML5 is work in progress, not a specification, but it is better to
consult the newest drafts, like http://www.w3.org/TR/html5/ or the
WHATWG "Living Standard"
http://www.whatwg.org/specs/web-apps...ork/multipage/
(The URL you mention refers to a draft that is over three years old.)

> and there is no explicit
> mention of types, although I guess, as you say below, that values have
> to be settable via HTML, which would preclude boolean values.

When no restriction has been set, an attribute value is any string. HTML
attribute values are always strings from the JavaScript perspective; to
use them as other than string, they need to be type-converted,
implicitly or explicitly.

> In fact I'm not setting it in HTML. I'm setting it in JavaScript, as in:
>
> ptr.dataset.myflag = true;

It seems that the dataset property behaves somewhat oddly: the above
assigns the string "true", not the truth value true. Here, too,
displaying typeof(ptr.dataset.myflag) shows that it is a string.

So it would be better to set it to "true" and do the test accordingly.

--
Yucca, http://www.cs.tut.fi/~jkorpela/

In article <jrdl47$nev$>,
"Jukka K. Korpela" <> wrote:

> 2012-06-14 23:49, Tim Streater wrote:
>
> > OK, I looked here:
> >
> > <http://www.w3.org/TR/2009/WD-html5-20090423/dom.html>
> >
> > (whether this is the correct place I know not)
>
> HTML5 is work in progress, not a specification, but it is better to
> consult the newest drafts, like http://www.w3.org/TR/html5/ or the
> WHATWG "Living Standard"
> http://www.whatwg.org/specs/web-apps...ork/multipage/
> (The URL you mention refers to a draft that is over three years old.)

Well I knew *that* , but it was late and it was the best I could come
up with quickly.

> > and there is no explicit
> > mention of types, although I guess, as you say below, that values have
> > to be settable via HTML, which would preclude boolean values.
>
> When no restriction has been set, an attribute value is any string. HTML
> attribute values are always strings from the JavaScript perspective; to
> use them as other than string, they need to be type-converted,
> implicitly or explicitly.
>
> > In fact I'm not setting it in HTML. I'm setting it in JavaScript, as in:
> >
> > ptr.dataset.myflag = true;
>
> It seems that the dataset property behaves somewhat oddly: the above
> assigns the string "true", not the truth value true. Here, too,
> displaying typeof(ptr.dataset.myflag) shows that it is a string.

Well I'll be hornswoggled.

> So it would be better to set it to "true" and do the test accordingly.

Er possibly but I'm done with that code for the moment. Anyway - thanks
for the useful clarifications.

--
Tim

"That excessive bail ought not to be required, nor excessive fines imposed,
nor cruel and unusual punishments inflicted" -- Bill of Rights 1689