Highlight differences between books?

[radius]

Is there any convenient way to show the difference between two editions of the same book?

For example, say I have an epub from MobileRead (with MR edits) and one from Project Gutenberg and I want to see what was cleaned up, or added and so on.

To do that by hand, I would have to:

1) Unzip
2) Concatenate the XHTML files in order according to manifest
3) Strip out all the tags
4) Normalize quotes, punctuation and other characters that might appear as HTML entities versus unicode characters and so on
5) Use a diff program to see the difference in the text

I am trying to clean up my library, but when all I want to do is decide which version of a duplicated book in my library I want to keep, this is a lot of work for a single title.

Or as another example, for some SF stories I have, I am unsure which file is from the magazine version and which one is the novel version and so on.

[jackie_w]

One thing you could try is to convert both epubs to TXT using calibre. Then use your Diff utility to compare the two TXTs. At least it would reduce the number of steps.

[GrannyGrump]

@radius --- take a look at this thread. WinMerge might be the answer.

https://www.mobileread.com/forums/sh...light=WINMERGE

All you should have to do is change the file extension from epub to .ZIP.
But for the kind of details you are examining, that might not be enough "pre-exam" preparation.

Keep us posted on what works for you.

[radius]

Originally Posted by jackie_w

One thing you could try is to convert both epubs to TXT using calibre. Then use your Diff utility to compare the two TXTs. At least it would reduce the number of steps.

Ah, I didn't know you could dump text from calibre. I might install and try that.

[radius]

Originally Posted by GrannyGrump

@radius --- take a look at this thread. WinMerge might be the answer.

Thanks for the suggestion. If I was comparing two edits of the same book this would work great. But I am actually trying to compare "editions" of the same work. That is, maybe something I had marked up from plain text versus a new Distributed Proofreaders version or a GrannyGrump edition

WinMerge would highlight all of the HTML markup and obscure the difference in the actual text of the book.

Edit: oh yeah, forgot to mention that my day to day machines are a Linux box and a Mac laptop so Windows is an inconvenience too haha

[Doitsu]

Originally Posted by radius

Edit: oh yeah, forgot to mention that my day to day machines are a Linux box and a Mac laptop so Windows is an inconvenience too haha

In that case, have a look at this Linux post by SBT.

If you prefer GUI tools, I'd recommend Meld and Beyond Compare.

[JSWolf]

The only way to do this is to convert each book to text and compare the two text files.

[radius]

Originally Posted by Doitsu

In that case, have a look at this Linux post by SBT.

If you prefer GUI tools, I'd recommend Meld and Beyond Compare.

Good ideas. I like Diffmerge myself. The last time I tried Meld it felt slow to me, and AFAIK Beyond Compare is commercial. I don't think they solve the "look inside a zip archive" and "ignore HTML" issues though.

[radius]

Originally Posted by JSWolf

The only way to do this is to convert each book to text and compare the two text files.

Yes. I've been thinking that I will script this myself. After thinking about it, I realized that I don't need a 100% solution, but rather an 80% solution would work fine. After all, this is only for my personal convenience.

[Doitsu]

Originally Posted by radius

I don't think they solve the "look inside a zip archive" and "ignore HTML" issues though.

They don't. BTW, you also might be able to use pandoc to convert epubs to plain text files before comparing them. (Use -t plain to force plain text output.)

[rkomar]

I can't remember if you (@radius) use linux or not, but if so, you can use unzip to extract the contents of the epub file, and then use lynx -dump -nomargins to extract the text from the htm/html files. You can then use diff to compare the extracted text.

[radius]

Ah pandoc looks like an excellent general purpose tool! Thanks! I guess trying to normalize plain text is easier than html so that might be the way to go.

@rkomar, I'm actually an elinks user and I've tried the dump command before. My problem with it was that it tried to preserve too much formatting from the html. I don't think elinks supports a nomargin option unfortunately. That's a good tip.

[rkomar]

Originally Posted by radius

@rkomar, I'm actually an elinks user and I've tried the dump command before. My problem with it was that it tried to preserve too much formatting from the html. I don't think elinks supports a nomargin option unfortunately. That's a good tip.

Have you tried using the "-b" option with diff? It suppresses differences in white space, so maybe it will be less sensitive to formatting between the text outputs.