[Webkit-unassigned] [Bug 41238] RegExp performance slow on Dromaeo benchmark

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=41238





--- Comment #5 from Gavin Barraclough <barraclough at apple.com>  2010-06-28 16:04:59 PST ---
(From update of attachment 59949)
Hey Michael,

As discussed:

* There is a tab in the changelog,
* Please remove the initialization of m_lastMatchStr & the call to resize on m_lastOVector from the RegExp constructor.
* Please remove redundant braces & else in the code checking the cache.
* Re the UString change, instrumenting & checking if a comparison of the rep()s rather than a compare of data() sounded like a good plan.

As discussed, this code can be made much cleaner by both removing the output param & never resizing the array to zero – but I agree with you that this would be better done as a separate patch.

Also, one other thing I missed,

> Are all the (Vector<int, 32>&) casts necessary? - could you check if these are still needed? – if the coding style requires a static_cast<> here instead of a C-style case (which could silently be doing nasties like casting away constness).

Oh, a second last thought! - I'm tempted to suggest ASSERTs in the switch statements, where you have removed the checks of size – these may just make it more obvious to someone else coming to the code why the lack of check is safe.  What do you think?

r-ing this for now, please attach a new patch with the fixes above.
cheers, G.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.