Snag if class="chapter"

[Quoth]

Occasionally I convert an azw3 or epub and on viewing each paragraph is a new page.

It's caused by the person/program doing original formatting choosing to name the regular body style paragraph
<p class="chapter">some stuff</p>

Just use Calibre editor to rename the class to something like "p-indent"

Is there a way in any of the default conversion settings to detect this and automatically rename the class?

The default Structure detection is normally useful
Detect chapters at (X-Path expression)

//*[((name()='h1' or name()='h2') and re:test(., '\s*((chapter|book|section|part)\s+)|((prolog|prologue|epilogue)(\s+|$))', 'i')) or @class = 'chapter']

That splits the file as the only 100% reliable way to have a new page (in all formats) is a new file.

[JSWolf]

What I do sometimes is remove the x-path, do the conversion, and put the x-path back. That works without having to change any of the code.

[kovidgoyal]

Well one could come up with some kind of heuristic for it I suppose. Maybe if there are consecutive runs of paragraphs in a single HTML file with that class or something. I generally like to avoid heuristics though as they can occasionally have false positives and lead to strange behavior.

[Quoth]

Originally Posted by JSWolf

What I do sometimes is remove the x-path, do the conversion, and put the x-path back. That works without having to change any of the code.

Yes, I've done that too. Though when a conversion produces madly unexpected results I look in the editor and then it's trivial to change.

[Quoth]

Originally Posted by kovidgoyal

Well one could come up with some kind of heuristic for it I suppose. Maybe if there are consecutive runs of paragraphs in a single HTML file with that class or something. I generally like to avoid heuristics though as they can occasionally have false positives and lead to strange behavior.

I agree, a heuristic might be a bad idea. It doesn't happen often.
I just wondered was there a way in existing "Look & feel" to automatically replace class="chapter" with class="p-body" that occurs in ANY <div> or <p> tag, and rename the CSS too.
I've seen a couple of mad ebooks with no <p> tags at all. They used <div> instead.

Obviously it would be very broken if the CSS wasn't changed.

Also class="p-body" might exist, so that's a bad choice. One would need a reliable new unique class name.


However I've also seen the chapters or other headings use <p> and you'd want to detect Chapter there. Calibre actually uses <p> instead of <h> on docx conversion unless the default names heading1 etc used. But it's common to have different styles of heading at same level so the default generic style names can't be used. I use the Span & Div edit tool to replace <p> with <h2> where there is the class that is obviously a Heading style. Then for some styles I have to add normal to the CSS as H styles are automatically bold and some fonts used for headings are already heavy/bold and then look garbage if bold applied

[JSWolf]

Originally Posted by Quoth

Yes, I've done that too. Though when a conversion produces madly unexpected results I look in the editor and then it's trivial to change.

If the conversion didn't go correctly, then yes, digging in the source code is the best way to handle it.