Re: [xml-dev] genx - consensus items

On Jan 23, 2004, at 10:31 AM, Karl Waclawek wrote:
>> In genx, the only significant string comparison is checking for dupe
>> attributes (I think; haven't written the code yet).  Since you only
>> check each one once, the cost of doing an intern() (what I assume you
>> mean by atomize) is probably higher than just doing the string
>> comparison. -Tim
>
> What algorithm are you using?

Well, since I want genx to generate canonical XML, I'm going to have to 
sort the atttributes anyhow, and avoiding premature optimization, I'm 
just gonna collect 'em and do a qsort() and then do an O(N)*strcmp run 
through.  I can think of lots of clever ways to optimize this if it 
produces measurable overhead. -Tim