Ignore empty title attributes

[winniehell]

Fixes #199.

The fix is not complete though as some tests still fail. This is because conversion is not reversible for empty title attributes:

$ kramdown -ihtml -okramdown <<< '<p><img src="a" alt="b" title="" /></p>'
![b](a)

but

$ kramdown -ikramdown -ohtml <<< '![b](a)'
<p><img src="a" alt="b" /></p>

and therefore assert_equal fails.

Any suggestions? Should I avoid these tests?

[gettalong]

Thanks for the pull request!

Before I can merge it:

• The #convert_title method should be renamed to #prepare_title or something similar because all #convert_NAME methods indicate the conversion of an kramdown element.
• If you want to test HTML input, please create a new file with the extension ".htmlinput". At first I didn't know why you added the tests because the normal conversion direction is from text to HTML.

Thank you!

[winniehell]

Sorry, i don't get the htmlinput mechanism. 😞

The bugfix does not touch the html-to-html conversion but only the html-to-kramdown conversion. I do not understand how i would write the test such that the backwards conversion (kramdown-to-html) is skipped.

[gettalong]

The test cases were written originally written to be in the direction text -> HTML. However, later on other directions were added, like text -> kramdown -> HTML, HTML -> kramdown -> HTML, ...

Normally, when testing the HTML -> kramdown direction, the .text file is compared against the converted .html file. However, sometimes it is explicitly wanted to only test the HTML -> kramdown direction and not the other way around (like it is in this pull request). To facilitate this, you can create a file with the extension .htmlinput and the output file with .text. As the test code sees that there is no .html associated, the kramdown -> HTML direction is not tested.

[winniehell]

Seems like I finally fixed it... ✨ 🚀

[gettalong]

Thanks for the fix! I have squashed the commits and attributed the final one to you.

[winniehell]

Thank you for merging! 👍