LilbreOffice 6.0 epub2/epub3 output test

[Doitsu]

Since the upcoming LibreOffice 6.0 version will have epub2/epub3 output support, I tested a LibreOffice 6.0 pre-release version with a demo file originally created by Kovid Goyal to test the Calibre DOCX filter.

I wasn't exactly impressed by the output. The epub2 version had duplicate header title tags in each file and both versions had one broken link.

The converter also couldn't handle footnotes and endnotes and, most annoyingly, all generated HTML files contained inline styles.

[Bertrand]

Wow, what a soup ...
Good luck if you need to edit the code outside Libreoffice.

[Sarmat89]

Well, EPUB is not a semantic markup format, and it cannot be created with a text editor, so this is expected.

Plus, the LibreOffice has a one-trick pony XHTML exporter, which doesn't even try to simplify the output.

[DaleDe]

Better to use Atlantis word processor as it creates very nice compact ePub files.

[JSWolf]

Originally Posted by Sarmat89

Well, EPUB is not a semantic markup format, and it cannot be created with a text editor

Actually, ePub can be created with a text editor. It's not all that easy, but it can be done.

[arjaybe]

Originally Posted by DaleDe

Better to use Atlantis word processor as it creates very nice compact ePub files.

It has poor ODT support and is only for Windows.

[JSWolf]

Originally Posted by arjaybe

It has poor ODT support and is only for Windows.

And those points are an issue why?

[mobama]

@Doitsu
How does it compare with the epub plugin in the earlier version of Libreoffice?

[Toxaris]

Originally Posted by Sarmat89

Well, EPUB is not a semantic markup format, and it cannot be created with a text editor, so this is expected.

Yes it can. If I can write an add-in that does it (for Word, granted), than they can do it as well. The other example given in this thread, Atlantis, is a fine Word processor and it can do it as well with reasonable results.

[Doitsu]

Originally Posted by mobama

@Doitsu
How does it compare with the epub plugin in the earlier version of Libreoffice?

I no longer have the old version installed, but since its converter was based on writer2html, I converted the same test file to epub2 and epub3 files using the standalone version of writer2html.

IMHO, the results are much better. It also created lots of inline styles*, but at least footnotes and lists survived the conversion.
If I had to pick one, I'd definitely pick the old converter.

* Inline styles can be easily converted to classes with KevinH's RemoveInLineStyles Sigil plugin.

[Tex2002ans]

Originally Posted by Doitsu

I wasn't exactly impressed by the output. The epub2 version had duplicate header title tags in each file and both versions had one broken link.

This EPUB support is still in the barebones stages. I found out about it on the LibreOffice Wiki for v6.0:

https://wiki.documentfoundation.org/....0#New_filters

This also linked to the author's blog post about adding EPUB3 support.

This is a good first step.

Originally Posted by JSWolf

And those points are an issue why?

Because many people use ODT documents and are on non-Windows OSes?

As the LibreOffice EPUB Import/Export grows, there will be less of a need for a lot of the third-party solutions now (Save as DOCX and run through Calibre, etc., etc.).

And maybe some of the libraries used in LibreOffice will tangentially lead to better ODT input on Calibre's side of things.

[Notjohn]

Originally Posted by DaleDe

Better to use Atlantis word processor as it creates very nice compact ePub files.

Thanks for that information! I collect votes on behalf of purpose-built softwares like Atlantis, Jutoh, Scirvener, Vellum etc, since I don't use any of them myself.

[roger64]

Originally Posted by Doitsu

I no longer have the old version installed, but since its converter was based on writer2html, I converted the same test file to epub2 and epub3 files using the standalone version of writer2html.

IMHO, the results are much better. It also created lots of inline styles*, but at least footnotes and lists survived the conversion.
If I had to pick one, I'd definitely pick the old converter.

* Inline styles can be easily converted to classes with KevinH's RemoveInLineStyles Sigil plugin.

writer2xhtml can create directly quite clean Epub2 or 3 without any inline style if you select carefully its -little convoluted- configuration options. The source odt is expected to be written using styles that writer2xhtml obviously cannot invent on behalf of the user.

I opened the source docx file with LO6 and converted it to odt format. Then I used odtImport (a Sigil plugin) to export directly an Epub3. In the joint zip file you'll find the odt and the Epub3.

I only batch renamed the html files to xhtml. Everything else is straight from the converter. Epubcheck reports a missing image . writer2xhtml displays all four images because there are two green dots which are the same image, which it displays with different relative settings.

This version of writer2xhtml is an alpha version 1.6. compiled by Doitsu from the source repository.

Edit: for reasons I fail to explain, the odt file has a 4.3 MB size but the resulting Epub has a 32 k size...

[Sarmat89]

Your ODF file contains 18 embedded fonts.

[Apparition B5]

According to the update in this blog post, EPUB support has been substantially approved in the LibreOffice master trunk.