Re: [docbook-apps] Thanks re docbook to ePub3

[Bob Stayton <bobs@sagehill.net>]

I'm coming to the conclusion that the EPUB3 standard changed after the Docbook epub3 stylesheet was created. The EPUB3 standard is based on XHTML5, and XHTML5 does permit @width on <img> elements. Further evidence of the change is that in the DocBook template that converts banned attributes to style attributes, there is a specific exception carved out for @width on img elements to pass it through to the epub3 output. The comment in the XSL says it is permitted, and I know I've produced epub3 output that passed previous versions of epubcheck without complaining about @width.

So this needs to be updated in the DocBook stylesheet. Could you please file an issue on the Github site for the stylesheets:

https://github.com/docbook/xslt10-stylesheets

In the meantime, I'll describe the source of the problem in case you want to fix it in your stylesheets. In the xhtml5/html5-element-mods.xsl files there is a template that starts with <xsl:template match="*" mode="convert.to.style"> that handles the conversion of hard attributes to styles. The culprit is this line in that template:

 <xsl:when test="local-name() = 'width' and $element != 'img'">

that needs to be changed to:

 <xsl:when test="local-name() = 'width'">

to remove the exception for img elements.

Bob Stayton
bobs@sagehill.net

On 3/25/2021 9:21 PM, Bob Stayton wrote:

I presume you are using the epub3 stylesheet, which is better supports the current EPUB standard. The DocBook epub3 stylesheet imports the xhtml5 stylesheet as its starting point. In the xhtml5 stylesheet, banned attributes like @width in an <img> element are supposed to be converted to a single @style attribute by processing the generated img element in mode="convert.to.style".Â

So I'm not yet clear on is where that process is going wrong such that you are seeing @width in your output. I'll need to investigate more.

Bob Stayton
bobs@sagehill.net

On 3/25/2021 4:46 AM, Radu Coravu wrote:

Hi Philo,

So the problem which remains is that the XHTML docs from the generated EPUB archive have the image reference looking like this:

<img src="" width="50%"/>

right? Does this way of specifying the proportional width work with the EPUB reader? So is it only a problem of validation errors reported by the epub checker but actually the EPUB reader interprets the proportional width correctly?

Probably the image width would need to be specified like this in the XHTML output:

<img src="" style="width: 50%; height: 50%"/>

The DocBook publishing is based on XSLT stylesheets, there is an XSLT stylesheet named "graphics.xsl" located in the Oxygen installation here:

OXYGEN_INSTALL_DIR/frameworks/docbook/xsl/xhtml/graphics.xsl

with a template named:

<xsl:template name="process.image.attributes">

which could probably be overwritten with a Docbook customization to produce a @style attribute instead of directly the @width attribute.

Maybe you can start a different thread for the double numbers in the PDF output and attach a small DocBook sample to reproduce the problem.

Regards,

Radu

Radu Coravu
Oxygen XML Editor

On 3/25/21 04:45, Philo Calhoun wrote:

Thank you Robert and Richard for your suggestions:

The color issue was solved using your Âsuggestion:

<para><phrase role="red">Step 1:</phrase>ÂDescription of Step 1 etc. </para>,Â

and editing the docbook-epub.css.xml file in Oxygen XML to include:

span.red {
 color: rgb(200,0,0);
 font-weight: bold;
}

However, the issue of image scaling seems more problematic, as I have different % scaling of different images within a single topic (from 25% to 100%). The only way I can see to make this work is to have the conversion add some sub style of img, like img.half img.quarter etc. and Iâm not sure how to do it in docbook. I can edit the ePub in Sigil and add that sub style of img, but I then have to add the designation for each and every image file, and that is a lot of work.

I am also having issues with pdf output of cross-references. They work fine in ePub but generate double numbers in pdf. Iâm fairly new to Oxygen XML and docbook, and understand I need to brush up on the conversion algorithms.

I am not using the docbook bibliography designation, as it seems pretty complicated for the mixed types of references (web links, books journals, etc.). So I have the bibiography as a section and designate the entries like:

<listitem>
  <para xml:id=âSmith01â>Smith, J. This is my title. Scientific Journal 2021; 37(4):435-438. </para>
ÂÂ </listitem>

AndÂcross reference like:

[<link linkend=âSmith01"/>]

As a background, I have a couple of books that need to be in both pdf and ePub outputs and have a lot ofÂcross-references, tables, figures, and mathematical formulas. Iâve used InDesign and Framemaker for them in the past, but migrated to MadCap Flare. Unfortunately, while Flare creates beautiful web output, its pdf and ePub outputs are buggy and limited. This is particularly so with mathematical formulas and fine tuning of the table of contents. I get about 95% of the way there, but have to do a lot of creative workarounds due to problems/bugs with Flare. Due to this frustration, I now publish the print version using LaTeX, which works perfectly, but getting precise ePub output without a lot of carryover junk code (the way Flare does) has been much easier in Oxygen than in Flare. At this point, I still am having to do a fair amount of precise formatting of the Oxygen generated ePub in Sigil. (I tried Jutoh, but it corrupts MathML code unfortunately.) My goal is to get both a clean output and specific image formatting for ePub directly from OxygenXML. The extra level of structure using XML vs XHTML is a huge benefit when it comes to crossreferencing and maintaining consistent formatting.

Thanks again,

Philo