XHTML closing tag issue

[ptm@xact.co.uk]

Simply by parsing and committing an xhtml file, using the container, I see the following changes to the content.

Before parsing:
<span x="y"/>
<div id="w"/>

After committing:
<span x="y"></span>
<div id="w"></div>

Is there a 'tweak' setting or other method I can use to preserve the compact version of for tags with only attributes and no content?

[kovidgoyal]

No there isn't, IIRC.

[ptm@xact.co.uk]

I have not looked at the Calibre code. Hopefully there is an option on the serializer to use compact closing tags (i.e. <sometag/>) instead of <sometag></sometag>. If this is the case, how difficult do you think it would be to add a tweak to set this option? Is this something you or I can look at? I would need a few hints where to start looking.

[kovidgoyal]

This is done deliberately, see line 441 in oeb/base.py

[JSWolf]

Originally Posted by ptm@xact.co.uk

I have not looked at the Calibre code. Hopefully there is an option on the serializer to use compact closing tags (i.e. <sometag/>) instead of <sometag></sometag>. If this is the case, how difficult do you think it would be to add a tweak to set this option? Is this something you or I can look at? I would need a few hints where to start looking.

Most tags are <sometag>Text from the book</sometag>. That's why it should not be done. <sometag/>This is text from the book will not work.

[ptm@xact.co.uk]

Originally Posted by kovidgoyal

This is done deliberately, see line 441 in oeb/base.py

I see the comment about some browser based renderers. I wonder if this is still true. I've no idea when this code was written. But my boss needs self closing tags. So if, at line 160 we had:

def close_self_closing_tags(raw):
    if self_closing_tags_permitted():
        return raw
    for_unicode = isinstance(raw, str)
    repl = as_string_type(r'<\g<tag>\g<arg>></\g<tag>>', for_unicode)
    pat = self_closing_pat(for_unicode)
    return pat.sub(repl, raw)

And in 'tweaks', wherever they are coded, the self_closing_tags_permitted() function is added. Returning false by default, but true if so configured by the user. Then we get the best of both worlds. Am I on the right track?

[kovidgoyal]

I'm not going to change that as it is a completely cosmetic change that has the potential to break rendering in the real world. You are welcome to run calibre from source and make the change for yourself.

[ptm@xact.co.uk]

Originally Posted by kovidgoyal

I'm not going to change that as it is a completely cosmetic change that has the potential to break rendering in the real world. You are welcome to run calibre from source and make the change for yourself.

May I ask you to reconsider? The Client using our plugin says it is a requirement. They suggest <tag></tag> (with no text between) is badly formed XML. The suggestion in my previous post would maintain backwards compatibility. So I do not believe it would affect anyone, unless they use the 'tweak' setting.

Implementing such a change myself (apart from not having a Linux system to do the build) would mean our Client could no longer install any future Calibre updates.

Fingers crossed, best regards
Paul

[kovidgoyal]

<tag></tag> is perfectly well formed XML. I suggest you tell your client to read the XML specifications.

You don't need Linux to do a build, indeed you don't need to do a build at all. calibre can run from source without needing building see https://manual.calibre-ebook.com/develop.html

I am not going to accept this change in calibre as it has the potential to break things for people for no actual benefit to any calibre user.

[DNSB]

Is that first line in your sample real? The x in x="y" is not a valid attribute.

Before parsing:
<span x="y"/>
<div id="w"/>

After committing:
<span x="y"></span>
<div id="w"></div>

[Quoth]

Originally Posted by ptm@xact.co.uk

Simply by parsing and committing an xhtml file, using the container, I see the following changes to the content.

Before parsing:
<span x="y"/>
<div id="w"/>

After committing:
<span x="y"></span>
<div id="w"></div>

Is there a 'tweak' setting or other method I can use to preserve the compact version of for tags with only attributes and no content?

It's only historic accident that regular content tags can be self closing.

First line before opening <html>
<?xml version='1.0' encoding='utf-8'?>
Only inside <head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>
<meta otherstuff />
<link rel="stylesheet" type="text/css" href="stylesheet.css"/>
</head
Inside the <body >
Valid tags for ebooks that are self closing:

<hr />
<br />

<img src="whatever" alt="sometext"> and height & width should be via a class.
maybe <col /> to set properties of a table column, but tables in ebooks are tricky
</body>

An ebook isn't a web page, nor an XML application for a browser.

Such a setting in general is crazy.
These are not converted self closing tags, but tags with no content.

<span x="y"></span>
	<div id="w"></div>

This is obsolete for empty tags:

<span x="y"/>
	<div id="w"/>

As ebooks are "zipped", it saves nothing. Compact (as with clever C that should take 3 lines fitted as one statement) is not always a good idea.

The HTML WG has discussed this issue: the intention was to allow old (HTML-only) browsers to accept XHTML 1.0 documents by following the guidelines,

So, yes, every browser will accept <div />, <span /> etc if the doc type allows it, but it's actually now wrong and totally inappropriate for ebooks.
mobi is HTML3
azw3 & epub2 is HTML5
epub3 is extended.


