How can I use a script to change some text in an epub file?

Question

I've recently bought a Nook Simple Touch. I use Calibre to manage my ebooks, and to transfer them to the Nook.

Due to a non-standard implementation of the epub specification on B&N's part, the Nook ST does not display cover images when they are brought over from many sources. The issue is described here: http://john.nachtimwald.com/2011/08/21/nook-covers-not-showing-up/ Basically the Nook ST requires the XML attribute for the cover to be in the format:

<meta name="cover" content="id5" />

But many epub creators have them around this way:

<meta content="id5" name="cover" />

And the Nook ST then ignores the cover image entirely.

I have been manually editing the content.opf file in my epub files. So far they have all had the image meta, but it was always around the "wrong" way (wrong, according to the Nook).

Recently I've been playing around with REGEX, mostly to try to automate the cleaning up of epubs converted by Calibre from PDF files. I'm still very much a beginner with REGEX.

What I was wondering is how I might go about automating the swapping of the 'name' and 'content' attributes? I figure it can be done with a combination of REGEX and scripting. I know some of the other epub related scripts I have are in Python. I am on a Mac (OS X) and they seem to run fine. AppleScript might be a good option too, although I'd like something that people can run on any platform, as I am sure other folk will find this useful.

Here are the steps I foresee:

~ Extract epub file

~ Use REGEX to look for:

<meta content="???" name="cover">

~ If found, use REGEX to change it around to:

<meta name="cover" content="???">

~ Zip extracted files back into an epub using the correct zipping process.

I found info here: http://www.mobileread.com/forums/showthread.php?t=55681 explaining how to zip up an epub file correctly. Basically it requires these two commands:

zip -X0 "full path to new epub file" mimetype
zip -rDX9 "full path to new epub file" * -x "*.DS_Store" -x mimetype

I'd like to post the resulting script online where ever it might be found and made use of (until B&N resolve their poor epub/XML implementation). Posting it on the Calibre forums and the mobileread forums comes to mind (since they are two I am familiar with, and have seen people discussion manual fixes to this issue).

Is there someone who can walk me through how to create such a script? Ideally, I'd love to actually know how to create the script, so that over time I can start to figure out these sorts of things myself (especially the REGEX part, as I see more and more how useful it is).

Thank you.

Jonathan

@Haldean: ADDED to illustrate what I mean in a comment to Haldean regarding making his script work through all content.opf files in all subfolders recursively.

> My_expanded_epubs
- -> epub_one_expanded
- - - -> content.opf
- -> epub_two_expanded
- - - -> content.opf
- -> epub_three_expanded
- - - -> content.opf
etc.

Answer 1

If you're willing to go with a shell script (which I think is a better option) then you can use a sed one-liner:

sed 's/<meta content="\(.*\)" name="cover" \/>/<meta name="cover" content="\1" \/>/' [your-file]

That should replace all meta lines where the content attribute comes first with one with the correct order. An equivalent Python translation of that would be:

import re
import sys
with open(sys.argv[1]) as f:
  for line in f:
    # Match this line to the wrong-way-around meta tag, put the content in group 1
    m = re.match(r'<meta content="(.*)" name="cover" />', line)
    if not m:
      print line
    else:
      print '<meta name="cover" content="%s" />' % m.group(1)

Answer 2

I would suggest that you use sed to work with the unpacked file, and do something like:

sed -e 's/<[ ]*meta[ ]*content[ ]*=[ ]*"\(.*\)"[ ]*name[ ]*=[ ]*"cover"[ ]*\/*[ ]*>/<meta name="cover" content="\1" \/>/g'

Note that this version will cope with extra or missing space, or slashes.

You may like to subsequently use an xml processor (I would suggest a python script using lxml) to verify that your edit has not for any created invalid markup.

Using any kind of XML tool to perform the manipulation is unattractive in the extreme, because a fully compliant XML processor may make other changes which are completely legal, and also trigger other bugs in your nook. Using sed allows you to edit only the parts of the document you want to.

Answer 3

Personally I wouldn't do this with regex (its the wrong tool). Could you use XSLT?

---

EDIT:

Here is a demo. http://www.xsltcake.com/slices/nvLRJ6

There are a number of XSLT librarys for python.

---

EDIT:

If you insist on doing it with regex you'll want a patten like this:
<meta content="([^"]+)" name="([^"]+)" \/>

I say this with the disclaimer that this is the wrong tool and there are edge cases that make this unreliable and I don't recommend it.

http://regexr.com?301uq

Answer 4

I agree with zapthedingbat's answer: this is an XML problem, so let's use tools specifically designed for XML, namely XSLT.

Since you're new to XSLT, you'll need an XSLT processor to try this solution. If you are using *nix, xsltproc is a command-line processor and is almost surely installed by default and you can take this solution at face value. If not, you'll need to see if your language of choice has an API for perform XSL transforms.

Here's a very simple general solution for reordering the attributes:

<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">

  <xsl:template match="@*|node()">
  <!-- copy everything as is -->
    <xsl:copy>
      <xsl:apply-templates select="@*|node()"/>
    </xsl:copy>
  </xsl:template>

  <xsl:template match="meta">
    <!-- except for the <meta/> element, reverse the attribute order -->
    <meta name="{@name}" content="{@content}"/>
  </xsl:template>
</xsl:stylesheet>

Here's your example:

<root>
  <meta content="id5" name="cover" />
</root>

Running the XSLT with xsltproc:

$ xsltproc so.xsl so.xml

and the result:

<root>
  <meta name="cover" content="id5"/>
</root>