HTML modes create "dirty" HTML

[nphilipp]

Kajiki's HTML modes default to produce "dirty" HTML, e.g. unclosed

tags. There's a hacky way to improve this, but it still doesn't use shortened end tags for e.g. <meta ... />, <img ... />, plays loose with inserting whitespace (e.g. two blank lines at the beginning of rendered documents):

# make kajiki render nicer HTML
from kajiki import html_utils
html_utils.HTML_OPTIONAL_END_TAGS.clear()

You can say many things about Genshi, but the (X)HTML it renders is beautiful. I realize that Kajiki uses a totally different way to create the templates which might make "staying close to the original template" difficult, but it should at least preserve closed and shortened end tags.

[amol-]

In case XML syntax is desired shouldn't user just use XML mode? HTML allows for unclosed tags and actually deprecates self closing tags, but if I remember correctly kajiki XML mode does actually what you want to achieve.

[nphilipp]

There are a couple of things that don't really cope well with the differences between HTML and XML (when delivered with an XML content type to the browser), e.g. some third-party Javascript modules expect tags to be uppercase in the DOM.

HTML5 explicitly allows self-closing void start tags, so I guess I'm in the clear there (except maybe for the suggested shortcut to using FontAwesome icons, i.e. a self-closing <i ... /> tag, but there's a licit alternative in <embed ... />).

Is there a way to use Kajiki XML mode in Turbogears and have it delivered as text/html? If so, I haven't found the way to do it.

[amol-]

I'm not saying that self closed tags are wrong, jus that they are not required and they are actually just ignored for void elements, so kajiki is doing the right thing in HTML mode as those tags are self closing in any case.

Regarding how to switch kajiki to xml mode setting in turbogears templating.kajiki.force_mode = xml in your app_cfg.py should do what you are looking for. (See http://turbogears.readthedocs.io/en/latest/reference/config-options.html#tg.renderers.kajiki.KajikiRenderer) hope that helps!

[nphilipp]

They're not required for HTML, but they're good if you want the rendered content not make stuff expecting XML go belly up (say the XML mode of your editor trying to balance out tags, or indent things intelligently 😉). I'm not expecting Kajiki to self-close all void start tags, only to preserve it if I do it in the template.

I'll try what you suggested about forcing Kajiki to use XML mode in TG, but as I said above, it needs to be delivered with content type text/html so client-side doesn't have the changes in DOM etc. that make other things (JS libraries) break.

[nphilipp]

So, setting base_config['templating.kajiki.force_mode'] = 'xml' rendered void elements with self-closing start tags (not so with e.g. <script> but that is expected with what the HTML standard says about it), and delivers the page as text/html. That works for me and trims down this issue to "playing loose with whitespace" but that's not a huge problem (unless you optimize for bytes sent ;).

That leaves me only issue #21 before I can make the switch :).