css - Is !important bad for performance?

Question

I hate them, it defies the cascading nature of CSS, and if you don't use them with care you end up in a loop of adding more !important.

But I want to know are they bad for performance?

EDIT
From the (fast) replies I can conclude it won't have a (significant) impact on performance. But it's nice to know, even if it's just as an extra argument for discouraging others ;).

EDIT 2
BoltClock pointed out that if there are 2 !important declarations the specs says it will pick the most specific one.

Answer 1

It shouldn't have any discernible effects on performance. Seeing Firefox's CSS parser at /source/layout/style/nsCSSDataBlock.cpp#572 and I think that is the relevant routine, handling overwriting of CSS rules.

It just seems to be a simple check for "important".

  if (aIsImportant) {
    if (!HasImportantBit(aPropID))
      changed = PR_TRUE;
    SetImportantBit(aPropID);
  } else {
    // ...
  }

Also, comments at source/layout/style/nsCSSDataBlock.h#219

    /**
     * Transfer the state for |aPropID| (which may be a shorthand)
     * from |aFromBlock| to this block.  The property being transferred
     * is !important if |aIsImportant| is true, and should replace an
     * existing !important property regardless of its own importance
     * if |aOverrideImportant| is true.
     * 
     * ...
     */

---

1. Firefox uses a top down parser written manually. In both cases each CSS file is parsed into a StyleSheet object, each object contains CSS rules.

2. Firefox then creates style context trees which contain the end values (after applying all rules in the right order)

From: http://taligarsiel.com/Projects/howbrowserswork1.htm#CSS_parsing

Now, you can easily see, in such as case with the Object Model described above, the parser can mark the rules affected by the !important easily, without much of a subsequent cost. Performance degradation is not a good argument against !important.

However, maintainability does take a hit (as other answers mentioned), which might be your only argument against them.