In 3.7.2, spaces before <em> or <i> tags get removed after conversion to KEPUB

[hengyu95]

Bug Checklist

These items are mandatory. If you need help finding this information submit the
bug report with as much completed as you can and ask for help finding the rest.

• I am using the latest version of calibre to report this bug, which is:
• I am using an official calibre release, not one from a third party (e.g.
  your Linux distro, Flatpak, Chocolatey package, Homebrew, etc.)
• I am using the latest version of this plugin, which is:
• My operating system is Windows 11
• I have included the full, complete, unmodified debug log from calibre
  • Directions for getting the debug log are under the "Logs" header below.
• I have translated the text in any screenshots and logs to English, or all
  screenshots and logs included are in English.

These items are optional. Fill in as much of them as possible. If something is
not applicable to your bug report, note that.

• I have installed the Scramble Epub plugin (see
  https://www.mobileread.com/forums/showthread.php?t=267998) and will attach
  a scrambled copy of the book I'm having problems with (attach a file by
  dragging and dropping onto the Github editor).
  • If this is a conversion bug, I will also attach a scrambled copy of
    the converted book.
• The path to my calibre library or to a book in my calibre library has
  non-ASCII characters: yes/no
• If I am using Windows 10, I (have/have not) enabled Windows' beta support
  for Unicode (see
  https://www.mobileread.com/forums/showpost.php?p=3988195&postcount=2052)
• If I am using Windows 10, does this bug happens with beta Unicode support
  both enabled and disabled, only when enabled, or only when disabled?

Describe the bug

In the EPUB to KEPUB conversion, spaces before italics all get removed.

Steps to Reproduce

Convert any EPUB to KEPUB via sending it to Kobo

Expected behavior

Spaces before italics remain

Actual behaviour

Spaces before italics are deleted. This is corroborated by many other users on the mobileread forums.

Screenshots

Original EPUB

Converted KEPUB

Logs

N/A, conversion itself "works"

Additional context

Add any other information you think might be helpful.

[chocolatechipcats]

Also appears to affect some other tags: https://www.mobileread.com/forums/showpost.php?p=4474070&postcount=3159

[hungrytoast7]

The space between the chapter number and the chapter title is also missing after converting the file to Kepub. "Chapter 3: The Sun is Rising" appears as "Chapter 3:The Sun is Rising" in both the table of contents and the header. This issue may be related to the < a id="page-42"/ > tag.

[gabri25ele]

I confirm the problem with italics, I'm currently back to 3.7.0

[GAntiko]

I confirm the problem with italics, but it happens also with bold. I tested it with 3.7.2, Calibre 7.22 and Kobo Libra Colour

[onecrayon]

If this is verified to be working in 3.7.0 and failing in 3.7.2, there's actually not a lot of changed lines that could be the issue. Has anyone tested to see if the problem exists in 3.7.1? Because when I look at the diff between 3.7.2 and 3.7.0 there are only two changes that have any likelihood of triggering this difference to my eye; one was committed as part of 3.7.1 and the other as part of 3.7.2:

• The 3.7.1 change: e30f7a4#diff-e95207e1df7f5c5850ab1aa4833068c58447916ddf8031042d7a84f92c8f1501R609 is skipping over text that is nothing but whitespace
• The 3.7.2 change: e30f7a4#diff-e95207e1df7f5c5850ab1aa4833068c58447916ddf8031042d7a84f92c8f1501R609

[jadehawk]

If this is verified to be working in 3.7.0 and failing in 3.7.2, there's actually not a lot of changed lines that could be the issue. Has anyone tested to see if the problem exists in 3.7.1? Because when I look at the diff between 3.7.2 and 3.7.0 there are only two changes that have any likelihood of triggering this difference to my eye; one was committed as part of 3.7.1 and the other as part of 3.7.2:

• The 3.7.1 change: e30f7a4#diff-e95207e1df7f5c5850ab1aa4833068c58447916ddf8031042d7a84f92c8f1501R609 is skipping over text that is nothing but whitespace
• The 3.7.2 change: e30f7a4#diff-e95207e1df7f5c5850ab1aa4833068c58447916ddf8031042d7a84f92c8f1501R609

I downgraded to 3.7.1 on my KLC and resend the book in currently reading. I do not see the missing space anymore when Italics is involved. This was the most noticeable issue for me with 3.7.2 so I can't comment on nothing else. staying on 3.7.1 unless I see anything else weird in the books I read.

Either way, Thank you for the plugging.

[dragid10]

The bug was definitely introduced in commit: https://github.com/jgoguen/calibre-kobo-driver/commits/e30f7a4

I did a git bisect and this was the only version with the errored spacing

[Boswell-Scrubbs]

I have this problem on my Kobo Libra 2 when using 3.7.2. Downgraded to 3.7.1 and the problem went away.

P.S. Thank you for the awesome plugin.

[dragid10]

Looking at it even further, it seems to be caused by the new regex used to fix the highlight breaks between sentences in commit: 8638dc0.

I added this paragraph to the page_github_106.html (This is just a random excerpt from one of the books I have)

<p class="nonindent">
		“Even in Russia, <i class="calibre2">women</i> are”
</p>

and then added a bunch of print statements in container.py under the _append_kobo_spans_from_text method.

The proper text should read:

Even in Russia, women are

but it instead reads:

Even in Russia,women are

It seems that this new regex is causing spaces before indentation tags, to be missed (or its causing space after regular text, but before an indentation tag, to be missed).

Edit: Okay I've re-checked, I think this specific code change is actually what is causing the problem. Ignoring the space between punctuation and indentation perhaps? But I believe that line was added to work with the new regex mentioned above. Unfortunately I struggled for about 4 hours with trying to set up my development environment to even try to test this. I'd love to help submit a fix that isn't just removing that change, but the calibre dev environment setup was a bit frustrating for me 😅