Cleaning a stylesheet of unused styles

[DiapDealer]

XML doesn't allow orphaned classes. So yes, it does break the XML rules and thus breaks the ePub rules.

I think I'm going to doubt you on this particular transitive property, rule-break assertion. I could be wrong, but I'm going out on a limb and saying "No." Final Answer.

[theducks]

Originally Posted by JSWolf

XML doesn't allow orphaned classes. So yes, it does break the XML rules and thus breaks the ePub rules. But, unused styles are not breaking anything.

AFAIK the rule is that unsupported styles must simply be ignored.
(ADE is great at that , they Ignore the whole stylesheet)

I will take that as the style has not been properly defined (orphaned) should be ignored, thus validate.

[meme]

Originally Posted by Hitch

Well...the iterations would be fairly massive, I'd think.

I mean, you'd have to (I don't program, so--those with real brains, leap in) aggregate ALL the styles and elements; probably doing so from the top-down, so most-used to least-used makes the most sense; search for all instances of "p," first, and then all classes used in the ePUB--and then do a string-match in the CSS...does that sound right, ye olden programmers? I mean, it seems more efficient to do it by element and then by class, but sometimes, what seems elementary (sorry, sorry, you guys KNOW I can't resist bad punnage!) isn't.

We use something SOMEWHAT like this in two places; we have a PERL script and an NTPro script that both check for this--but it's in the HTML. I don't know how we'd adopt it to run inside Sigil, unless we took the additional steps of copying the file, merging it, and exploding it, but even so...it would be a SLICK addition in Sigil. I would infinitely rather the "used in the ePUB but not in the stylesheet" functionality than simply cleaning up the Stylesheet. I know some folks need that--and it would be nice fudge topping--but it's not as needed as the inverse. We have a House CSS that we use, so cleanup isn't as important for us, in the CSS itself...but I think everyone would love the missing classes thing.

JMHO,
Hitch

I don't know CSS that well, yet, but from what I've seen, the selectors you can use in CSS are very flexible, making it hard to parse. Unless EPUB only uses a subset - something I still need to find out. There are parsers out there, and it may be that one of them might help.

You would have to go through every element and then compare it against every selector to see if there is a match. For finding used but undefined entries you could simplify things by restricting what's checked (ie checking only class attributes and seeing if they match a selector. Checking just for the name doesn't cover everything as a class could only be defined for certain tags...). I created an issue to track this - just reporting on classes used in the HTML.

[Jellby]

Originally Posted by theducks

AFAIK the rule is that unsupported styles must simply be ignored.
(ADE is great at that , they Ignore the whole stylesheet)

I will take that as the style has not been properly defined (orphaned) should be ignored, thus validate.

Aren't you mixing CSS and XHTML?

I think in this thread we are talking about "unused styles" when they are defined in the CSS stylesheet, but they never actually apply to anything. And we are using "orphaned classes" when a class is used in the XHTML file, but there's no defined style that applies to this class.

The rule about ignoring unknown/unsupported styles is, as far as I know, a CSS one, and it refers to properties and values, so if you use "margin_top" instead of "margin-top", it is unknown and ignored.

Again, as far as I know, there is no rule that says that an orphaned class (a class used in XHTML but with no specific style defined for it) is wrong. Maybe in XML it is, but in XHTML? How would you apply a custom stylesheet to any document if you don't know all the class names used? (An XHTML could use <p> for its paragraphs, another <p class="text"> and another <p class="normal">, all of them should work fine if you just define a style for "p" in the CSS, no need for "p.text" and "p.normal".)

[Hitch]

Originally Posted by meme

I don't know CSS that well, yet, but from what I've seen, the selectors you can use in CSS are very flexible, making it hard to parse. Unless EPUB only uses a subset - something I still need to find out. There are parsers out there, and it may be that one of them might help.

You would have to go through every element and then compare it against every selector to see if there is a match. For finding used but undefined entries you could simplify things by restricting what's checked (ie checking only class attributes and seeing if they match a selector. Checking just for the name doesn't cover everything as a class could only be defined for certain tags...). I created an issue to track this - just reporting on classes used in the HTML.

Yup, that's the problem. There's no limit, literally, to the number of classes you can have for an element. I mean, it's the equivalent of not knowing what you don't know.

It's not the elements...it's the limitless number of classes that a designer can create, particularly if they're on the fly, or, worse (for us, anyway), we get an INDD file. Man...hard for me to say enough about that.

Hitch