[Webkit-unassigned] [Bug 33222] Make text decoders and resource loaders read from segmented SharedBuffer

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=33222





--- Comment #14 from Yong Li <yong.li.webkit at gmail.com>  2010-01-06 10:15:07 PST ---
(In reply to comment #11)
> 
> This is still a quite-inefficient algorithm for creating a string out of
> multiple decoded chunks; any loop that appends to a String is suboptimal
> because making a new String by appending two strings *always* involves
> allocating a fresh buffer each time and copying all the data; this inefficiency
> might make some real world cases pathologically worse if we find ourselves with
> multiple segments. It might actually be more efficient to do the dumb thing the
> old code was doing, and concatenate them all, just because the decoded data
> would be dealt with all at once even though the encoded data ends up being
> copied.
> 
> I think at some point we will want to change this. Even now I think we could do
> better with Vector<UChar>.
> 
> Maybe we want to get rid of the use of String in the interface to decoders
> eventually?
> 
> I am conflicted about the patch. It seems like a good idea in theory, but in
> practice it seems that this patch claiming to speed things up might always end
> up slowing things down instead. And might cause us to use more memory too.
> 
> > +    String decode(const SharedBuffer& data);
> 
> I don't think the argument name is needed for clarity here so normally we would
> leave it out.
> 
> I am going to say r=me but I do have serious reservations. I'd like to hear
> more concrete evidence about why this is a helpful change to make.

You're right, this patch is probably not helpful at all. I'll leave it here. I
don't like the way that webkit uses String to store and parse scripts.
SegmentedString is a good concept, but only html parser is using it.

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