Sigil changes epub, even with "Clean Source" unchecked; ADE crashes

[DonPayne]

(I'm new to Sigil today.)

Sigil makes spurious changes to my epub file. These changes cause ADE to crash and make comparison of epub files practically impossible using "diff" or similar text tools. Process:

I open Sigil version 0.9.2 on Windows 8.1.
I go to menu > Edit > Preferences; under "Clean Source" > "Mend XHTML Source Code On", I uncheck both boxes "Open" and "Save"; click OK to confirm.
I open an epub file (generated by Adobe InDesign CC (Creative Cloud) version 2014.2 on Mac OS X), do nothing to it, and save it.
I try to open this epub file in ADE (Adobe Digital Edition) version 4.5 on Windows 8.1; it "stops responding", then crashes. The original file opens in ADE with a delay (it's 30+ MB) but no errors.
I unzip both original and Sigil-saved epub files, then diff them.

Some differences:

- under directory OEBPS, original directory "font" is renamed by Sigil to "Fonts", "image" to "Images" and "css" to "Styles". All XHTML files, which in original are directly in the OEBPS directory, Sigil has moved to a subdirectory "Text". The XHTML files are adjusted to point to fonts, images and CSS files in the new directories. As you can imagine, this introduces many, many differences merely in the directory references in "href" and "src" attributes in the XHTML files.

- In META-INF/container.xml and OEBPS/content.opf, and OEBPS/toc.ncx, "standalone" attribute was removed:
< <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
---
> <?xml version="1.0" encoding="UTF-8"?>

- In the XHTML, the original doc has a space before the close of an "empty" XHTML element. Sigil's version removes that space. Example from container.xml:
Original:
<link href="css/idGeneratedStyles.css" rel="stylesheet" type="text/css" />
Sigil:
<link href="../Styles/idGeneratedStyles.css" rel="stylesheet" type="text/css"/>

- In encryption.xml, fonts appear in a different order from the original. Example:
original:
<encryption xmlns="urnasis:names:tcpendocument:xmlns:conta iner" xmlns:enc="http://www.w3.org/2001/04/xmlenc#">
<enc:EncryptedData>
<enc:EncryptionMethod Algorithm="http://www.idpf.org/2008/embedding" />
<enc:CipherData>
<enc:CipherReference URI="OEBPS/font/Calibri.ttf" />
...
Sigil:
<?xml version="1.0" encoding="UTF-8"?>
<encryption xmlns="urnasis:names:tcpendocument:xmlns:conta iner" xmlns:enc="http://www.w3.org/2001/04/xmlenc#">
<enc:EncryptedData>
<enc:EncryptionMethod Algorithm="http://www.idpf.org/2008/embedding"/>
<enc:CipherData>
<enc:CipherReference URI="OEBPS/Fonts/SFCartoonistHand-BoldItalic.ttf"/>
...

- in OEBPS/content.opf, order of attributes was changed:
< <package version="3.0" xmlns="http://www.idpf.org/2007/opf" unique-identifier="bookid" prefix="ibooks: http://vocabulary.itunes.apple.com/rdf/ibooks/vocabulary-extensions-1.0/">
---
> <package prefix="ibooks: http://vocabulary.itunes.apple.com/rdf/ibooks/vocabulary-extensions-1.0/" unique-identifier="bookid" version="3.0" xmlns="http://www.idpf.org/2007/opf">

< <meta name="generator" content="Adobe InDesign 10.2" />
---
> <meta content="Adobe InDesign 10.2" name="generator"/>

< <item id="cover" href="cover.xhtml" media-type="application/xhtml+xml" />
[... long list of items ...]
---
> <item href="Text/cover.xhtml" id="cover" media-type="application/xhtml+xml"/>

