[GUI Plugin] Modify ePub

[botmtl]

Originally Posted by Rev. Bob

Who on Earth is building ebooks that way, instead of using image files like a sane person? There are times when text-encoded images make sense on the web, but in a static ebook file? That's just plain inefficient.

For the example I created, I doesn't make much sense. However, there is lots of fancy stuff you can do efficiently with data-uris. The book I bought used data-uris to make fancy/themed <hr>s that didn't require much data at all.

Besides, I think there a a lot of users who do download web pages (that potentially have images encoded this way) and make them into ebooks (fanficfare, news recipes, read-it-later type websites, etc.). I am one of them.

[Rev. Bob]

Originally Posted by botmtl

For the example I created, I doesn't make much sense. However, there is lots of fancy stuff you can do efficiently with data-uris. The book I bought used data-uris to make fancy/themed <hr>s that didn't require much data at all.

It's still more efficient to actually store the image, unless we're talking about something so tiny that the overhead of creating a file at all is a significant factor. (In which case, I'll still point to the cover image as the biggest part of almost any ebook and stick to my guns.)

Is the book in question self-published or traditionally published?

Originally Posted by botmtl

Besides, I think there a a lot of users who do download web pages (that potentially have images encoded this way) and make them into ebooks (fanficfare, news recipes, read-it-later type websites, etc.). I am one of them.

I think that's a much smaller population than you think it is. It may seem larger from the inside, but I doubt it's a drop in the bucket as a proportion of Ebook-Reading People.

[davidfor]

Originally Posted by botmtl

For the example I created, I doesn't make much sense. However, there is lots of fancy stuff you can do efficiently with data-uris. The book I bought used data-uris to make fancy/themed <hr>s that didn't require much data at all.

Unless they were all different, then using an image for the divider probably takes less space. Especially if there is only a few used a lot of times. But, I am curious to see it. Can you point it out?

Besides, I think there a a lot of users who do download web pages (that potentially have images encoded this way) and make them into ebooks (fanficfare, news recipes, read-it-later type websites, etc.). I am one of them.

Really? I don't think I've seen a page that did this for a long time. And I've never downloaded one for later reading. As you mention FanFicFare, can you point to a site doing it?

[Rev. Bob]

Originally Posted by botmtl

I have noticed that "data-uris" are detected as broken links and removed by this plug-in.

Would it be possible to fix this behavior?

BTW, it's really helpful if you actually tell me what you're doing instead of being this vague. This plugin has a lot of features and switches and multiple modes of use; hows about at least a hint of how I can replicate the behavior?

I mean, after I got the EPUB to the point where it would validate, I ran my standard cleanup pass on the file and got no errors. That means that either the validation fixes resolved the issue (in which case, not a bug) or you're doing something I normally don't.

This doesn't mean I'll make the change you're after - only that if I elected to investigate, I'd need more information than you've provided. In addition, that also means the behavior's in code I had nothing to do with, which would make changes more problematic.

[JSWolf]

Originally Posted by Rev. Bob

BTW, it's really helpful if you actually tell me what you're doing instead of being this vague. This plugin has a lot of features and switches and multiple modes of use; hows about at least a hint of how I can replicate the behavior?

I mean, after I got the EPUB to the point where it would validate, I ran my standard cleanup pass on the file and got no errors. That means that either the validation fixes resolved the issue (in which case, not a bug) or you're doing something I normally don't.

This doesn't mean I'll make the change you're after - only that if I elected to investigate, I'd need more information than you've provided. In addition, that also means the behavior's in code I had nothing to do with, which would make changes more problematic.

A data URI is a way to display an image without having to have a graphic file.

https://css-tricks.com/data-uris/

That link will tell you about data URIs. Data URIs do work with ADE and Calibre.

[botmtl]

Originally Posted by Rev. Bob

BTW, it's really helpful if you actually tell me what you're doing instead of being this vague.

Sorry! I totally forgot to to indicate the steps to reproduce the problem. I was using "remove broken image pages" on some epubs and noticed that images with data-uris are considered invalid. I then made a silly epub with the type of images that get destroyed with the "remove broken image pages" switch on.

