Bug in rendering tal:attributes in case of boolean values?

[gh-PonyM]

We discovered a changement in rendering of boolean attributes in the option tag since release 3.8.0 (using python 3.8).
Example:
<select> <option value="" tal:attributes="selected True">Selected</option> <option value="a" tal:attributes="selected False">Not selected</option> <option value="ok" tal:attributes="selected None">OK</option> </select>
will render to:
<select> <option value="" selected="True">Selected</option> <option value="a" selected="False">Not selected</option> <option value="ok">OK</option> </select>
Can somebody confirm this?
The workaround for now would be to pin down the version to 3.7.4.

[malthe]

You need to configure boolean_attributes. It's a template option that takes a set of attributes, for example "selected", "checked", etc.

For an example, see https://github.com/zopefoundation/z3c.pt/blob/master/src/z3c/pt/pagetemplate.py#L40.

[stevepiercy]

@DadaDaMotha I confirm the same behavior. Chameleon 3.7.4 does not exhibit this behavior. I have pinned my dependencies to Chameleon < 3.8.0.

In Deform, all CheckboxChoiceWidgets (and probably other checkbox widgets) render a checked attribute regardless of its actual value. If the value is False, it should not render a checked attribute.

I'll follow up after I deal with the fire drill this change caused.

[malthe]

@stevepiercy let me know how you fare – it should be relatively straight-forward to get those boolean_attributes in there, after which the behavior should essentially be back to normal.

[stevepiercy]

Ha, you assume I know what I'm doing! 😬

I honestly don't know whether to look first in Deform or pyramid_chameleon. I spend more time staring blankly at the code and drooling than understanding it. Ah, well, no time like the present to start learning.

[stevepiercy]

My first guess was to scan for any imports of chameleon in either library. I then tried to understand code for each result, but I'm failing to comprehend it.

[stevepiercy]

Our issues were fixed by Pylons/deform#418

[malthe]

Superb!

…

On Sun, 28 Jun 2020 at 03.38, Steve Piercy ***@***.***> wrote: Our issues were fixed by Pylons/deform#418 <Pylons/deform#418> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#318 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAGOJMYH34DSF6VKPZF4JDRY2NLDANCNFSM4OJFFBQA> .

[gh-PonyM]

Dear maintainers,
By now, I don't think this issue is fixed. By using the BOOLEAN_HTML_ATTRIBUTES in https://github.com/zopefoundation/z3c.pt/blob/master/src/z3c/pt/pagetemplate.py#L40 like:

from chameleon.zpt.template import PageTemplate
ts = "<select> <option value="1" tal:attributes="selected True">Selected</option></select>"
page = PageTemplate(ts, boolean_attributes=BOOLEAN_HTML_ATTRS)()

page will yield <option value="1" selected="selected">Selected</option>
Our custom integration shows the same...

[malthe]

I am not sure I understand. What is the expected output?

[gh-PonyM]

Hi malthe, thx for your reply. I would expect <option value="1" selected">Selected</option>.

[mauritsvanrees]

[Edit: our replies crossed.]

The comment in z3c.pt says about BOOLEAN_HTML_ATTRS:

List of Boolean attributes in HTML that should be rendered in minimized form (e.g. <img ismap> rather than <img ismap="">)

So I think he expects this as output:

<option value="1" selected>Selected</option>

I think the generated html is fine, at least better than what it was according to the original issue comment, especially when it could be selected="False".
I wonder if the comment in z3c.pt is simply wrong.
Note: I have not tried any of this.

[stevepiercy]

I think the comment is a misinterpretation of the W3 specification.

        # List of Boolean attributes in HTML that should be rendered in
        # minimized form (e.g. <img ismap> rather than <img ismap="">)
        # From http://www.w3.org/TR/xhtml1/#guidelines (C.10)
        # TODO: The problem with this is that this is not valid XML and
        # can't be parsed back!

The link should be https://www.w3.org/TR/xhtml1/#C_10, which is quoted below.

Some HTML user agents are unable to interpret boolean attributes when these appear in their full (non-minimized) form, as required by XML 1.0. Note this problem doesn't affect user agents compliant with HTML 4. [emphasis added] The following attributes are involved: compact, nowrap, ismap, declare, noshade, checked, disabled, readonly, multiple, selected, noresize, defer.

In other words, it is perfectly OK, and actually better to use, the full form, such as selected="selected". The full form is required by XML 1.0 and is understood by any user agent that supports HTML 4, which is over a decade old.

[gh-PonyM]

Okay, now I understand and true, I was mislead by that comment. Thanks very much for the clarifications.
However, I found something else that is really part of the bug. I still get <option selected=False></option> if used inside a macro. I'll try to figure out why.

[stevepiercy]

@malthe @mauritsvanrees should I submit a PR to update that comment? I think its current state is a misinterpretation of the specification.

[malthe]

@stevepiercy yeah I think that would be helpful. Thanks!