- in OEBPS/toc.ncx, nested XHTML that was packed into one line in original was expanded to multiple lines by Sigil. Example:
< <navPoint id="navpoint1" playOrder="1"><navLabel><text>DEDICATION</text></navLabel><content src="Pickle_Don_sandbox.
xhtml#_idParaDest-1" /></navPoint>
[... many other navPoint elements ...]
---
> <navPoint id="navpoint1" playOrder="1">
> <navLabel>
> <text>DEDICATION</text>
> </navLabel>
> <content src="Text/Pickle_Don_sandbox.xhtml#_idParaDest-1"/>
> </navPoint>

- Even though most of the XHTML files are in the same directory "Text", Sigil links from one to another by an unnecessary path up one directory then back down into Text:
original:
< <p class="TOC_Chapter"><a href="Pickle_Don_sandbox.xhtml#_idTextAnchor000">D EDICATION</a></p>
Sigil:
> <p class="TOC_Chapter"><a href="../Text/Pickle_Don_sandbox.xhtml#_idTextAnchor000">DEDICAT ION</a></p>

- Sigil changes empty "td" elements into an open and close tag:
< <td class="No-Table-Style" />
---
> <td class="No-Table-Style"></td>

All these changes appear innocuous, but something must be making ADE crash. And with all the spurious changes, I cannot see what the real problem is.

[JSWolf]

The problem is really quite simple. You are trying to edit an ePub 3 eBook and Sigil does not work with ePub 3. It only works with ePub 2.

If you are the one who made the eBook, make it as an ePub 2 eBook. If not, don't edit it in Sigil.

[KevinH]

FYI, Sigil has always done this. It is how it operates. When you load an epub into Sigil, Sigil reformats it to fit its internal layout. Parses all files and reserializes them to make sure their src and href links are properly updated, etc. None of the changes you mention should break anything.

So some questions:

In BV, does your epub function after loading?

Were all the files well-formed when loading?
If not, enable Mend on Open and reopen the epub.

Was this a epub2 or an epub3?

Was this a fixed layout?

If an epub2, please try running the Flightcrew plugin to see if it detects any breakage.

Or you can run epubcheck (via plugin) and again see if any breakage is detected, before writing the epub out.

That should point to anything funny.

Let us know what it reports and we can try to help.

KevinH

[KevinH]

If JSWolf is right and this is an epub3, then Sigil is not yet ready for it unless you are doing only very vanilla editing, using none of the built in tools, etc. This will change greatly in Sigil 0.9.3, the next release (current Sigil master right now) which has much better epub3 support.

The entire Sigil 0.9.X series is meant to gradually add support for epub3 in stages while keeping full epub2 support. We are still early in that process, sorry about that.

Please try again when Sigil 0.9.3 is released.

KevinH

[DiapDealer]

Double mimetype.

[JSWolf]

Originally Posted by DonPayne

> <package prefix="ibooks: http://vocabulary.itunes.apple.com/rdf/ibooks/vocabulary-extensions-1.0/" unique-identifier="bookid" version="3.0" xmlns="http://www.idpf.org/2007/opf">

Where it shows version="3.0" shows that this is an ePub 3 eBook and thus, not compatible with Sigil 0.9.2 or earlier.

[DonPayne]

Thanks, all, for educating me. I created the ePub (actually, InDesign did), but I want to keep ePub3 format for adding video, etc. I'll use Calibre to edit it.

[odamizu]

Originally Posted by JSWolf

The problem is really quite simple. You are trying to edit an ePub 3 eBook and Sigil does not work with ePub 3. It only works with ePub 2.

Is there a plug-in or command to turn an ePub3 into ePub2?

(Sorry in advance for the newbie question )

[Toxaris]

Originally Posted by odamizu

Is there a plug-in or command to turn an ePub3 into ePub2?

(Sorry in advance for the newbie question )

No, you have to do that manually. Since certain things in ePUB3 are not supported in ePUB2, conversion can be troublesome.

[odamizu]

Oh well. Wishful thinking. Thanks for the quick reply!