Here is a list of self closing tags for web pages. The ones used in ebooks listed above and in bold. <col /> should be avoided in ebooks.

Name	Description
< !DOCTYPE html>	Special Case, never use a forward slash. Used to tell the browser what type of document you are using.
 <br />	Used to create a breaking space on a page. Content following this tag will be moved to the next line in the browser.
<area />	Used for image mapping, making specified areas of an image clickable by associating a hyperlink
<base />	Creates a base url for all other hyperlinks on a given webpage so that relative urls can be used. The base element is placed inside the head.
<col />	This element represents a column or several columns in a column group
 <command />	 A multipurpose element for representing commands.
 <hr />	 Horizontal rule – creates a horizontal line across the page.
 <embed />	 Used to embed multimedia objects such as a video into a page.
 <input />	 This element allows a visitor to enter information such as username, password, email address etc…
 <link />	 Used to link external pages to a webpage such as external style sheets.
 <wbr />	 Word Break Opportunity. Specifies where in a line of text it would be OK to break into the next line on different screen sizes.
 <track />	 Used in conjunction with audio or video to add subtitle or caption tracks.
 <meta />	 A multipurpose element used to represent metadata. Metadata is information about data.
 <param />	 Defines parameters for object elements such as audio or video
 <source />	 Allows multiple media sources to be specified for audio and video elements.
 <keygen />	 Control used to generate a key pair (public – private) and allows the public key to be submitted to the user.
 <img />	 Image – Used to add images  to a web page.

This isn't definitive.

However while <span /> <div /> etc can be legal in all browsers in web pages, I've never seen such in ebooks in 12 years and never used them in 30 years of web site editing.

[Quoth]

Originally Posted by ptm@xact.co.uk

They suggest <tag></tag> (with no text between) is badly formed XML. The suggestion in my previous post would maintain backwards compatibility. So I do not believe it would affect anyone, unless they use the 'tweak' setting.

No, it's not badly formed XML. Depending on the properties of <tag> it might be pointless and do nothing, but that's not the same thing as malformed.

It might confuse people and break ebook conversions.

[Quoth]

Self closing tags are also alternatively known as void tags, empty tags, singletons tags, etc. i.e these tags do not have contents and also can not have any child.
Container tags can be empty <tag></tag>

The Void / Empty / Singleton / Inherently self closing tags may end in />, but some do not.
They can't be written as an opening and closing tag. Illegal (malformed) XML/HTML
<hr></hr>
<br></br>
<col></col>

So container tags can be <tag></tag>, i.e. have no content.
But singleton tags are never a pair of opening and closing tags, they are self closing, complete, though some end in /> and some contexts will accept <br> and <hr>.

The comment is a special case of void or singleton
<!-- This is a comment -->

But this also is a valid comment tag. All the enclosed HTML is disabled.

<!-- 
<ul class="top_menu_right">
			<li>
				<form action="search.php" method="post">
				<input type="hidden" name="do" value="process" />
				<input type="hidden" name="showposts" value="0" />
				<input type="hidden" name="childforums" value="1" />
				<input type="hidden" name="securitytoken" value="1723152374-234a53b9ef86581494348ed4767ee7d9f993eecf" />
				<input type="hidden" name="s" value="" />
				<input type="text" class="button" name="query" size="20" style="width:120px" />
				<input name="search" value="Search" type="submit" class="button" style="width: 5em;" />
				</form>
			</li>
			</ul>
-->

An ebook can have comments, but unlike web pages, shouldn't need them.

[Quoth]

A further search

They're called "void" elements in HTML 5. They're listed in the official W3 spec.

A void element is an element whose content model never allows it to have contents under any circumstances.

As of April 2013, they are:

area, base, br, col, command, embed, hr, img, input, keygen, link, meta, param, source, track, wbr

As of December 2018 (HTML 5.2), they are:

area, base, br, col, embed, hr, img, input, link, meta, param, source, track, wbr

Void type elements are never <tag></tag>
Container elements with no content may work as <tag />, but it's as likely to break in a web browser, hence ALWAYS use <tag></tag> even if no content, such as <script type="text/javascript" src="external.js">< /script> because it will break without a separate closing tag.
NEVER <script type="text/javascript" src="external.js" />
Most ereaders don't support script at all.

As noted earlier, ebooks only support a subset of void elements.

[Quoth]

And finally!

C.3. Element Minimization and Empty Element Content
Given an empty instance of an element whose content model is not EMPTY (for example, an empty title or paragraph) do not use the minimized form (e.g. use <p></p> and not <p />).

from https://www.w3.org/TR/xhtml1/#dtds

Using <title /> instead of <title></title> actually causes a blank page with all the content <body>stuff</body> missing and only visible via view code on some browsers.

But remember, ebook CSS and HTML (and XML) is a special subset, even if rendered by a web page renderer, which might render stuff that breaks totally on real ereaders.