This is what I get when I run modify-epub with that switch on:
Looking for html pages containing only broken image links
- Broken image link: "data:image/png;base64,iVBORw0KGg[...] in: ModifyEpubBug2.html
- Broken image link: "data:image/png;base64,iVBORw0KGgoAA in: ModifyEpubBug.html

[Rev. Bob]

Originally Posted by JSWolf

A data URI is a way to display an image without having to have a graphic file.

No kidding!

I already KNEW that part; I remember base64 encoding from back when it supplanted UUEncoding on Usenet. Doesn't help one bit in diagnosing the plugin problem reported, though.

botmtl - I suspect from your description that "data:" isn't getting recognized as a protocol and is instead being treated as a filename... which, of course, doesn't exist and thus gets flagged as broken. It seems the official EPUB validation software has had some issues with that, in the context of pages saved in subdirectories.

Anyway, if that's the case, this may be a fairly simple fix - just treat a string beginning with those five characters as automagically okay. (True, that won't catch malformed/corrupted data streams, but I think that's an acceptable risk.) I'm not at the computer right now, though, so it'll have to wait. I'll also have to watch out for that subfolder issue, but it really depends on the guts of the image link validation routine and where the path gets added to the specified "filename."

[botmtl]

I have a suggestion for a fix if you like it. In container.py (get_page_image_names), line 784, instead of

images = XPath('//h:img[@src]')(data)

use

images = XPath('//h:img[not(starts-with(@src,"data:"))]')(data)

?

**disclaimer: I'm about as good with python as I am programming LISP or assembly.
**disclaimer: I haven't actually tried it in the plugin, but I known the XPath is valid at the very least.

[BetterRed]

If anyone is looking for the discussion on greyscale cover images, it's here Send/Save Cover in Greyscale

BR

[senhal]

Originally Posted by kiwidude

Remove non dc: metadata from manifest

Is it possible to remove them also from html files?
Thanks!

[JSWolf]

Originally Posted by senhal

Is it possible to remove them also from html files?
Thanks!

Try it and see. There is an option to do this. But I don't know if it is for OPF only. I've never used that option.

[DNSB]

Originally Posted by Rev. Bob

It's still more efficient to actually store the image, unless we're talking about something so tiny that the overhead of creating a file at all is a significant factor. (In which case, I'll still point to the cover image as the biggest part of almost any ebook and stick to my guns.)

Is the book in question self-published or traditionally published?

I think that's a much smaller population than you think it is. It may seem larger from the inside, but I doubt it's a drop in the bucket as a proportion of Ebook-Reading People.

I do download quite a few bits off the web to convert into epubs and I have seen a page that used data URIs for images exactly once a couple of years back when I was downloading a web short story to convert to an epub. Generating an image file from the data and referencing that file reduced the final epub size by a negligible amount -- the recurring data strings compressed quite happily -- but it made it a lot easier for me to read the code without an multi-line block of gibberish every time a horizontal rule was applied which was about every second paragraph. The author was also excessively fond of em dashes and ellipses.

[AlanHK]

Originally Posted by DNSB

an multi-line block of gibberish every time a horizontal rule was applied which was about every second paragraph.

Presumably the same image? If so, could have put it in CSS and had it once in the CSS file. I just found a retail book that did that; I think it was the publisher's logo. I can see the logic in that if you're making a series of books. Just copy one single CSS file and the images it needs can't be misplaced (unless some plugin erases them).

[DNSB]

Originally Posted by AlanHK

Presumably the same image? If so, could have put it in CSS and had it once in the CSS file. I just found a retail book that did that; I think it was the publisher's logo. I can see the logic in that if you're making a series of books. Just copy one single CSS file and the images it needs can't be misplaced (unless some plugin erases them).

There were 4 horizontal rule blobs with negligible changes in the width of the rule plus minor changes in the overall size of the image. I replaced the 4 with a single image and haven't really noticed any difference in my reading enjoyment.

[Barb-B]

Praise and gratitude for this fabulous tool.

It would be very nice to have an option to simply overwrite the existing epub file, rather than create a backup with the ".ORIGINAL_EPUB" extension. Since I'm running the routine immediately after importing the book to Calibre, I already have a backup in the original imported file. If the results should go disastrously wrong (though they never do), I can always re-import the original. Would save the extra step of having to delete the backup copies every single time.

Even though I already made a $ contribution some time ago, I would seriously consider doing it again if my wish were granted.

Thanks!